Research by Ramon de Punder and Mathijs Dijkstra questions that reassurance. They studied 694 Dutch home transactions. Appraisals were much closer to agreed purchase prices than their statistical test would predict under independent valuation. The key question is not just how close the figures are, but how the valuation was formed. [1]
A purchase price is one outcome
A home's price reflects the property, its location and the market. But it also reflects the particular buyer. Perhaps their family lives nearby. Perhaps, after months of searching, they are determined not to lose again. These circumstances can lead one buyer to pay more than the next would.
The researchers call this the transaction-specific component of the price. For an independent valuation, they ask a different question: what would this home be expected to sell for under the same market conditions, to another buyer? One agreed purchase price does not provide the full answer.
An appraisal that confirms your offer does not establish that the next buyer would pay the same amount.
More than four in ten appraisals almost match the purchase price
In the sample, 21.3% of appraisals exactly matched the purchase price. But that does not tell the whole story. Appraisers commonly use round figures: 95.1% of appraisals were multiples of €5,000, compared with 47.7% of purchase prices.
The researchers therefore also examined the difference on that rounding scale. In 41.5% of cases, the gap was no more than €2,500: rounding the difference to the nearest €5,000 gives zero. And 60.8% of appraisals were within €5,000 of the purchase price. These groups overlap; the percentages should not be added together. [1]
The concentration around the purchase price
How much agreement could happen by chance?
An appraisal and a purchase price both describe the same home. Finding that they are close is therefore not enough to demonstrate dependence. This is where the study adds something useful.
For each home, the researchers constructed a distribution of possible selling prices using SlimBieden's statistical model. Its inputs are property characteristics and macroeconomic information available before the transaction. Neither the final purchase price nor the appraisal is used as an input.
They then held each appraisal fixed and repeatedly drew new selling prices from the corresponding distribution. Across 100,000 simulations, this creates a reference in which prices are independent of appraisals. In the observed data, the figures are much closer. The test provides very strong evidence against independence. For this comparison, the model needs to describe both the expected price level and the spread of possible selling prices well. The researchers checked both. [1]
The additional analysis also shows how strongly appraisals follow purchase prices. Approximately 80% of the difference between the purchase price and the model’s expected price is passed through to the appraised value.
Why it matters to a buyer
The paper describes a home bought for €721,000. The model's expected selling price is approximately €663,000. The appraisal is €725,000. The appraisal almost confirms the purchase price, while the model suggests a lower expected price level.
With the purchase fully financed, the mortgage on this home is €721,000. If the buyer runs into payment difficulties and the home has to be sold, the question is whether the proceeds will cover that debt. A sale at €663,000 leaves a shortfall of €58,000, before selling costs.
The financial risk arises as soon as the mortgage exceeds the expected resale proceeds. An appraisal close to the purchase price can obscure that risk. Rising house prices may narrow the gap; falling prices may widen it. The study’s scenario analysis shows how the probability of a shortfall on sale develops as house prices rise or fall while the buyer makes mortgage repayments.
What the study does and does not establish
The analysis covers 694 homes from 800 appraisal reports supplied by mortgage adviser Viisi, spanning December 2025 to August 2026. A total of 106 homes were excluded because required characteristics were missing or the property fell outside the model's application domain.
The researchers check whether the model describes selling prices well. These calibration checks find no evidence of systematic discrepancies in either the predicted price level or the spread of possible selling prices. This supports using the price distributions as the reference for the test. They also examine how sensitive the findings are to model errors and to information about the home that buyers and appraisers may have but the model does not.
The results do not establish deliberate misconduct by any individual appraiser. The authors point to incentives in the process: buyers, sellers and other parties benefit when financing is approved and the sale proceeds. Those interests can conflict with the independent check an appraisal is intended to provide.
More insight before you bid
The practical lesson for buyers is that an appraisal after your offer has been accepted does not automatically establish whether the offer was sensible beforehand. Distinguish what you can finance, what you are willing to pay and the selling prices you might expect based on the home and the market.
The researchers see a role for transparently tested statistical valuations that do not use the agreed purchase price. Such models can show uncertainty alongside an expected price level. That helps explain the trade-off: a higher bid may improve your chance of winning, while also taking you further above the expected selling price.
The conversation need not end with one reassuring number. A more useful question is: what does this valuation tell me about the home, and what mainly reflects the price I have already agreed?
Know your chances of winning, before you bid
SlimBieden calculates your chance of winning for each bid, so you bid with optimal certainty, without unnecessarily paying thousands of euros too much.
About the research
[1] De Punder, R. F. A. & Dijkstra, M. R. G. (2026). Copying Transaction Prices: How Dutch Housing Appraisals Track Idiosyncratic Variation.
Based on the manuscript dated 9 November 2026, SB–2026–01/I. References: Section 4 and Tables 2–3 (sample and rounding), Section 5 (tests and regressions), Section 6 (collateral risk), Appendices A and E.5 (model checks and sensitivity).
Ramon F. A. de Punder and Mathijs R. G. Dijkstra are affiliated with the University of Amsterdam, Tinbergen Institute and SB Statistics, the research unit of SlimBieden. The study was conducted in collaboration with Viisi and uses SlimBieden’s model.