The state of Minnesota housing distress
This report publishes what Govire directly observes about distressed property in Minnesota, and what happened to those properties afterwards.
It does not publish a statewide foreclosure count. Coverage is not complete across every signal and every county, and a count drawn from partial coverage presented as a statewide figure would be wrong in a way a reader could not detect. Statewide counts are published by the Minnesota Homeownership Center and by commercial data providers, and this report points to them rather than competing with them.
What is unusual here is not the volume. It is the outcomes — following individual properties past the filing to what actually happened, which is sustained work rather than a snapshot and is why almost nobody publishes it.
What is tracked#
| Signal | Source | Coverage |
|---|---|---|
| Sheriff foreclosure sales | County sheriff feeds, published legal notices | Concentrated in metro counties with published feeds |
| Tax delinquency | County treasurer and auditor records | Varies by county publication practice |
| Tax forfeiture | County auditor, Minn. Stat. ch. 282 | Statewide under a single statute |
| Vacant and condemned buildings | City registers | Minneapolis, St Paul, some suburbs |
| Probate filings | Probate court records | Where published |
| Ownership and assessed value | County assessor, MnGeo parcel layer | 59 of 87 counties in the state programme |
| Recorded sale prices | eCRV, Minnesota Department of Revenue | Statewide |
Everything is normalised onto a parcel spine, so signals on the same property connect. That is what makes stacked signals visible — a parcel carrying tax delinquency, a vacant registration and an absentee owner is a different situation from one carrying any of them alone.
What resolved#
The central finding, and the part that does not exist elsewhere.
Across 326 redemption windows tracked from the sheriff's sale to resolution through recorded deeds and county ownership records:
| Outcome | Share |
|---|---|
| Owner redeemed and kept the property | 33.4% |
| Owner sold during the redemption window | 15.0% |
| Lender kept it or it was resold | Remainder |
Nearly half did not end with the owner losing the property outright.
Restricted to outcomes confirmed by a recorded document rather than inferred from a judgement, the redemption figure is 37.5% of 285 windows. Both are published because 41 of the 326 rest on an inference and the difference is material.
What separates the outcomes#
Equity, more than anything else.
| Sheriff sale bid vs assessed value | Redeemed | Windows |
|---|---|---|
| Under 50% | 58.1% | 31 |
| 50–80% | 44.2% | 77 |
| 80% or more | 20.0% | 50 |
And whether the owner lives there:
| Homestead status | Redeemed | Windows |
|---|---|---|
| Homesteaded | 38.6% | 171 |
| Not homesteaded | 24.6% | 114 |
Both fields are published on the county assessor record, so the strongest observed predictors of what happens to a foreclosed property are knowable in advance at no cost.
The bid-to-value cut is Hennepin only, because only Hennepin publishes the winning bid amount. And a survival model disagrees with this table about how much of the effect is equity and how much is something correlated with it — that disagreement is unresolved and documented in the companion piece on predictors.
How long it takes#
Across 1,336 tracked mortgage foreclosure windows:
| Time from sheriff's sale | Reached a foreclosure sale | Owner sold during the window |
|---|---|---|
| 6 months | 3.2% | 11.7% |
| 12 months | 33.0% | 15.0% |
| 18 months | 51.0% | 15.0% |
Minnesota's statutory redemption period is six months. At six months, 3.2% had resolved. Half were still open at eighteen.
Owner sales stop after the first year — 15.0% at twelve months and unchanged at eighteen. The option to sell during a redemption window is a first-year option.
Redemption by county#
| County | Redeemed | Windows | Confirmed only |
|---|---|---|---|
| Hennepin | 39.2% | 158 | 43.0% of 142 |
| Dakota | 26.2% | 61 | 27.1% of 59 |
| Washington | 24.2% | 66 | 25.0% of 64 |
County differences should be read cautiously. They reflect a mix of local conditions and how completely each county's windows have been worked. A county checked less thoroughly looks like a county where less happens. Counties below twenty confirmed windows are not published.
Tax forfeiture is a separate system#
Minnesota does not sell tax liens. Unpaid taxes lead to a judgment, then a statutory redemption period of three years for most homestead and agricultural property and five for other classifications, then forfeiture to the State of Minnesota. Counties then sell forfeited parcels under ch. 282.
Two clocks that must not be pooled:
| Mortgage foreclosure | Tax forfeiture | |
|---|---|---|
| Statute | Ch. 580 / 582 | Ch. 281 / 282 |
| Redemption | 6 months from the sale | 3–5 years from judgment |
| Ends with | Certificate holder or lender | The State of Minnesota |
| Average window length observed | Under a year | 921 days |
This is not an academic distinction. Pooling 335 forfeiture windows into a mortgage redemption analysis moved a published one-year figure from 33.0% to 18.8%, because forfeiture windows sit in the risk set indefinitely and can never experience a mortgage foreclosure outcome. The corrected figures are the ones above.
Forfeited land is concentrated in the north. St Louis County holds one of the largest programmes in the country, and Aitkin, Cass, Itasca, Koochiching and Beltrami carry substantial rural acreage. Hennepin and Ramsey hold far fewer parcels, and theirs are urban lots and structures rather than land.
What this report does not claim#
No trend. The tracking history is not yet long enough to distinguish a real change in distress volume from an expansion in coverage. A rising line that is actually more counties being added is worse than no line.
No statewide count. Coverage is partial by signal and by county.
No causal claim on equity. The association is strong in the rate table and absent in a stratified model, and both are published.
No forecast for an individual property. A survival model fitted on this data reaches a concordance index of roughly 0.66 to 0.70, which is better than chance and not strong enough to rank individual windows usefully. The published figures describe populations.
Corrections made this year#
Published because a reader assessing whether to trust these figures should see how the estimates were reached.
Statutory tracks separated. 335 tax forfeiture windows removed from the mortgage foreclosure timing analysis. One-year figure moved from 18.8% to 33.0%.
Superseded rows excluded. 762 duplicate tracker rows retired. Washington County had read 93 resolved windows against a true 66 before the filter was applied, and the statewide resolved count moved from 353 to 326.
Lender pattern list retuned. Reading back recorded owner strings found 56 already-resolved outcomes the pattern list had missed, including 23 parcels held by the Secretary of Veterans Affairs across four spellings. Resolved count moved from 252 to 412 before deduplication.
Homestead artefact corrected. Model concordance fell from 0.803 to 0.661 and the homestead term ceased to be significant once the statutory tracks were separated. The earlier significance was an artefact of the pooling error.
What the signals look like together#
The value in normalising signals onto one parcel spine is that combinations become visible, and combinations mean something a single signal does not.
One signal has an innocent explanation. Tax delinquency might be someone who moved and missed a bill. A code violation might be a fence.
Three unrelated innocent explanations on the same parcel is unlikely.
| Combination | What it usually describes |
|---|---|
| Tax delinquency + vacant registration | Nobody paying, nobody living there — heading to forfeiture |
| Sheriff sale + absentee mailing address | An investment or an inheritance, not a family home |
| Probate filing + tax delinquency + out-of-state heir | An estate nobody is administering |
| Code violations + long ownership + low assessed value | An ageing owner unable to maintain the property |
| Condemnation + tax delinquency | Heading to demolition or forfeiture |
| Deceased owner + no probate + years of unpaid tax | Tangled title, and the clearest signal of a family losing an asset |
That last one is the pattern worth naming. A property still recorded in the name of someone who died, with no probate opened and tax accumulating, will eventually forfeit. The heirs frequently do not know they own it and nobody has standing to pay the bill. Research on heirs' property nationally finds it concentrated in lower-income communities and identifies it as a significant route of intergenerational wealth loss.
It is also a substantial share of the vacant property in many cities, and it is visible in public records years before the forfeiture.
The measurement problem, stated plainly#
Why outcome data is rare, and why what exists should be read carefully.
Redemption leaves no positive record. No document is filed saying an owner redeemed. There is only the continued absence of a transfer. Detecting it means returning to each window after expiry and re-checking the record — sustained work rather than an extract.
Detection improves and figures move. Every improvement to the detection method so far has moved resolved counts up, because misses appear as unresolved rather than as errors. That means current figures are more likely conservative than inflated, and it also means they will change.
Absence of evidence is the default state. A window with no detected outcome is not a window where nothing happened. It is a window we have not resolved, and treating those two as the same is the error that makes most published foreclosure statistics describe filings rather than outcomes.
Which is the reason for publishing counts alongside every rate. A reader can see how much weight a figure carries, and a figure resting on 31 windows is visibly different from one resting on 326.
Where this goes next#
More counties. Sheriff sale tracking is concentrated where feeds are published most completely. Extending it improves precision and may shift central estimates if the added counties differ systematically.
Working the pending backlog. Windows currently unresolved will resolve, and each one updates the figures. Whether the redemption rate rises or falls as they do is not predictable in advance.
Testing the homestead prediction. Dakota County shows a 23-point observation imbalance between homesteaded and non-homesteaded pending windows. Working those down will either move the homestead gap in a predictable direction or show the raw difference was largely a coverage artefact. The result will be published either way.
A second state. The method transfers; the numbers do not. Every state has different statutes, different recording offices, different redemption rules and different data availability, and getting those wrong produces confident nonsense. Each new state is a research problem before it is an engineering one.
And more signals followed to outcome. Redemption windows are followed because a sheriff sale creates a dated event with a statutory expiry. Tax forfeiture, vacant registrations and probate filings could each be followed the same way, and are not yet.
For journalists and researchers#
Cite freely, with the sample size. A rate without its count is not a finding, and it is the discipline this dataset is built on.
Suggested form: Govire, observed Minnesota redemption outcomes, n=326 resolved windows, August 2026.
Cuts not published here are usually answerable from the underlying records — by county, by signal, by property characteristic, by time period. Access for research and reporting use is free.
Where a published figure moves materially, the previous version and the reason for the change are published alongside it. Nothing here is quietly updated, because these figures may be cited and a citation should remain checkable.