Another piece in this section, on the same terms as the rest: a named public dataset for every claim, every disparity stated with its denominator, the innocent explanation weighed before the damning one, and a fix a real office could carry out.
This one started as a piece about a missing dataset. The dataset turned out to exist.
An earlier article on this site, the cameras that follow you home, worked what Minnesota's license-plate-reader law permits agencies to do with the data — who can search it, who can share it, how long it survives. This piece asks a narrower question: where are the cameras. Not what they capture. Where they physically stand, and who lives on that block.
The law says the locations are public. It says so twice.
Minn. Stat. § 13.824 was enacted in 2015 and amended twice since — subdivision 6 in the 2021 first special session, and subdivision 1 plus a new subdivision 2a in 2024. I checked the Revisor's table of statutes affected by session laws across every session from 1994 forward: four records, all accounted for. Nothing in the 2025 or 2026 sessions changed this section. The posted text is the operative text.
Two subdivisions put camera locations in public view, and they do different jobs.
Subdivision 5 is a duty on the agency. The public use log every department must keep has to include, for a fixed reader, "the location at which the reader actively collected data and is installed and used." Then paragraph (b) goes further than most people realize:
The law enforcement agency must maintain a list of the current and previous locations, including dates at those locations, of any fixed stationary automated license plate readers or other surveillance devices with automated license plate reader capability used by the agency.
Current and previous. With dates. Then: "The agency's list must be accessible to the public, unless the agency determines that the data are security information as provided in section 13.37, subdivision 2."
Subdivision 8 is a duty on the state. Within ten days of installing or beginning to use a reader, an agency must notify the Bureau of Criminal Apprehension "of that installation or use and of any fixed location of a stationary automated license plate reader." And then:
The Bureau of Criminal Apprehension must maintain a list of law enforcement agencies using automated license plate readers or other surveillance devices with automated license plate reader capability, including locations of any fixed stationary automated license plate readers or other devices.
Subject to the security-information exception, "this list is accessible to the public and must be available on the bureau's website."
That is not a records-request right. It is an affirmative publication mandate that names the website.
Now the distinction that matters, because two different articles could be written here and only one of them is true. The audit provision is not a publication provision. Subdivision 6 requires each agency to arrange an independent biennial audit, says "The results of the audit are public", and requires a summary report to go to the commissioner of administration, to the chairs and ranking minority members of the House and Senate committees with jurisdiction over data practices and public safety, and to the Legislative Commission on Data Practices and Personal Data Privacy within 30 days of completion. Public, and delivered to named recipients — but nothing in subdivision 6 requires anyone to post an audit anywhere. Locations must go on a website; audits must merely be available to whoever asks the right office. So "the law requires disclosure and nobody complies" is the wrong story. The law requires it, and the location half of it is being done.
What is actually on the website
I pulled the BCA's list on September 10, 2026 and parsed it out of the page's own data payload rather than reading it by eye. Every count below is that day's. The bureau revises this list without announcing that it has — there is no revision date on the page, as I set out below — so a reader checking later should expect a different number and should treat these as a dated snapshot rather than a standing total.
| Agencies listed | 116 |
| Agencies naming at least one fixed reader location | 87 |
| Named fixed-reader location lines | 747 |
| Agencies disclosing only that locations are withheld | 4 |
| Locations covered by those withholdings | 17 |
| Agencies reporting no fixed readers at all | 25 |
The four agencies invoking the security-information carve-out say so on the face of the list, and three of them say how many locations it covers: the Crow Wing County Sheriff's Office (4), the Robbinsdale Police Department (2), and the University of Minnesota Police – Twin Cities (10). The fourth, the Edina Police Department, writes "Location withheld" in the singular and gives no number, so I have counted it as one. That is the law working as designed. Section 13.37 defines security information as data whose disclosure the responsible authority determines "would be likely to substantially jeopardize the security of information, possessions, individuals or property against theft, tampering, improper use, attempted escape, illegal disclosure, trespass, or physical injury." Section 13.824 then adds a check: the determination "is subject to in-camera judicial review, as provided in section 13.08, subdivision 4." Seventeen withheld locations against 747 published ones is not a loophole swallowing the rule.
The twenty-five agencies reporting no fixed readers include the Minnesota State Patrol and the St. Paul Police Department. Their readers move: all 116 agencies, these twenty-five included, answer "N/A" to the list's second question about other surveillance devices with reader capability. The list does not say how many mobile readers any of them run. Hold the St. Paul entry in mind.
The denominator trap, and how it nearly wrote the wrong article
Eighty-eight of the agencies are police departments whose jurisdiction is a single Minnesota city, which makes the denominator unambiguous. (I set aside the county sheriffs, the State Patrol, the airport, the university, the tribal department, and the three multi-city joint agencies — twenty-eight entries in all — because their readers sit in places their name does not identify.) Those 88 cities held 3,104,024 people in the 2020 Census and name 560 fixed readers between them: 1.80 readers per 10,000 residents.
Sort those cities by the share of residents who are non-Hispanic white alone, and you get a table that looks like a scandal.
| City % non-Hispanic white | Cities | Population | Readers | Per 10,000 |
|---|---|---|---|---|
| Under 50% | 3 | 431,787 | 15 | 0.35 |
| 50.0–64.9% | 11 | 699,356 | 102 | 1.46 |
| 65.0–76.2% | 20 | 934,543 | 177 | 1.89 |
| 76.3–84.9% | 26 | 753,140 | 138 | 1.83 |
| 85% and up | 28 | 285,198 | 128 | 4.49 |
Minnesota as a whole was 76.3 percent non-Hispanic white in 2020. Read that table quickly and you conclude that the whitest cities in the state are thirteen times as surveilled as the most diverse ones, which would be a strange finding for anybody expecting the opposite, and a story either way.
It is an artifact. Sort the same 88 cities by population instead:
| City population | Cities | Population | Readers | Per 10,000 |
|---|---|---|---|---|
| Under 5,000 | 12 | 33,086 | 23 | 6.95 |
| 5,000–14,999 | 23 | 216,644 | 97 | 4.48 |
| 15,000–34,999 | 28 | 652,479 | 187 | 2.87 |
| 35,000–74,999 | 17 | 919,649 | 171 | 1.86 |
| 75,000 and up | 8 | 1,282,166 | 82 | 0.64 |
Cameras per person falls almost perfectly as cities get bigger, and Minnesota's biggest cities are its most diverse. Hill City has 613 residents and two readers, which is 32.63 per 10,000 — the highest rate among these 88 departments. That is a fact about the denominator, not about Hill City. A town buys a camera for each way into town whether it has six hundred residents or sixty thousand, and a per-capita rate computed on 613 people will top any list you build.
So control for size. Among the 45 cities between 15,000 and 74,999 residents — 1,572,128 people, the band where most of these departments actually sit:
| City % non-Hispanic white | Cities | Population | Readers | Per 10,000 |
|---|---|---|---|---|
| Under 65% | 10 | 290,820 | 83 | 2.85 |
| 65.0–76.2% | 12 | 531,147 | 126 | 2.37 |
| 76.3–84.9% | 17 | 622,808 | 95 | 1.53 |
| 85% and up | 6 | 127,353 | 54 | 4.24 |
The gradient is gone. No monotonic relationship, the most diverse band and the whitest band both above the middle, and the ranking inside the band is driven by which council voted for a contract: Maple Grove (77.3 percent white, 70,253 people) has zero, Blaine (72.8 percent, 70,222 people) has thirty-two. Columbia Heights and Golden Valley have none. Edina withheld its one. At the municipal level, Minnesota's fixed camera map is a map of purchasing decisions, not of demography. Statewide the same conclusion holds at the census-tract level: of the 747 lines, 505 resolved to a point, landing in 322 of Minnesota's 1,505 tracts, home to 1,339,687 people or 23.5 percent of the state — and those tracts are 72.2 percent non-Hispanic white against 76.3 percent statewide. A four-point difference, in the direction urbanization alone would predict.
One illustration of purchasing over targeting: the Ramsey County Sheriff's Office lists 33 fixed readers, and 19 of them are in North Oaks — a city of 5,272 people, 85.5 percent non-Hispanic white. Nine are in St. Paul, a city of 311,527. More than half of a county sheriff's fixed camera network sits inside a suburb holding 0.95 percent of Ramsey County's 552,352 residents.
Inside one city, the picture changes
Municipal boundaries are the wrong unit for the question I actually want answered. Whether Blaine buys cameras is a budget question. Where Blaine puts them is a siting question, and only a handful of departments list enough locations to ask it.
Minneapolis lists 30. Twenty-nine geocoded; one — "4th Street SE and I-35 West" — came back matched to "N 4th St & 35th Ave N" in North Minneapolis, which is a different place entirely, so I dropped it rather than let a bad match push the result my way. That leaves 28 readers, sitting in 23 distinct census tracts.
Those 23 tracts held 84,214 people in 2020, 19.6 percent of the city. They are 39.0 percent non-Hispanic white. Minneapolis is 58.0 percent. Weighted by the number of readers rather than by tract, the figure is 43.2 percent.
Run the identical method on the Ramsey County Sheriff's nine readers inside St. Paul and it produces nothing: host tracts 47.6 percent non-Hispanic white, city 48.8 percent. Same state, same vintage, same code, no gap. Whatever is happening in Minneapolis is not an artifact of how I am counting.
The innocent explanation, tested twice
Readers go on busy roads. That is the first thing anyone who has actually placed one will tell you, and in Minneapolis it is verifiably true.
MnDOT publishes annual average daily traffic as a segment layer. There are 1,235 counted segments inside Minneapolis; the median carries 5,100 vehicles a day, the 75th percentile 8,974, the 90th percentile 15,838. Every one of the 28 readers sits within 75 metres of a counted segment. The median volume at a reader location is 13,644 vehicles a day — the 86th percentile of every counted street in the city. Twenty-seven of the 28 sit on an interstate, a county state-aid highway, or a municipal state-aid street; the twenty-eighth is a city street beside the University. West Broadway, Lake Street, Franklin, Penn, University, Washington, Lyndale, Chicago, Cedar. Nobody put a camera on a cul-de-sac.
And in Minneapolis, the arterials run through the neighborhoods the highway program cut up eighty years ago. Traffic volume and racial composition are not independent variables in this city. So the test is not whether readers are on arterials — they are — but whether, among comparable arterials, the camera went to one neighborhood rather than another.
Take every Minneapolis tract carrying at least one street of 10,000 vehicles a day or more. There are 85 of them.
| Arterial tracts (AADT ≥ 10,000) | Tracts | Population | % non-Hispanic white |
|---|---|---|---|
| With a fixed reader | 20 | 75,317 | 39.9% |
| Without a fixed reader | 65 | 240,334 | 61.9% |
Twenty-two points, with the road held constant. At the segment level the same thing: the 31 arterial segments with a reader within 75 metres sit in tracts averaging 41.6 percent non-Hispanic white; the 232 without average 57.5 percent. The arterial rule is real and it is not sufficient.
Then add the second explanation, which is the one the departments would actually offer. I pulled the City of Minneapolis's own 2025 incident layer — every reported homicide, criminal sexual conduct, robbery, aggravated and domestic assault, motor vehicle theft, and carjacking — 8,731 records, all of them placed inside a tract.
Among those same 85 arterial tracts, the ones with a reader had a median 31.0 such incidents per 1,000 residents in 2025. The ones without had 17.5. The reader tracts are not merely less white; they are where the reported crime is.
So hold crime constant too. Rank the 85 arterial tracts by 2025 incident rate and take the top 20 — every tract above 29.8 incidents per 1,000 residents:
| Top 20 arterial tracts by 2025 incident rate | Tracts | Population | % non-Hispanic white | Median rate |
|---|---|---|---|---|
| With a fixed reader | 11 | 41,598 | 34.4% | 42.4 |
| Without a fixed reader | 9 | 32,779 | 34.2% | 38.1 |
Two tenths of a percentage point. The gap is gone — and I should say that the two tenths themselves are not the finding. A verification pass reran this with five different rules for assigning a road segment to a tract; the arterial gap held at 21 to 24 points every time, while this top-twenty comparison moved around within a few points of zero in both directions. What is robust is that the gap does not survive the crime control. The exact margin is noise. Nine of the twenty highest-incident arterial tracts in Minneapolis have no fixed camera at all, and demographically they are indistinguishable from the eleven that do. Widen the comparison to the top 40 and a gap reopens — 38.7 percent against 47.3 — but by then the group without cameras carries a median 26.4 incidents per 1,000 against 35.1 for the group with them, so the comparison has gone back to measuring crime.
The honest finding: fixed camera placement in Minneapolis is explained by traffic volume and reported crime. Once both are held constant, the racial composition of the neighborhood carries no additional signal. The people living under these cameras are disproportionately Black, Native, and immigrant Minnesotans. That is true, and worth saying plainly. The mechanism is where reported crime concentrates, not a demographic filter applied to a map.
What cuts against me
The targeting story does not survive the data, and I went looking for it. I expected the statewide table to show cameras clustering in the most diverse cities; it showed the reverse. I expected the Minneapolis gap to survive the controls; it did not. Anyone selling a clean Minnesota story about surveillance following race should be asked for the denominator.
"Explained by reported crime" is not the same as justified. Reported crime is partly a function of where officers already are and who feels able to call 911. Controlling for it moves the question one step upstream rather than closing it. The cameras follow the incident map; I cannot show that the incident map is a neutral measurement, and I am not going to pretend those are the same sentence.
My within-city sample is 28 readers in one city. Minneapolis and St. Paul are the only cities on this list big enough and varied enough for a tract-level comparison, and St. Paul's nine fixed readers belong to the county rather than the city. One city is a finding, not a state.
And the cameras solve cases. Nothing above argues that no reader should exist. The argument is about who decides where they go, and whether that decision is reviewable by the people driving past.
What the list cannot tell you
I could run all of the above because Minnesota publishes this. I could barely run it because of how Minnesota publishes it. The 747 lines are free text, typed by 87 different agencies with no shared schema, and it shows.
- No coordinates. 505 of 747 lines resolved to a point through the Census geocoder; 242 did not. Some of that is the geocoder, which handles address ranges better than intersections — but a state-published surveillance inventory should not require geocoding at all.
- No counts. Some agencies use a parenthetical number as a device count — Deephaven's "20225 Cottagewood Road (4)", Roseville's "1595 Highway 36 West (3)" — and St. Cloud writes it out as "Four devices at Ninth Avenue South and University Drive South". But the Ramsey County Sheriff uses parentheses as labels: "Maycomb Road entry (1)", "Maycomb Road (2)", "Maycomb Road entry (3)". The same notation means two different things on the same page, so the number of fixed license-plate readers in Minnesota cannot be derived from the state's list of them. 747 is a count of lines.
- No direction field. 229 of 747 lines record a direction of travel. Thirteen agencies do it on every line, 58 on none, 16 inconsistently. So "51st Avenue North and Brooklyn Boulevard, northbound" and "...southbound" are plainly two devices, while Coon Rapids' "Springbrook Drive Northwest at Coon Rapids Boulevard Northwest" appears twice, identically, and you cannot tell whether that is two cameras or one typed twice. Four lines are exact duplicates within a single agency's entry.
- No dates, and no history — even though subdivision 5(b) tells every agency to keep "current and previous locations, including dates at those locations". The BCA list carries no installation dates, no removal dates, and no publication or revision date of its own. I checked the page's underlying data for a modified timestamp; there is none. A reader cannot tell whether an entry is from last week or from 2016.
- No vehicle-mounted inventory. Twenty-five agencies report no fixed readers at all, which means their readers move. The list says nothing about how many, or how many the other 91 agencies also run on squad cars alongside the fixed ones.
- Ordinary transcription errors, unreviewed. Minnetonka lists "Bred Road and Shady Oak Road" one line below "Bren Road and Smetana Drive", places a reader at "Interstate 384", which does not exist in Minnesota, and lists "Ridgedale Center" with no street at all. Woodbury lists "Tamarack Main". Fridley lists "West More Lake Drive". Forty-one of the 747 lines are neither an address nor an intersection: "Hill Farm entrance in North Oaks", "Market Place North", "Woodlands Trails Park". The agency itself is listed as "Sauk Center Police Department"; the city is Sauk Centre.
- A jurisdictional seam. The St. Paul Police Department reports no fixed readers. The Ramsey County Sheriff's Office operates nine inside St. Paul. Both entries are accurate and complete. A St. Paul resident who checks their own police department learns nothing about the cameras on Rondo Avenue, because the list is organized by who owns the camera rather than by who drives under it.
The legislature looked at this section in the 2026 session. HF 4661 was introduced to amend § 13.824 — "updating automated license plate reader provisions to address the role of third-party service providers" — reaching subdivisions 1, 6, and 7 to define sharing and vendor access, guarantee auditors access to vendor-held query histories, and require written authorization for outside queries routed through a vendor platform. Good work, and it did not pass; the Revisor's table shows no 2026 change to this section. It also never touched subdivision 5 or subdivision 8. The disclosure provisions are the part of this statute nobody is arguing about, which is exactly why they have not improved in eleven years.
What I could not verify
I could not obtain city-level income data. The Census Bureau's American Community Survey API now requires a key, which this project does not have; I confirmed the refusal from the response header rather than assuming it, and I confirmed the network path was fine by running the Census geocoder, which answered normally. Every demographic figure above therefore comes from the 2020 Census redistricting file, which carries race and Hispanic origin but no income. Where I say a city is whiter, I mean whiter. I do not know whether it is richer.
I could not enumerate the Department of Administration's ALPR page. It sits behind a bot wall that redirects automated requests to a CAPTCHA validator. I retrieved one document from that office directly — a July 2016 Data Practices Office guidance sheet on the biennial audit review, which confirms audits began August 1, 2017 for agencies operating readers when the law took effect — but I could not inventory what else is posted there. I am not asserting that no ALPR audits have been published. I am asserting that § 13.824, subdivision 6 does not require them to be, which is a different and verifiable claim.
I have no data on vehicle-mounted readers, no installation or removal dates, no device counts, and no way to confirm the list is current. I did not verify HF 4661's authorship or committee history. The Minneapolis crime layer is reported incidents for calendar year 2025 as the city published them, with all the known limits of reported-crime data. My 75-metre matching radius between a reader point and a traffic segment is my choice, not the state's, and a different radius would move the segment-level numbers somewhat — the tract-level numbers, which drive the conclusion, do not depend on it.
What we can do
1. Make the location list a dataset instead of a web page. This requires no new obligation on any department. Subdivision 8 already commands publication; it says nothing about format. A standard reporting form with fixed fields — latitude, longitude, street description, device count, direction of travel, install date, removal date, owning agency, operating agency — and a downloadable file would turn a document that took an afternoon and a shapefile to parse into something a city council member, a reporter, or a resident could open. The BCA is already receiving these notices within ten days. It is receiving them as prose.
2. Publish the history the statute already requires. Subdivision 5(b) obligates every agency to keep current and previous locations "including dates at those locations". That history exists, by law, in ninety-one separate agency files. Route it to the same list. Without it there is no way to know whether a camera has watched an intersection for a month or for a decade.
3. Index the list by geography, not only by owner. A resident should be able to ask what cameras are on their street without first knowing which of five agencies bought them. The same records, indexed by city and county as well as by department, closes the St. Paul seam and costs nobody anything.
4. Put the audits on the same website as the locations. Subdivision 6 already makes audit results public and already sends a summary to the commissioner of administration and three sets of legislators. One sentence would post them where the location list lives. The work of producing them is already done; handing them out one request at a time is the only thing between Minnesotans and a compliance record.
5. Require a stated siting basis. Subdivision 7 already makes every search carry "a legitimate, specified, and documented law enforcement purpose" and a recorded factual basis. Placement carries no such requirement. If a department is putting a fixed reader at an intersection because of traffic volume and crash history, or because of a mapped robbery pattern, it should say which on the same form where it reports the location. That one field would have answered in a line the question that took me a traffic layer, a crime layer, and a census file to answer for a single city — and it would let every other city's residents run the check for themselves.
6. Keep the security-information exception, and keep the judicial check on it. Four agencies used it for 17 locations, and told the public they were using it and — in three of the four cases — for how many. That is what a narrow exception looks like when it works. Section 13.37 already requires an entity denying a request on security grounds to "provide a short description explaining the necessity for the classification", and § 13.824 already routes the determination to in-camera judicial review under § 13.08, subdivision 4, where an aggrieved person can recover fees and the court may impose a civil penalty of up to $1,000. Leave it alone.
Minnesota did the hard part of this in 2015, when it decided that the public gets to know where the cameras are and put the mandate in a statute with a website named in it. Eleven years later the mandate is being met and the product is a list of 747 typed sentences with a misspelled street, a nonexistent interstate, and no dates. The gap between those two facts is not a scandal. It is a form nobody ever designed.
First the facts. Then the fix.
Sources
Statute. Minn. Stat. § 13.824, "Automated license plate readers", retrieved as raw text from revisor.mn.gov on September 10, 2026 and quoted verbatim: subd. 5(a)(4) ("for a reader at a stationary or fixed location, the location at which the reader actively collected data and is installed and used"); subd. 5(b) ("The law enforcement agency must maintain a list of the current and previous locations, including dates at those locations, of any fixed stationary automated license plate readers or other surveillance devices with automated license plate reader capability used by the agency." and "The agency's list must be accessible to the public, unless the agency determines that the data are security information as provided in section 13.37, subdivision 2."); subd. 6(b)–(c) ("The results of the audit are public." and "A report summarizing the results of each audit must be provided to the commissioner of administration, to the chairs and ranking minority members of the committees of the house of representatives and the senate with jurisdiction over data practices and public safety issues, and to the Legislative Commission on Data Practices and Personal Data Privacy no later than 30 days following completion of the audit."); subd. 7(b) ("a legitimate, specified, and documented law enforcement purpose"); subd. 8(a) ("Within ten days of the installation or current use of an automated license plate reader or the integration of automated license plate reader technology into another surveillance device, a law enforcement agency must notify the Bureau of Criminal Apprehension of that installation or use and of any fixed location of a stationary automated license plate reader."); subd. 8(b) ("The Bureau of Criminal Apprehension must maintain a list of law enforcement agencies using automated license plate readers or other surveillance devices with automated license plate reader capability, including locations of any fixed stationary automated license plate readers or other devices." / "this list is accessible to the public and must be available on the bureau's website" / "A determination that the location of a reader or other device is security information is subject to in-camera judicial review, as provided in section 13.08, subdivision 4"). Minn. Stat. § 13.37, subd. 1(a) (definition of "security information") and subd. 2(b) (short description on denial), and § 13.08, subd. 4(a) (in-camera inspection, fees, and the civil penalty of up to $1,000), both retrieved raw from revisor.mn.gov the same day and quoted verbatim. Amendment history from the section's own history line (2015 c 67 s 3; 1Sp2021 c 11 art 3 s 4; 2024 c 127 art 3 s 2, 3) and from the Revisor's Statutes Changed (Table 2) queried for § 13.824 across all sessions 1994 to present, which returns four records and none in 2025 or 2026. HF 4661, 94th Legislature (2026), introduction text at revisor.mn.gov, title quoted verbatim; not enacted, per the same table.
The camera list. Minnesota Bureau of Criminal Apprehension, "Agencies that use License Plate Readers (LPR)", dps.mn.gov/divisions/bca/data-and-reports/agencies-use-lprs-lpr, retrieved September 10, 2026. Parsed from the page's own structured data payload rather than by hand: 116 agency entries; 87 naming at least one fixed location, for 747 location lines; 4 disclosing only a withholding under subd. 8(b) covering 17 locations (Crow Wing County Sheriff's Office 4, Robbinsdale Police Department 2, University of Minnesota Police – Twin Cities 10, and Edina Police Department, whose notice is singular and states no number, counted as 1); 25 reporting none, all 116 answering "N/A" on other devices with reader capability. Format counts (538 intersections, 168 street-number addresses, 41 neither; 229 lines recording direction; 4 exactly duplicated lines) computed from those 747 strings.
Demographics. U.S. Census Bureau, 2020 Census Redistricting Data (P.L. 94-171), Minnesota file mn2020.pl.zip, downloaded from www2.census.gov on September 10, 2026. Population from table P2 field P0020001 and non-Hispanic white alone from P0020005, joined by logical record number to the geographic header, at summary levels 040 (state), 160 (place), and 140 (census tract). State check: 5,706,494 residents, 76.3 percent non-Hispanic white alone. City-to-agency matching done by name against the 913 Minnesota places in that file; the 28 county, state, tribal, campus, airport, and multi-city agencies were excluded from the per-capita tables for want of a defensible denominator, and are listed in the working notes. The American Community Survey was unavailable: api.census.gov returned HTTP 302 with header X-DataWebAPI-KeyError: 1 for both ACS and decennial endpoints, so no income figure appears anywhere in this article.
Geography. U.S. Census Bureau geocoder (geocoding.geo.census.gov), Public_AR_Current benchmark, Census2020_Current vintage, used to resolve location strings to points and 2020 census tracts; 505 of 747 statewide lines resolved. Census Bureau TIGER/Line 2020 tract boundaries for Minnesota (tl_2020_27_tract.zip) used for point-in-polygon assignment of traffic segments and crime incidents to tracts; that routine was validated against the geocoder's own tract answers on 38 of 38 points before use. The Minneapolis entry "4th Street SE and I-35 West" was excluded because the geocoder matched it to "N 4th St & 35th Ave N", a different location.
Traffic. Minnesota Department of Transportation, AADT_SEGMENT_CURRENT feature service at webgis.dot.state.mn.us (annual average daily traffic segments, current), queried September 10, 2026 for COMMUNITY = 'Minneapolis': 1,235 segments carrying a current volume, count years 2007 through 2025. Citywide median 5,100, 75th percentile 8,974, 90th percentile 15,838, maximum 204,383. Reader-location volume is the highest-volume segment within 75 metres of the geocoded point; all 28 Minneapolis readers had one, median 13,644, which sits at the 86th percentile of the 1,235 segments.
Crime. City of Minneapolis, Police_Incidents_2025 feature service, queried September 10, 2026 for UCR codes 01, 03, 04, 05, and 08 plus offense code CARJCK: 8,731 incidents, all of which carried usable coordinates and were assigned to a 2020 census tract.
This piece ran long — past the house 1,100–1,400 word norm — because the statutory reading, the statewide table, the within-city test, and the two controls on it are each load-bearing and none could be cut without leaving a claim unsupported. No income figure appears anywhere: the Census ACS API requires a key this project does not have, verified from the response header, with the Census geocoder run as a control to confirm the failure was the key and not the network. The Department of Administration's ALPR page could not be enumerated — it redirects automated requests to a CAPTCHA validator — so no claim is made about whether biennial audit results are posted there; the claim made is that § 13.824, subd. 6 does not require them to be. The count of 747 is a count of published location lines, not of devices: the list's parenthetical numbers function as device counts in some agencies' entries and as device labels in others, so a device total cannot be derived from it. Direction of travel appears on 229 of 747 lines, so duplicate-looking entries cannot reliably be separated from paired opposite-direction cameras. No installation or removal dates exist on the list and it carries no publication or revision date, so currency is unknown. The within-city analysis rests on 28 readers in one city and does not generalize to the state. Reported crime is not a neutral measurement of crime, and controlling for it moves the question upstream rather than settling it. HF 4661's authorship and committee history were not verified. The 75-metre reader-to-segment matching radius is the author's choice. This piece reads a public statute against a public dataset for a general audience. It is not legal advice, it concerns no client, case, or investigation of mine, and reading it creates no attorney-client relationship. Corrections: campaign@madgettformn.com.