# Literature notes — Part B: Empirical evidence and measurement (WP19)

Scope: empirical and measurement literature bearing on WP19's mechanisms (seller urgency, time on market, forced sales, cash versus mortgage buyers, sale selection and price indices, volume–price comovement, duration and censoring). Search theory proper (Wheaton, Krainer, Albrecht et al., Díaz–Jerez, Head et al., Ngai–Tenreyro, Guren, Hedlund, Guren–McQuade, Taylor, Yavas–Yang) is covered in Part A and only referenced here where an empirical paper depends on it. Every key below corresponds to a row of research/lit_part_B.csv and a Crossref record or evidence file in research/verification/B_<key>.*. Statements are based on abstracts actually retrieved (Crossref, IDEAS/RePEc, NBER, CESifo, OpenAlex) or on content known with certainty; anything else is marked "[unconfirmed]". Compiled 2026-10-20.

# Theme 1 — Time on market, price and seller motivation

The micro literature on marketing time began as an application of optimal-stopping theory to listings data. haurin1988duration showed that houses with atypical characteristics, which face a wider dispersion of offers, remain on the market longer, and zuehlke1987duration fitted Weibull hazards to MLS listings and found stronger time dependence of the sale hazard for vacant than for occupied houses (the vacancy split is reported by secondary sources; the original abstract was not retrievable [unconfirmed magnitudes]). jud1996brokerage likewise reports positive duration dependence and finds that pricing strategy, not broker identity, drives marketing time. These early studies treat the seller as homogeneous apart from observable house traits, so the estimated duration dependence confounds genuine state dependence with unobserved heterogeneity; WP19's prediction that a mixture of urgency types produces apparent negative duration dependence is therefore not contradicted by them, but it does require a reading of the positive-dependence results as reflecting list-price revisions or learning.

Seller heterogeneity entered explicitly with springer1996motivation and glower1998motivation. Springer finds that sellers flagged as eager or motivated, relocated sellers, vacant homes and foreclosures all obtain lower prices, but only foreclosures sell faster, and he concludes that the list price is the seller's primary instrument. Glower, Haurin and Hendershott, using a survey of sellers linked to outcomes, show that a planned move date shortens marketing time, that a shorter horizon to the planned move shortens realised time on market, and that motivation lowers the sale price but not the list-price markup; their closing recommendation that "theoretical models of the housing search process should be recast to allow for heterogeneous sellers" is essentially the research programme of WP19. The price–time tradeoff itself is documented by anglin2003tradeoff (a higher degree of overpricing raises expected time on market, more strongly where list prices are informative) and knight2002listing (large list-price revisions are associated with longer marketing and lower final prices; vacant and overpriced homes are the most likely to be cut), and formalised with a joint list/reservation-price model in haurin2010listprices, where greater offer dispersion raises the list-to-expected-price ratio. huang2001withdrawal provides an estimator that recovers the seller's reservation price jointly with duration and finds that reservation prices fall with elapsed time on market, consistent either with learning or with rising effective carrying costs. levitt2008agents gives the cleanest estimate of the marginal value of waiting: agents selling their own homes obtain about 3.7% more while waiting about 9.5 days longer.

Equity and reference points move the reservation price in the opposite direction from urgency. genesove1997equity shows that high loan-to-value sellers in Boston set higher asking prices, expect longer marketing periods and obtain higher prices conditional on sale (the often-cited 12% asking-price gap between 100% and 80% LTV sellers was not visible in the abstract [unconfirmed]); genesove2001lossaversion attributes a similar pattern to loss aversion, with asking prices higher by roughly 25–35% of the prospective nominal loss and a markedly lower hazard of sale. For WP19 these papers argue that the distribution of carrying costs should admit "locked" or patient sellers with effectively negative urgency, not only motivated ones, and that constrained sellers can raise, rather than lower, transaction prices conditional on sale. The remaining papers in this theme measure the market state. genesove2012search estimates, from NAR surveys, that seller time on market falls sharply as demand rises and that the elasticity of the seller contact hazard with respect to the buyer–seller ratio is 0.84 under constant returns, a direct calibration target for WP19's matching function. carrillo2013heat recovers a seller-bargaining-power index from list price, sale price and TOM, and carrillo2015tightness shows that sale probabilities and bargaining power forecast subsequent appreciation and turning points; han2014biddingwars documents the rise of above-list sales from 3–4% to over 30% of transactions in hot markets, and han2016askingprice shows that the asking price directs buyer search without being a ceiling. hendel2009fsbo documents endogenous sorting of patient sellers to the FSBO platform (no price penalty but longer time to sell and lower sale probability), a platform-level instance of selection on urgency, while merlo2004bargaining provides the only dataset with every list-price change and every offer, and anenberg2016information shows that seller uncertainty and learning can produce momentum, an alternative mechanism WP19 must distinguish from liquidity-driven selection.

# Theme 2 — Forced and distressed sales, liquidity discounts

campbell2011forced is the reference point: in twenty years of Massachusetts transactions, foreclosure sales trade about 27% below comparable houses, death- and bankruptcy-related sales also sell at discounts, and a foreclosure within 0.05 miles lowers a neighbouring house's price by about 1%. clauretie2009foreclosure cautions that the earlier literature's discounts above 20% are roughly one-third too large once property condition, the endogenous relationship between marketing time and price, and spatial dependence are controlled, so a substantial but smaller component is attributable to the forced nature of the sale per se. penningtoncross2006reo shows that the REO shortfall relative to metro appreciation varies with legal regime, market conditions and marketing time, and mayer1998auctions establishes that the price of immediacy is state dependent: auction discounts of 0–9% in the Los Angeles boom versus 9–21% in post-bust Dallas, consistent with a seller trading an immediate sale against a long expected search. andersen2017firesales identifies forced sales from sudden death, so the supply shock is orthogonal to current demand, and finds discounts that increase with urgency, with poor market conditions and with the seller's financial constraints, which is the sharpest causal support available for WP19's interaction between carrying cost and market liquidity. Externality and aggregate evidence complete the picture: harding2009contagion estimates a peak contagion discount of roughly 1% per nearby foreclosed property that decays with distance; anenberg2014foreclosures shows that sellers react in the exact week a nearby REO is listed, with competition effects everywhere and disamenity effects only in dense, low-price areas; mian2015foreclosures uses judicial-foreclosure laws to show that foreclosures caused large declines in house prices, residential investment and consumption in 2007–2009. Collectively these papers establish that an increase in the mass of urgent sellers lowers prices for urgent and non-urgent sellers alike without any change in fundamentals, which is the empirical content of WP19's liquidity-shock proposition; what they do not provide is a decomposition of the discount into bargaining under high carrying cost versus selection of which houses transact.

# Theme 3 — Cash versus mortgage buyers and financing contingencies

The cash discount has been measured for three decades with widely varying magnitudes. asabere1992cash reports a discount of roughly 13% on a small sample of row houses; jauregui2019fundingprice shows that a large part of raw differentials across cash, conventional, FHA and VA funding reflects selection on property and neighbourhood characteristics, leaving an unexplained "financing premium"; aroul2023cashdiscount finds that cash discounts widen in distressed markets, at lower price levels and for smaller or poorer-condition properties. Two recent papers supply the mechanism WP19 formalises. han2024cashisking shows that in Los Angeles a cash offer closes 29% faster and carries a 2–3.9% discount, that the discount is smaller in booms and in larger markets, and, crucially, that closing risk alone cannot rationalise it: the seller must also price the risk that a property returned to the market after a failed contingency does not sell. reher2024mortgagecash estimates an 11% mortgage premium over 1980–2017 against a 3% prediction from a representative-seller model; heterogeneity in selling conditions closes half of the gap and belief distortions are invoked for the remainder in high-transaction-risk states. The gap between the 2–4% structural discount of Han–Hong and the 11% reduced-form premium of Reher–Valkanov is exactly the space where WP19's heterogeneous urgency and selection into cash sales operate. lin2025contingencies (CESifo WP 9947 version: Gordon, Seiler, Siebert and Winkler 2022) documents that contingency clauses in Miami-Dade carry price premia from −3.7% to +2.2% depending on clause and market state, and buchak2026ibuyers shows that intermediaries who sell immediacy (iBuyers) earn about a 5% spread and survive only in liquid, easy-to-value markets.

The qualification side is documented by mortgage-market studies that WP19 uses only as inputs to its buyer-arrival and approval parameters. anenberg2019creditavailability builds a loan frontier showing credit expanding for all borrowers in the boom and contracting most for low-score borrowers; bhutta2021insurance shows that a 50 basis point cut in FHA premiums raised home buying by 14% among FHA-reliant households through DTI thresholds, i.e. qualification is highly elastic at the margin; defusco2020leverage finds that the Ability-to-Repay rule eliminated 15% of the affected market while raising rates only 10–15 basis points; fuster2019technology measures approval speed, with FinTech lenders processing applications about 20% faster. allen2019searchfrictions is the only verified Canadian structural evidence: the Canadian insured-mortgage market is itself a negotiated-price search market in which frictions cost about $12 per month per borrower. No verified Canadian or Québec study of residential time on market or cash discounts was found (see lit_unverified_B.md).

# Theme 4 — Selection, indices and volume–price comovement

The measurement literature establishes that transacted houses are not a random sample of the stock and that the bias co-moves with the cycle. jud1994selection found Heckman-type selection to be significant for appreciation estimates; gatzlaff1997selection showed that repeat-sales indices built from homes that sell more than once are biased and that the bias is strongly correlated with changes in economic conditions; gatzlaff1998selection extends this to local hedonic indices [details unconfirmed]; hwang2004selectivity, using all sales in Sweden's largest metropolitan area, shows that repeat-sales estimates of the value of the stock are systematically biased once selectivity, quality change and mean reversion are modelled. The paper closest to WP19's selection proposition is goetzmann2006reservation: when sellers have heterogeneous reservation prices, falling volume biases index returns upward and rising volume biases them downward, and a reserve-conditional index differs substantially from the standard repeat-sales index for Los Angeles. In commercial real estate, fisher2003liquidity and fisher2007tbi build constant-liquidity (demand-side) and supply-side indices and find that constant-liquidity values lead transaction-based indices and are more volatile, while clayton2008liquidity finds that time variation in liquidity is best explained by the option value of waiting. The classical index tools (bailey1963repeatsales, case1989efficiency with its finding of positive serial correlation in 39,210 repeat sales, clapp1992assessed with the assessed-value alternative) assume random sampling of the stock and are the objects whose bias WP19 characterises.

Bargaining power and information are the remaining confounders. harding2003bargainingpower identifies bargaining effects from the characteristics of both parties in the American Housing Survey and shows that seasonal bargaining variation explains anomalies in quality-adjusted indices; harding2003bargainingeffects finds that vacant sellers receive lower prices and that bargaining alters attribute prices. kurlat2015asymmetries and garmaise2004asymmetries document adverse-selection channels (informed sellers exit before declines; buyers avoid informationally disadvantaged trades) that WP19 excludes by assuming symmetric information; piazzesi2020segmented shows that broad searchers spread local shocks across segments, a dampening mechanism absent from a single-segment model. On comovement, berkovec1996turnover and hort2000turnover [details unconfirmed] find that turnover responds to demand before prices; clayton2010pricevolume finds, across 114 MSAs, that price declines reduce volume while increases do not, that volume Granger-causes prices only where supply is inelastic, and that the positive correlation is mostly co-movement from common shocks; andrew2003transactions documents a structural break in the British price–transactions relation; leamer2007businesscycle makes the macro case that housing volumes lead the cycle. dachis2012transfertax is the one verified Canadian natural experiment: Toronto's 1.1% land transfer tax cut sales by about 15% and lowered prices by roughly the tax. cheng2013liquidityrisk translates uncertain marketing time into a liquidity risk premium on private-asset returns, which is large in down markets.

# Theme 5 — Duration, censoring and withdrawn listings

Observed time on market is right-censored by withdrawals and relistings, and the treatment of these margins changes results. anglin2003tradeoff reports that withdrawn homes have higher and more variable list prices and that withdrawal limits what can be learned about marketing time; benefield2015measurement shows that five alternative definitions of TOM and the handling of relisted properties substantially change coefficient estimates and significance; huang2001withdrawal treats withdrawal as censoring within a joint reservation-price/duration estimator. carrillo2012potatoes characterises the distribution of marketing time (the "hot or cold potatoes" framing) and anglin2006liquidity studies the joint movement of value and liquidity as conditions change, although neither abstract could be retrieved [details unconfirmed]. The duration-dependence evidence (zuehlke1987duration, jud1996brokerage) is positive in parametric models without unobserved heterogeneity; WP19's mixture-over-urgency hazard predicts negative dependence in the observed hazard unless list-price revisions or learning dominate, which makes a competing-risks (sale vs withdrawal) specification with heterogeneity the appropriate empirical counterpart.

# Empirical facts WP19 must be consistent with

Price–time tradeoff: a higher list markup or longer optimal waiting raises the expected price and the expected time on market; marginal value of waiting about 3.7% for roughly 9.5 extra days in Levitt–Syverson (anglin2003tradeoff, knight2002listing, haurin2010listprices, levitt2008agents, springer1996motivation).
Seller urgency lowers price and shortens duration: planned-move sellers sell faster; motivated, vacant, relocated and foreclosed sellers obtain lower prices; vacant sellers have weaker bargaining power (glower1998motivation, springer1996motivation, harding2003bargainingeffects, knight2002listing).
Patient or constrained sellers do the opposite: high-LTV and loss-facing sellers set higher asking prices (about 25–35% of the prospective loss), sell more slowly and obtain higher prices conditional on sale (genesove1997equity, genesove2001lossaversion, hendel2009fsbo).
Forced-sale discounts: about 27% for foreclosures in Massachusetts, smaller for death and bankruptcy sales; roughly one-third of naive foreclosure discounts is condition/TOM selection; the discount rises with urgency, poor market conditions and seller financial constraints; auction (immediacy) discounts 0–9% in booms versus 9–21% in busts (campbell2011forced, clauretie2009foreclosure, andersen2017firesales, mayer1998auctions, penningtoncross2006reo).
Spillovers from urgent sellers: about 1% lower price per nearby foreclosure, operating within the week of REO listing through competition; aggregate foreclosure waves cause large price declines (campbell2011forced, harding2009contagion, anenberg2014foreclosures, mian2015foreclosures).
Cash discount: 2–3.9% structurally with 29% faster closing in Los Angeles; 11% reduced-form mortgage premium nationally versus 3% from a representative-seller model; about 13% in an early small sample; discount smaller in booms and larger markets, larger in distressed markets; contingency price effects between −3.7% and +2.2% (han2024cashisking, reher2024mortgagecash, asabere1992cash, aroul2023cashdiscount, lin2025contingencies, jauregui2019fundingprice).
Qualification is elastic and delay is measurable: a 50bp rate cut raised FHA-reliant purchases by 14%; a leverage rule removed 15% of the affected segment; FinTech processing is about 20% faster; credit frontiers expanded for all and contracted most for low scores (bhutta2021insurance, defusco2020leverage, fuster2019technology, anenberg2019creditavailability).
Matching and tightness: seller TOM falls strongly with demand; contact-hazard elasticity with respect to buyer–seller ratio about 0.84; bidding wars from 3–4% to over 30% of sales in hot markets; tightness indicators lead appreciation (genesove2012search, han2014biddingwars, carrillo2013heat, carrillo2015tightness).
Volume–price comovement with volume leading: turnover responds before prices; price declines reduce volume asymmetrically; the correlation is mostly common-shock co-movement; a 1.1% transfer tax cut Toronto sales by about 15% (berkovec1996turnover, hort2000turnover, clayton2010pricevolume, goetzmann2006reservation, dachis2012transfertax, leamer2007businesscycle).
Index selection bias: indices from sold homes are biased, the bias co-moves with the cycle, falling volume biases returns upward; constant-liquidity indices lead and are more volatile than transaction indices (gatzlaff1997selection, goetzmann2006reservation, hwang2004selectivity, fisher2003liquidity, fisher2007tbi, jud1994selection).
Duration dependence and censoring: parametric hazards without heterogeneity show positive duration dependence; withdrawn listings have higher, more variable list prices; TOM definition and relisting treatment change estimates materially (zuehlke1987duration, jud1996brokerage, anglin2003tradeoff, benefield2015measurement, huang2001withdrawal).
Immediacy has a market price: iBuyers earn about a 5% spread and operate only in liquid segments; liquidity risk on private assets is large in down markets (buchak2026ibuyers, cheng2013liquidityrisk, clayton2008liquidity).

# Identification challenges the model should speak to

Liquidity discount versus quality: forced-sale and cash discounts shrink substantially when condition, location and TOM are controlled (clauretie2009foreclosure, jauregui2019fundingprice); WP19 should state which part of D = (p* − p)/p* survives conditioning on house quality, i.e. the bargaining component among identical houses.
Selection into sale versus price change: observed indices average over sellers who chose to transact; with heterogeneous urgency the composition of sellers shifts with liquidity (goetzmann2006reservation, gatzlaff1997selection, fisher2003liquidity). The model should deliver an explicit expression for the index bias as a function of the urgency distribution and tightness, and conditions under which a constant-liquidity index recovers p*.
Selection into cash: cash buyers choose distressed or low-price properties (aroul2023cashdiscount, jauregui2019fundingprice); WP19's decomposition of the cash discount into bargaining (contingency risk) and selection should map to the Han–Hong versus Reher–Valkanov gap.
Duration dependence versus heterogeneity: positive dependence in homogeneous hazards versus negative dependence predicted by mixing (zuehlke1987duration, jud1996brokerage); the model's observed hazard should be stated as a mixture and the competing risk of withdrawal (anglin2003tradeoff, benefield2015measurement) should be part of the observable mapping.
Reverse causation between price and TOM: list price, TOM and sale price are jointly determined (knight2002listing, carrillo2013heat); the structural mapping from reduced-form coefficients to bargaining power and arrival rates in Carrillo's framework is the template WP19 can extend to urgency types.
Equity and reference dependence: constrained sellers raise, rather than lower, reservation prices (genesove1997equity, genesove2001lossaversion); a model with only positive carrying costs will understate the dispersion of reservation prices and predict the wrong sign for some sellers.
Competing frictions: information (anenberg2016information, kurlat2015asymmetries, garmaise2004asymmetries), agency (levitt2008agents) and segmentation (piazzesi2020segmented) generate similar reduced-form patterns; the novelty claim should identify a moment (e.g. the response of urgent versus patient sellers' prices to a pure matching-efficiency shock, as in andersen2017firesales) that separates liquidity from these alternatives.