# WP19 — Literature review and novelty audit (Phase 1)

Date: 2026-10-20. Basis: 132 references verified against Crossref (129 with a DOI resolved; 3 with a publisher/RePEc evidence
file), merged from two independent searches (`lit_part_A.csv`, theory, 73 rows; `lit_part_B.csv`, evidence and measurement, 66
rows; 7 papers in both; 4 duplicate keys reconciled, see `literature_merge_report.md`). Raw Crossref records are in
`verification/`. Unverified or deliberately excluded candidates are listed in `lit_unverified_A.md` and `lit_unverified_B.md`
(notably: no formal theory of financing contingencies beyond `bian2018bargaining`; no verified Québec or Canadian time-on-market
study). Detailed thematic notes: `lit_notes_A.md`, `lit_notes_B.md`. Model-design implications are collected in §8.

## 1. Housing search theory: from vacancies to liquidity

The matching approach to housing starts with `arnott1989housing` (idiosyncratic tastes generate equilibrium vacancies) and
`wheaton1990vacancy` (a mismatched owner searches while carrying a vacant unit; the Nash-bargained price depends on the vacancy
rate through expected time to sell and the cost of holding two houses). `yavas1992simple` adds two-sided search intensity;
`williams1995pricing` prices heterogeneous real assets as optimal-stopping values under Poisson buyer arrivals, giving the
first explicit search-induced discount. `krainer2001theory` is the reference theory of residential *liquidity*: buyers draw
match values, sellers set reservation prices anticipating the aggregate state, and time on market lengthens in downturns because
waiting is rational; `krainer2002equilibrium` values an illiquid asset relative to a liquid benchmark as a function of arrival
intensity. `han2015microstructure` organises the field (one-sided search, random matching, directed search, bargaining and
auctions, brokerage). WP19 is a random-matching, Nash-bargaining model with a posted-price robustness check and inherits from
this strand the two objects it formalises: the sale hazard and a discount against a frictionless benchmark.

The important distinction for WP19 is that the "liquidity shock" of `krainer2001theory` is a shock to the dividend (housing
services), whereas WP19 studies shocks to matching efficiency and buyer qualification with the dividend distribution fixed.
`ngai2014hot` is the closest precedent for a non-fundamental driver: thick-market effects on match quality turn small seasonal
drivers into large price and volume swings. `novymarx2009hot` shows that feedback between the buyer–seller ratio and bargaining
positions amplifies demand shocks; `diaz2013house` (random matching, bargaining, entry) and `head2014search` (search with
construction and buyer entry, momentum from time-varying time to sell) reproduce the cyclical comovements; `ngai2020decision`,
`ngai2024ins`, `moen2021buying`, `gabrovski2019beveridge`, `caplin2011trading`, `guren2018house`, `anenberg2016information`,
`leung2017flipping` and `piazzesi2020segmented` complete the aggregate-dynamics branch. WP19 is stationary and local in scope: it
asks what a fall in $A$ or $\phi$ does to the *cross-section* of hazards and prices and to the composition of transactions.

## 2. Seller urgency and heterogeneity

The nearest antecedent of WP19's central device is `albrecht2007opportunistic`: buyers and sellers switch from a "relaxed" to a
"desperate" flow value at a Poisson rate, and the bargained price depends on the pair's states; the expected price conditional on
time on market falls with duration while its variance is hump-shaped. Other treatments of urgency are deadlines
(`salant1991sale`), the seller's discount rate (`arnold1999search`), the endogenous cost of holding two homes
(`anenberg2020endogenous`), debt overhang that *raises* asking prices (`head2023default`), and an estimated motivation parameter
(`carrillo2012empirical`). The empirical call for heterogeneous-seller theory is explicit in `glower1998motivation`: planned-move
sellers sell faster and motivation lowers the sale price but not the list markup; `springer1996motivation` finds lower prices for
eager, relocated, vacant and foreclosed sellers; `harding2003bargainingeffects` finds weaker bargaining power for vacant sellers;
`levitt2008agents` provides the marginal value of waiting (about 3.7% for about 9.5 extra days). Equity and reference dependence
work in the opposite direction: high-LTV and loss-facing sellers set higher asking prices and sell more slowly
(`genesove1997equity`, `genesove2001lossaversion`), and patient sellers sort into FSBO platforms (`hendel2009fsbo`).

In all theoretical treatments urgency is a two-state process, a deadline, a debt position or a nuisance parameter. None models a
continuous, permanent carrying-cost type whose cross-sectional distribution determines the stock of listings, the composition of
transactions and, through free entry, market tightness. That is WP19's modelling choice. The evidence on constrained sellers
implies that the carrying-cost distribution should admit very patient types (low $c$) and that the model's predictions for
"locked" sellers are limited to the case $c\ge0$ (Limitations).

## 3. Forced sales and the size of liquidity discounts

`campbell2011forced` (foreclosure discount about 27% in Massachusetts, smaller discounts for death and bankruptcy sales, about
1% spillover per nearby foreclosure), `clauretie2009foreclosure` (naive discounts above 20% overstated by about one third once
condition, endogenous marketing time and spatial dependence are controlled), `penningtoncross2006reo`, `mayer1998auctions` (price
of immediacy 0–9% in a boom versus 9–21% in a bust) and `andersen2017firesales` (sudden-death forced sales: discounts rising with
urgency, poor market conditions and seller constraints, with demand held fixed) establish the empirical range of the urgency
discount and its state dependence. Spillovers (`harding2009contagion`, `anenberg2014foreclosures`, `mian2015foreclosures`) show
that a larger mass of urgent sellers lowers prices for all sellers without a change in fundamentals, the empirical content of
WP19's liquidity-shock proposition. What this literature does not provide is the decomposition the model delivers: how much of
the discount is bargaining under high carrying cost and how much is selection of which houses transact.

## 4. Credit, qualification and the cash–mortgage gap

On the theory side, `hedlund2016illiquidity`, `hedlund2016cyclical` and `garriga2020mortgage` make search-based illiquidity a
state variable interacting with long-term defaultable mortgages; `guren2020foreclosures` show how foreclosures freeze the
non-distressed market; `eerola2018borrowing` find that a moderate credit tightening raises time on market and idiosyncratic
price dispersion; `gabrovski2021search` open a channel from credit to liquidity with multiplicity; `diaz2024housing` has wealthier
buyers paying to transact faster under competitive search; `stein1995prices` and `ortalomagne2006housing` are the
frictionless-with-credit benchmarks. On the evidence side, qualification is elastic and delays are measurable
(`anenberg2019creditavailability`, `bhutta2021insurance`, `defusco2020leverage`, `fuster2019technology`); `allen2019searchfrictions`
is the only verified Canadian structural evidence (insured mortgages as a search market).

The cash–mortgage gap is the empirical object WP19's financing extension speaks to. `reher2024mortgage` estimate an 11% mortgage
premium (1980–2017) against about 3% from a representative-seller model and find that heterogeneity in selling conditions closes
half of the gap; `han2024cash` estimate a 2–3.9% cash discount with 29% faster closing and show, with a calibrated search model,
that closing risk alone is insufficient — the risk that a relisted property fails to sell is needed, so the discount is smaller
in booms and larger markets. `asabere1992cash`, `jauregui2019fundingprice`, `aroul2023cashdiscount`, `lin2025contingencies` and
`buchak2026ibuyers` (about 5% immediacy spread) add magnitudes and selection caveats. The only formal bargaining theory with
financing is `bian2018bargaining`. No verified paper decomposes the measured gap into a bargaining component (compensation for
delay and failure risk, which depends on seller urgency) and a selection component (mortgage deals require a better match), with
the buyer's outside option differing by financing type. That decomposition is WP19's Proposition 7.

## 5. Selection, indices and volume–price comovement

The measurement literature establishes that sold houses are not a random sample of the stock and that the bias co-moves with the
cycle: `jud1994selection`, `gatzlaff1997selection`, `gatzlaff1998selection`, `hwang2004selectivity`, and above all
`goetzmann2006reservation`, where heterogeneous reservation prices make falling volume bias index returns upward and rising volume
bias them downward. In commercial real estate, `fisher2003controlling`, `fisher2007tbi` and `clayton2008liquidity` build
constant-liquidity indices that lead transaction indices and are more volatile; `lin2007illiquidity`, `bond2007marketing`,
`cheng2010illiquidity`, `lin2008real` and `cheng2013liquidityrisk` show that ignoring the uncertain marketing period biases
returns upward and risk downward, and that risk is seller-specific. The classical index tools (`bailey1963repeatsales`,
`case1989efficiency`, `clapp1992assessed`) assume random sampling. Bargaining and information confounders are documented by
`harding2003bargainingpower`, `kurlat2015asymmetries`, `garmaise2004asymmetries`; volume leads prices in `berkovec1996turnover`,
`hort2000turnover`, `clayton2010pricevolume`, `andrew2003transactions`, `leamer2007businesscycle`; `dachis2012transfertax` is the
Canadian transfer-tax experiment.

The literature recognises the bias but no verified paper derives its *sign* from primitives in a search equilibrium. WP19's
Proposition 6 does so and finds that the sign is not a general law: it depends on whether the elasticity of the sale hazard to
the arrival rate rises or falls with urgency, which is governed by the product of the hazard function and the mean residual
valuation of the match-value distribution. This nuance is new relative to `goetzmann2006reservation` and `fisher2003controlling`.

## 6. Duration dependence, censoring and withdrawals

`zuehlke1987duration`, `haurin1988duration`, `jud1996brokerage`, `anglin2003tradeoff`, `knight2002listing`,
`haurin2010listprices`, `huang2001withdrawal`, `benefield2015measurement`, `carrillo2012potatoes` and `anglin2006liquidity`
document the price–time tradeoff, positive duration dependence in parametric hazards without heterogeneity, and the sensitivity
of results to withdrawals and relistings. `taylor1999time` (stigma) and `anenberg2016information` (learning) are competing
explanations of duration effects. WP19's observed hazard is a mixture over urgency types with a competing withdrawal risk
(Lemma 3), which predicts spurious negative duration dependence; the paper must state that positive dependence in the data
requires list-price revisions, learning or stigma, which are outside the model.

## 7. Foundations and tools

Diamond–Mortensen–Pissarides architecture (`diamond1982aggregate`, `mortensen1994job`, `pissarides2000equilibrium`,
`mortensen1999new`, `rogerson2005search`, `petrongolo2001looking`, `shimer2005cyclical`); efficiency (`hosios1990efficiency`,
with the two-sided-entry caveat of `gabrovski2025note`); reservation values (`lippman1976economics`); log-concavity
(`bagnoli2005logconcave`); bargaining (`nash1950bargaining`, `rubinstein1982perfect`); directed search (`moen1997competitive`,
`wright2021directed`, `albrecht2016directed`); price dispersion (`burdett1983equilibrium`, `maury2014search`, `sagi2021asset`);
OTC asset-market liquidity premia (`duffie2005over`, `lagos2009liquidity`, `rocheteau2011liquidity`), from which WP19 differs by
the indivisibility and match-specificity of the housing unit; matching evidence (`genesove2012search`: contact-hazard elasticity
0.84 with respect to the buyer–seller ratio under CRS; `carrillo2013heat`, `carrillo2015tightness`, `han2014biddingwars`,
`han2016askingprice`, `merlo2004bargaining`, `merlo2015home`, `horowitz1992role`, `yavas1995strategic`).

## 8. What is known, what is open, and the model-design implications

Known: urgency lowers reservation and transaction prices and shortens duration; liquidity is state dependent; credit tightening
lengthens time on market and raises dispersion; cash buyers obtain discounts; transaction indices are selection-biased; tightness
amplifies shocks. Open: (i) a stationary equilibrium with a continuous urgency distribution, free entry and an explicit
frictionless benchmark; (ii) analytical conditions under which a lower arrival or qualification rate lengthens time on market
(monotone hazard) and widens dispersion; (iii) the sign and size of the index selection bias from primitives; (iv) the
decomposition of the cash–mortgage gap into bargaining, outside-option and selection terms; (v) the welfare role of bargaining
power with heterogeneous sellers and one-sided entry.

Design implications adopted in `model_specification.md`: carrying cost $c$ is a continuous permanent type with a right-skewed
distribution (most sellers patient, a minority urgent, as in the motivation and forced-sale evidence); an exogenous withdrawal
hazard with a positive own-use value $w_0$ (withdrawals and relistings matter empirically); free entry of buyers pins the buyer
value, which both isolates the seller-side comparative statics and generates tightness feedback; the baseline qualification
screen is nested in a financing-contingency extension with a pending state, cash and mortgage buyers, so that $A$ and $\phi$ are
observationally distinct only in the extension; the frictionless benchmark is defined as the Walrasian free-entry price and the
urgency discount relative to the most patient seller is the primary liquidity-discount object, because the forced-sale
literature measures exactly that comparison; match-value distributions include IFR cases (truncated normal, Weibull), the
exponential boundary and a heavy-tailed Pareto case to display the conditions of Propositions 2 and 6; listing inflow is
exogenous (one-sided entry) so that the Hosios benchmark applies in the homogeneous case and heterogeneity is the only wedge.

## 9. Novelty gate decision

Propositions 1 and the directional content of 2–4 restate known comparative statics in a new setting and are labelled as
benchmarks in the manuscript, with `albrecht2007opportunistic`, `krainer2001theory`, `lippman1976economics`,
`eerola2018borrowing`, `ngai2014hot` and `novymarx2009hot` credited as antecedents. The contributions that survive the audit are
the joint treatment (continuous urgency × free entry × pending financing), the Pareto counterexample to monotone hazards with
its exact condition, the index-selection proposition with its sign condition, the three-term decomposition of the cash–mortgage
gap, and the numerical welfare analysis with heterogeneous sellers. The gate is passed with these labels; see `novelty_matrix.csv`.