Flats resold without renovation by the flipper produced gains of 28.66% to 32.82% on the purchase price. These are the strictly speculative groups under Methods I and II: 68 of 136 flips in one sample and 695 of 3,426 in the other. The paper rests on this gross buy-sell spread. Its returns do not specify what has been deducted, and none of the profit figures includes tax, fees, financing or renovation costs. Reading 30% as a rent requires doing that netting independently.
Where the spread comes from
The paper asks where flippers make their money and offers four possible sources: renovation or resolution of title defects, timing a rising market, sales-side work, and arbitrage. Sales-side work covers staging, marketing, exposure and negotiation. A broker can supply those services for 2% to 4% of the sale price, most commonly 3%. Arbitrage means purchasing below market from a hurried or uninformed seller, then reselling at market. One version casts the flipper as a middleman compensated for supplying immediate liquidity to a distressed seller (Bayer et al., 2011).
Chamier Cieminski hand-assembled repeat transactions from the Polish Real Estate Price Registers, working county by county. The dataset contains 1,131,046 flat transactions from 14 cities monitored by the National Bank of Poland between 2005 to 2023. Those cities account for 3.68% of counties and 43.22% of national flat transactions. Any pair of sales at the identical address within 24 months counts as a flip. Over 61,000 meet that definition, recording mean profit of 36.45% over a mean duration of 8.40 months.
Across all 14 cities, flippers bought smaller flats than the city median. In Warsaw, the comparison is 46.00 sqm against 50.59. They also paid less per square metre in 13 of 14 cities: Warsaw shows 8,576 against 9,185 PLN/sqm. Kraków is the exception, at 6,826 against 6,715.
Physical renovation never appears directly in the records, so the paper infers it twice. Method I reproduces the four-transaction design of Bayer et al. (2011) for flats sold at least four times. It flags improvement when the post-flip price is above 1.20 times the pre-flip price after scaling for city price growth. The filter leaves 136 flips, or 0.23% of all flips. Exactly 50% show no evidence that the flipper renovated them. Those flats earned 32.82%, compared with 47.03% for the remainder.
Method II is Chamier Cieminski's own three-transaction design and reasons backwards. When the buyer from the flipper later resells at more than 20% above trend, the subsequent buyer is treated as having paid for the work, leaving the flipper as the party who did not renovate. This method covers 3,426 flips. Of those, 695 (over 20%) are what he calls definitively not renovated by the flipper, earning 28.66% versus 39.42%. Short flips add 13,919 observations, carry a two-month mean holding period and returned 26.30%.
Arbitrage absorbs everything else
The classification is broad by design. Staging, negotiation and improved liquidity are "functionally classified within this framework as tools used of arbitrage". The limitations section makes the same concession. The data cannot separate the financial contribution of intermediary services, small improvements such as home staging, or stronger negotiation skills, and "these activities inherently function as tools of efficient arbitrage". The unimproved group's 30% therefore remains after removing renovation, while every other action by the flipper stays inside the residual. The paper answers with a price comparison: ordinary broker fees are approximately 3% of the sale price, far below the 28.66% to 32.82% baseline.
The 3% covers marketing. Carrying the asset is a different exposure. Brokers warehouse nothing, while flippers hold the flat in a market without a continuous quote. Across the full population of over 61,000 flips, that hold averaged 8.40 months. The international comparison relies on two findings from a literature the paper describes as dominated by the United States. Flippers buy at an average 16% discount to other participants (Li et al., 2023) and sell at a 5% premium (Guntermann et al., 2011).
Chamier Cieminski treats those figures as evidence that flippers profit from market inefficiency. We draw another inference from his two citations rather than challenging his stated claim: most of the 20%-plus spread arises at the transaction points instead of during the holding period. For a trader, the distinction between extracting the Polish 30% from an uninformed seller and receiving it for underwriting and carrying an illiquid asset matters. The second source decays as capital arrives.
And capital was arriving.
The earlier Polish study cited in the paper shows flips climbing from 2% to 3% of transactions in 2017, then reaching 5.9% in 2020. Bydgoszcz recorded 14.5% and Katowice 11.2%. Flippers also bought 20% of all properties sold at auction. Their counterparty is a household unaware of its own price.
A boom-heavy measurement
The paper says exploitation of information asymmetry "increases significantly during housing market booms", as owners struggle to follow their own valuations. Both extracted samples lean heavily toward that setting. Method I assigns 74% of observations to 2014 to 2019 and 3% to 2020 to 2023. The corresponding shares for the full flipping population are 45.51% and 37.96%. Method II contains 63% and 16%.
The appendix makes the concentration starker. Of the 136 Method I flips, 40 come from 2018 alone. The years 2022 and 2023 supply no Method I observations at all, while Method II receives 85 of its 3,426. Chamier Cieminski identifies this as structural censoring rather than sampling error. The limitations section explains that recent transactions are underrepresented because a 2022 flip has had insufficient time to produce another recorded sale. The diagnosis is sound, and the consequence remains: the 50% figure measures the middle and later stages of a boom.
Chamier Cieminski also declines to sweep the threshold and explains the choice. He considers the 20% premium conservative already, equivalent to over PLN 1,350/sqm on average and over PLN 3,000/sqm in Warsaw in 2023. A dynamic adjustment, he writes, "would not fundamentally alter the core findings". That seems fair for the result's direction. Its size depends directly on the threshold because the dividing line determines both cohort averages. And size carries the argument: a roughly 30% arbitrage baseline beside a renovation premium of only 10 to 17 percentage points. Shift the line and both means shift. Chamier Cieminski acknowledges that Method II's 39.42% group simply contains flats whose next buyer did not upgrade immediately, which does not establish renovation by the flipper. One leg of the 10 to 17 is therefore a mixed group even on his description.
The design has another consequence. Any same-address resale within 24 months qualifies as a flip, without confirmation that the buyer was an investor or any exclusion for owner-occupier resales. Method II defines its unimproved group as flats whose next buyer resold more than 20% above trend, creating a selection toward the worst stock. Asset quality could explain its lower 28.66% as readily as renovation. Table 3 introduces a small inconsistency as well: it reports 3,427 Method II observations, while the text and appendix each report 3,426.
We could not test any of this. The traded object is a particular Polish flat, and the signal needs address-level register microdata, repeat-sale matching and evidence about physical condition. Our data contain none of those elements, while no listed proxy retains a mechanism based on uninformed sellers.
Where the paper earns its place
Only one earlier empirical study estimated Polish flipping from actual transaction records (Czerniak et al., 2021). This paper therefore provides the fullest transaction-level account of the market currently in print. Validation against Statistics Poland finds median volume alignment of 99.87%, although 4 cities were excluded from that comparison. Median transaction price alignment reaches 100.43%. City results are less tidy, with Lublin average prices at 126% and Łódź volumes at 83%.
The paper presents itself as exploratory and openly describes its methods. The abstract uses that label, the limitations section identifies heuristic thresholds and absent inferential tests, and the data-validation section says confidence intervals were deliberately omitted. Its strongest use is as an existence proof: Polish flats can generate gross spreads of 28.66% to 32.82% without construction work.
The same cohort gap after subtracting PCC, notarial fees, brokerage and financing would change my view. A threshold sweep would also help, along with quarterly city multipliers in place of annual ones for the trend adjustment. The evidence would improve further if the 2020 to 2023 share rose beyond the 3% and 16% currently represented in the two samples.