Turin's court auctions: only round 2 sits on the legal schedule, and round 5 jumps to 79.5%

In Turin, the median asking price of later court auction rounds is 75.3% of the first asking price in round 2, 61.2% in round 3 and 55.9% in round 4. Round 5 then reads 79.5%. These figures come from our reading of the court auction notices published for the city. They are asking prices, not sale prices.

One of these numbers lines up with a legal rule. The others do not, and the data we hold cannot yet say why.

Only round 2 sits close to the legal schedule

When a lot receives no bid, the court offers it again at a lower price. The IMF's July 2020 Italy technical note on non-performing assets describes the design. It says Italy's 2015 and 2016 foreclosure reforms allowed reductions of up to 50 percent from the estimated value of the asset. A footnote adds that the auction price can fall by 25 percent from the appraisal value after each failed auction, and up to 50 percent of the appraisal value after the fourth.

A round-2 median of 75.3% sits next to the 75% that one 25% step would give. That is a match between a median and a document, not proof that the rule produced the median. The first ask in a notice may not equal the appraisal, and the note describes the rules as they stood in 2020, so anyone relying on a specific lot should check the schedule now in force.

Rounds 3 and 4 sit above what the schedule alone would give

The footnote can be read two ways. If each step is a flat 25% off the appraisal, two steps leave 50%. If each step compounds, two steps leave about 56%. The round-3 median of 61.2% is above both. Round 4 at 55.9% is above a flat reading with its 50% limit, and well above a compounded three-step reading of about 42%.

So the series does fall, as a discount schedule would make it fall, but the size of the falls does not follow the rule. I can offer candidate reasons and none is tested. The first ask may sit below the appraisal in some notices. Lots that sell leave the sample, so the lots that reach round 4 may differ from those that reached round 2. The notices may also number rounds differently from the legal count of failed auctions.

That second point matters more than it looks. The median in each round is taken over the lots that reached that round, not over one lot descending the ladder. We hold no counts of notices per round, so we cannot say how many lots stand behind each number and do not describe the groups as large or small.

Round 5 at 79.5% shows the medians are not one lot's path

The median moves from 55.9% in round 4 to 79.5% in round 5. Because each round covers a different set of lots, this does not show any lot's price rising. It shows that the group present in round 5 relates its asks to its first asks differently from the round 4 group.

Two hypotheses fit, and neither is tested. One is composition: round 5 may hold lots whose first ask was already low, so their later asks are a higher share of it. The other is procedural: a lot whose sale lapsed or whose procedure restarted may be offered again at a reset price. Either would lift the median with no buyer changing a view.

The per-round notice counts would separate the two. If round 5 rests on very few notices, a single lot entering or leaving moves the median. If it rests on a number similar to round 4, composition or a reset is the likelier cause.

My reading, and it is only that, is that the round medians describe the notices in the sample better than they predict the next price on any lot. For someone screening court lots, the individual notice carries the evidence: its first ask, its round and its stated reduction. The Turin median is context.

What to watch

Two checks would settle most of this.

The first is the notice count behind each round. Our quarterly count of published auction rounds in Turin runs 71 in Q4 2023, 101 in Q1 2024 and 83 in Q2 2024, then 5 in Q3 2024 and 3 in Q3 2026, with 53 so far in Q4 2026, about five weeks into the quarter. Gaps that deep look like publication or coverage effects as much as changes in supply, so we draw no conclusion from them until the counts are audited. A round 5 median built on a handful of notices would be noise. One built on a count close to round 4 would need the procedural explanation.

The second is the 2026 H2 half-yearly value quotation for Turin's zones. Our median of the zones' mid-points, based on the quotations published for each zone, was 2,312 EUR per m² in 2025 H2 and 2,425 EUR per m² in 2026 H1. The quotations are a reference value, not transaction prices, so they cannot tell us whether auction discounts are wider than ordinary market discounts. They do give an outside marker for whether the reference market is firming or softening while the auction medians remain unexplained.

For context only, the ECB's interest rate statistics for Italian banks put the average rate on new house-purchase loans at 3.60% in September 2026. Our auction data holds no link between that rate and the round medians, and we do not draw one.