RESEARCH DOSSIER
Search Frictions and Bilateral Bargaining
Abstract. Decentralized exchange combines the search for feasible counterparties and the negotiation of mutually acceptable contractual terms. This monograph examines how costly information, stochastic encounters, waiting, private valuations, and institutional restrictions shape reservation utilities, negotiated prices, allocation, and welfare. It reconstructs sequential stopping rules, generalized Nash bargaining, alternating-offer equilibrium, two-sided matching, and the benchmark private-information limits to efficiency. It compares evidence from employment, financial and durable-asset markets, mortgage contracts, retail electricity, transport, and agricultural exchange, distinguishing survey descriptions, experimental effects, and structural counterfactuals. The analytical synthesis establishes why economically credible alternatives, rather than market size or observed prices alone, are necessary for the interpretation of negotiated outcomes; why contact and acceptance hazards require separate measurement; and why a reduction in search friction can change price, access, and welfare through different mechanisms. Numerical examples are explicitly illustrative and all empirical magnitudes are attributed to their original studies. The dossier integrates established scholarship without asserting a new theory or original causal estimates.
Keywords: search costs; bilateral bargaining; matching; outside options; reservation utility; alternating offers; private information; negotiated prices; price dispersion; matching externalities; market power; structural estimation; welfare
Research Object and Disciplinary Boundary
Exchange in a decentralized market requires the creation of a feasible trading relationship before parties can agree on its terms. A friction is a resource cost, information restriction, waiting requirement, or institutional obstacle that separates a potentially beneficial exchange from its realization. Search describes the acquisition of contacts and information. Matching describes the formation of a relationship in which exchange becomes possible. Bargaining describes a procedure by which the parties select the terms of exchange, including the decision to decline it.
A market can be large in aggregate and bilateral at the point of agreement. A mortgage borrower may have several potential lenders but negotiate with only one at any instant. An employer may advertise a vacancy to many workers and eventually contract with one. An aircraft owner may hold a specialized durable asset while waiting for a buyer who values its operating characteristics. These relationships illustrate different technologies of finding, screening, and contracting. They do not imply a single empirical distribution of prices or waiting times.
The economic object of this dossier is the interdependence of matching frictions, reservation utilities, and negotiated prices. Reservation utility is the value of an attainable alternative under the relevant information and time structure. A reservation price is the price at which a participant is indifferent between the proposed transaction and that alternative, after accounting for any transaction-specific costs. Bargaining power can refer to a surplus-sharing parameter, a strategic advantage in the offer protocol, or an empirically measured response to improved alternatives. These concepts must be distinguished: an impatient buyer can obtain a poor price despite having many potential counterparties; a well-informed buyer may receive a favorable price even when the formal bargaining rule is unchanged.
The benchmark of competitive equilibrium supplies prices at which compatible demands and supplies can be coordinated without representing the costly sequence of meetings. Search and bargaining research opens this coordination mechanism to analysis. The relevant contrast is between a market's potential opportunity set and the subset that agents can actually access at a given cost. The main scientific question is how decentralized access produces prices, transaction probabilities, selection, and welfare. The subject belongs to microeconomic theory, information economics, and contract theory; unemployment, financial trading, and development economics enter as applications of the same narrowly specified mechanisms.
The investigation uses theoretical reconstruction under stated assumptions and comparative review of published research. It neither estimates an original population parameter nor treats an illustrative calculation as observed data. Studies are selected for mechanism coverage, not as a systematic census of all papers. The historical evidence spans the formation of bargaining theory, the analysis of information costs, equilibrium matching, and modern empirical identification.
Historical Formation of Search and Bargaining Theory
The modern research program contains two related analytical lineages. Nash's bargaining problem (1950) made the utilities delivered by feasible agreements and the disagreement point central objects of a solution concept. Its axiomatic approach identifies a division of benefits from a relationship once the opportunity and relevant alternatives have been specified. The contribution concerns two-person agreement; it does not by itself specify the distribution of future trading contacts.
Stigler (1961) treated price discovery as economically costly. A dispersed distribution of quotations creates a benefit to canvassing additional sellers, balanced against the resources required to obtain information. His historical illustrations concern quotations and their comparability: differences in accompanying services can account for part of observed price dispersion. The distinction remains fundamental when constructing a price measure from electronic records, where different delivery dates, warranties, or financing conditions may be bundled with ostensibly identical goods.
McCall's job search study (1970) developed the sequential choice problem in employment. Diamond's model of price adjustment (1971) illustrated how a seemingly small positive search friction can support monopoly pricing under restrictive assumptions. Burdett and Judd (1983) showed that equilibrium price dispersion can arise in markets with homogeneous rational participants when the process reveals a variable number of quotations. Information acquisition thus shapes the sellers' equilibrium strategy as well as the buyer's stopping rule.
Rubinstein's strategic bargaining game (1982) replaced the unspecified act of agreement with a sequence of offers and responses. Rubinstein and Wolinsky's sequential bargaining market (1985) embedded strategic negotiation in a stochastic encounter process. Current agreements depend on the terms and timing of alternatives that a participant may obtain after leaving the present relationship. Their model also illustrates that vanishing frictions need not generate the competitive allocation under every market and entry specification. Such convergence is a result to establish within a model rather than an assumption that follows from the number of traders.
The private-information branch of the literature developed in parallel. Myerson and Satterthwaite (1983) analyze bilateral trade under hidden reservation values. Hosios (1990) and later matching research examine the consequences of individual search and entry decisions for aggregate welfare. Pissarides (2000) organizes equilibrium unemployment around the creation and destruction of trading relationships. Modern empirical work adds records of counteroffers and unsuccessful negotiations, randomized information treatments, and structural models of repeated transactions. The historical unity of the field arises from a common question: which of the potentially productive encounters become agreements, on what terms, and at whose cost?
Sequential Search and Reservation Values
Consider a risk-neutral buyer facing independent price draws P from a known distribution F. Assume one unit of a homogeneous good, value v, no discounting, recall of the best available quote p, and constant real resource cost k for another draw. A buyer who will purchase at the best previously observed price can achieve an expected one-draw saving of E[(p-P)^+], where x^+ = max(x,0). For a continuous distribution supported on nonnegative prices this saving equals the integral of F(t) from zero to p. An additional opportunity has value because the buyer can retain the old quote when the new one is worse.
In the stationary iid model with free recall, finite expected prices, and an interior solution, the optimal reservation price p* satisfies
The buyer stops once the best quote is at or below p*. For a buyer still deciding whether to participate, the outside activity and the option not to buy must also be included; the equation describes the conditional purchasing benchmark, not an automatic requirement to transact. With P uniformly distributed on [0,100], the expected saving for p in that interval is p squared divided by 200. At k=2, the threshold is the square root of 400, or 20. At k=8, it is 40. The probability of a successful draw is respectively 0.2 and 0.4. If draws are counted including the successful one and the process starts without a quote, expected counts are five and two and a half. These are synthetic values. They display the tradeoff between accepted price and search expenditure across two assumed cost regimes.
A sequential employment benchmark uses discounted payoffs. Suppose unemployed worker utility per period is b, an offer w is drawn each period, an accepted job provides w forever, and the next period starts after the current unemployed payoff. Let beta be the discount factor within (0,1), U unemployment value at the beginning of a period before receiving b, and W(w)=w/(1-beta). Then
The reservation wage w*=(1-beta)U satisfies w*=b+beta E[(w-w*)^+]/(1-beta). The formula follows by writing max(U,W)=U+max(W-U,0), multiplying by 1-beta and substituting. This is an explicit pedagogical reconstruction of the one-sided job search tradition rather than a representation of every assumption in McCall's original model. If unemployment benefits expire, job destruction is possible, offers arrive at different rates, or learning changes the wage distribution, the stationary rule must be replaced by a state-dependent value comparison.
The term reservation value is sometimes used loosely for an individual's primitive valuation. Search theory makes the distinction consequential. A worker's continuation value depends on expected hiring, alternative earnings, location, and timing. A seller's continuation value depends on expected future buyers and carrying costs. A liquid buyer and a credit-constrained buyer can have equal preferences for an asset yet differ in credible willingness to wait. A better outside option need not entail an increase in intrinsic taste for the current match.
Several qualifications matter. Search with recall differs from irreversible rejection. Correlated offers change the informational return to an additional draw. Finite deadlines change the threshold even under fixed search costs. A posted quote may cease to apply, and costly verification may alter net value. The observable number of contacts is an endogenous result of both search effort and opportunities; low search counts can indicate either high costs or immediate discovery of a satisfactory partner. These interpretations imply different policies, even when transaction prices look similar.
Bilateral Bargaining and Strategic Agreement
Let v be the buyer's gross match value and c the seller's real cost. A payment p yields buyer utility v-p and seller utility p-c. Let d_B and d_S be disagreement utilities expressed in the same units and over the same horizon. The joint surplus over disagreement is S=v-c-d_B-d_S. Trade with positive gains is feasible when S>0 and the price lies between c+d_S and v-d_B.
With transferable utility and a generalized Nash weight alpha in (0,1) assigned to the buyer, the selected feasible price maximizes (v-p-d_B)^alpha times (p-c-d_S)^(1-alpha). Differentiating the logarithm with respect to p, setting the derivative to zero, and checking strict concavity on the interior yields
Here the gains g_B and g_S are incremental to disagreement. The symmetric special case reflects Nash's original axiomatic approach; the weighted expression is a standard extension. With v=100, c=40, d_B=10, d_S=5 and alpha=0.6, the surplus is 45. The buyer receives 27 of incremental gain and the seller 18, with price 63. If only d_B increases by 5, price declines to 61. The total incremental surplus falls to 40; the buyer receives 24 incremental gain but total utility increases from 37 to 39. A strengthened alternative can improve a participant's absolute negotiated utility while reducing that participant's surplus measured against the improved alternative.
The derivative of p with respect to d_B is -(1-alpha); with respect to d_S it is alpha. These are partial effects holding the division weight and other values fixed. In a market equilibrium all four values and the distribution of matched pairs can move. A policy changing search technology may alter price through several channels simultaneously. Moreover, an explicit option to quit negotiation is not automatically the same as a payoff received during disagreement. Under an alternating-offer procedure, a nonbinding quit option does not necessarily shift the stationary division. The timing of exit, its irreversibility, and the availability of an outside activity during delay belong in the description of the game.
In the complete-information alternating-offer benchmark, two players divide one unit of transferable surplus. Player A offers first, the other player either accepts or counteroffers after one period, and the respective discount factors are delta_A and delta_B in (0,1). Let x be A's equilibrium share when A offers, and y B's share when B offers. The responder must obtain at least the next-period discounted continuation. Thus x=1-delta_B y and y=1-delta_A x. Substitution gives
In this model agreement occurs immediately; the threat of delay determines its terms. With identical discount factor delta, x=1/(1+delta). At delta=0.9, the initial proposer receives approximately 0.5263. For unequal continuous-time discount rates r_A and r_B and interval Delta, set delta_i=exp(-r_i Delta). The limit as Delta tends to zero is r_B/(r_A+r_B). The more impatient responder concedes more. These results reconstruct the fixed-discount case in Rubinstein (1982); they assume common knowledge and the prescribed feasible actions.
If negotiations actually involve several offers, this does not directly estimate an agent's impatience. Revisions can reflect private valuations, delays in verification, strategic signaling, interruptions, or requests for authorization. The analyst must distinguish the model's theoretical division of surplus from the behavioral mechanism generating a counteroffer. The two are separate objects of estimation.
Equilibrium Matching and Endogenous Outside Options
A market-wide matching technology closes the relationship between individual alternatives and the aggregate availability of partners. Let U denote unmatched workers or sellers and V denote vacancies or buyers, with both stocks strictly positive. Consider a constant-returns matching function
The parameter A is match efficiency. Tightness theta=V/U indexes the abundance of buyers relative to sellers. Per-worker contact intensity f=M/U=A theta^(1-eta), and per-vacancy filling intensity q=M/V=A theta^(-eta). These are continuous-time hazards, not probabilities bounded above by one. In a discrete period of length h, a constant Poisson rate f corresponds to probability 1-exp(-fh), provided other competing events are absent.
These equations are illustrative abstractions associated with the matching tradition developed in Equilibrium Unemployment Theory (Pissarides 2000). The matching function counts contacts that meet the definition imposed by the model; in a richer specification some of these are rejected after quality is revealed. Stocks, contact rates, completed trades, and active search intensity must consequently be reported separately.
For a risk-neutral employment benchmark with discount rate r>0, constant job destruction rate s, productivity y, outside flow b, wage w, and vacancy flow cost kappa, let U_0 be the value of unemployment, W employment value, J filled job value, and V_0 a vacancy value. In a stationary continuous-time model with homogeneous matches and a constant outside payoff after separation,
Free entry implies V_0=0 for active vacancies and kappa=qJ. The wage bargain allocates the total capitalized gains W-U_0+J-V_0. Under a worker share eta_w, the rule W-U_0=eta_w(W-U_0+J) determines a wage jointly with entry. Let X=W-U_0. Then (r+s)X=w-b-fX and (r+s)J=y-w. Combining the equations and substituting free entry yields the familiar benchmark wage relation
Indeed, fJ=f kappa/q=kappa theta. Thus the term kappa theta reflects the value of the aggregate opportunity environment in the wage formula. The derivation assumes match surplus is positive and free entry is feasible. It does not apply unchanged with bargaining over hours, capital specificity, firing taxes, or heterogeneous productivity.
Equilibrium tightness solves the entry condition alongside the wage and continuation equations. If the labor force is normalized to one, with unemployed share u and employed share 1-u, the stationary stock-flow equation is s(1-u)=f u, hence u=s/(s+f). The matching elasticity of unemployment, eta, is distinct from the worker's bargaining weight eta_w. The two coincide only when imposed by a specified efficiency condition or institutional mechanism.
Two-sided search entails an equilibrium fixed point. An increase in buyer entry can raise sellers' meeting intensity, reduce each buyer's meeting intensity, alter the expected price and induce further entry or exit. Greater bargaining power for one side can stimulate its participation while reducing participation by the other. A partial-equilibrium increase in a reservation value therefore cannot be extrapolated directly into the market-wide welfare effect of a new search technology.
Private Information and Failed Trade
Complete information about gross valuation and costs allows an immediate feasible exchange whenever joint gains over disagreement are positive in the linear benchmark. Actual parties often observe only their own valuation, production cost, urgency, or capacity to walk away. An offer is then both a proposed division and a signal about privately known information. Revealing high valuation can weaken the buyer's negotiating position; revealing low cost can weaken the seller's position. Strategic concealment and the risk of mutually beneficial disagreement arise from these incentives.
Myerson and Satterthwaite's bilateral trading analysis (1983) establishes a benchmark impossibility under independent private values, overlapping valuation supports and the relevant regularity conditions: full ex post efficiency conflicts with Bayesian incentive compatibility, interim individual rationality, and the absence of an external subsidy. The result concerns mechanism constraints, not an assertion that every private negotiation fails or that every actual mechanism attains an optimal second best. Disjoint supports with always-positive gains, public values, or a relaxation of participation or subsidy constraints change the problem.
A transparent numerical counterexample separates the private-information friction from the costs of real search. Suppose v and c are independent uniform draws on [0,1] and valuations are private. A fixed price of 1/2 with voluntary acceptance trades if c<=1/2<=v. Trade probability is 1/4. The unconditional expected gain from trade is the integral over this rectangle of v-c; conditional means are 3/4 and 1/4, producing an unconditional gain of 1/8. Under full-information efficient allocation, trade occurs when v>=c, with probability 1/2. The unconditional expected gain is the triangular integral of v-c equal to 1/6. Thus the simple fixed-price rule yields 75% of the full-information expected surplus. It is not asserted to be the optimal incentive-compatible mechanism. Its lost surplus can be calculated without claiming that the parties would reveal their values under another mechanism.
Search can relieve or intensify informational barriers. Multiple credible offers may help establish an alternative price and discipline strategic exaggeration. Conversely, greater ability to wait can encourage an agent to hold out for a better offer and change the composition of rejected counterparts. Information about an external market price does not necessarily reveal the current counterparty's private cost. A party may possess public information but lack the means to exploit it.
The empirical distinction between inefficient rejection and efficient sorting is especially important. A contract that fails because v<c has avoided a socially costly transfer of resources. A contract that fails because parties cannot agree despite v>c has lost a feasible gross gain. The latter criterion still requires adjusting for verification, search, and enforcement resources. A measurement protocol needs either observed values, credible bounds, or a model of the latent valuation distribution. Treating all failed negotiations as inefficiencies would be an error.
Posted Prices, Negotiated Prices, and Dispersion
Posted-price and negotiated-price institutions organize search differently. A posted price is an announced term that an eligible buyer can accept as offered. Negotiation creates a channel for the seller to condition the eventual contract on information conveyed through interaction. A market may combine posted terms, quote requests, bargaining, and auctions. The observability of each stage is essential for distinguishing the price mechanism from the allocation mechanism.
In a restrictive benchmark of sequential search with identical sellers, buyers who expect all firms to charge the same price incur a positive cost for further search if a seller slightly raises its price within the search-cost tolerance. This logic is associated with Diamond's 1971 equilibrium analysis; it illustrates why arbitrarily small positive search costs do not universally guarantee competitive pricing. By contrast, the noisy-search process analyzed by Burdett and Judd allows firms to compete for purchasers who receive multiple offers, while retaining some captive demand from buyers who receive one. Dispersed prices can be equilibrium outcomes under such conditions. Neither result establishes that any given empirical price difference is caused by search.
A list-price discount measures the distance from an announced number, not buyer surplus. If an identical good is posted at 100 and negotiated to 90, a buyer with value 92 obtains gross surplus 2. Another buyer paying 95 for an asset worth 130 obtains 35. The second pays more but gains more. A retailer can post a high reference price and offer a large discount while collecting a higher final price than a rival. Comparative analysis therefore requires a matched contract specification, alternatives actually available to the buyer, fees, and expected quality.
Price dispersion should be decomposed before attributing it to market power. Observable heterogeneity in product, contract maturity, delivery speed, credit risk, and service may account for different prices. Residual variation can reflect unobserved quality, timing, information, stochastic bargaining, or differential outside options. The covariance between the number of offers and the final price is not a causal effect of competition: buyers with demanding preferences may both search more and purchase expensive specialized goods; buyers in urgent circumstances may search less and pay high prices; sellers may disclose more quotes to buyers who are likely to leave.
Repeated interaction introduces an intertemporal margin. A firm can quote a favorable acquisition price to win a customer and later exploit costly switching, while competitors anticipate future earnings when competing for initial business. The relevant cost to a consumer is the discounted value of the contract sequence, renewal payments, and transition costs. A static one-period discount ranking is insufficient when contract renewal creates lock-in. The appropriate market comparison must follow acquisition and renewal cohorts, distinguish a competing quote from a credible switch, and account for customers who exit the observed panel.
Welfare and Matching Externalities
Welfare involves the resources consumed by search, waiting, failed coordination, and intermediation as well as the allocation of goods and services. For an isolated voluntary agreement with quasi-linear utility, a pure price transfer cancels in unweighted buyer-plus-seller surplus. If the resource cost of finding the buyer is k, the contemporaneous joint gain is v-c-k, irrespective of p within the feasible interval. Distributional evaluation remains sensitive to p, and equilibrium quantity, entry, and search intensity may respond to the expected distribution.
In matching markets, an individual entrant creates benefits for some alternatives and congestion for entrants on the same side. With the homogeneous constant-returns function above, a marginal unmatched worker increases total matching by partial M/partial U=eta M/U=eta f; a marginal vacancy increases it by partial M/partial V=(1-eta)q. If the transfer rule rewards workers with eta_w of joint matching surplus and vacancy owners with 1-eta_w, the equality eta_w=eta aligns these margins in the standard frictional benchmark. This is the intuition of the Hosios efficiency condition (1990), under the remaining assumptions for constrained efficient entry and bargaining. The simple marginal-product argument is explanatory rather than a proof of unrestricted optimality. It abstracts from changing match quality, other entry distortions and redistribution.
When additional buyers also change expected match surplus, the standard condition no longer internalizes every externality. Mangin and Julien (2021) generalize the condition by accounting for the effect on expected gains from the match as well as its probability; their generalized Hosios condition applies to the particular dynamic environments specified in their paper. A market with apparently appropriate bargaining weights can still generate inefficient entry when the composition of matched agents changes. Consequently the elasticity of total matching is not interchangeable with the elasticity of expected value produced by the marginal searcher.
A proposed comparison must distinguish at least three benchmarks: the unconstrained frictionless allocation, the allocation attainable under the same search and information technology by a planner, and the best feasible mechanism under private information and participation constraints. Measured distance to a frictionless first best combines the cost of the technology and inefficient use of that technology. A policy may improve constrained welfare while leaving substantial frictionless losses. A claim about efficient negotiated prices is uninterpretable without identifying the benchmark.
A numerical illustration makes the difference between prices and resources clear. A buyer values delivery now at 120, the seller can produce it at cost 70, and a bilateral meeting requires resource expenditure 6. Suppose the fee charged by a matching platform is 10 and its real processing cost is 3. If the fee is a pure transfer to the platform and the meeting expenditure and processing cost are separate, total surplus is 41. Counting the fee itself as a further social resource loss double-counts part of the surplus. If the fee discourages otherwise worthwhile participation, however, lost trades create an additional welfare effect at the level of the market. The distinction between transfers and distortions has to be maintained in both theory and measurement.
Measurement and Empirical Identification
A useful dataset is a history of opportunities as well as transactions. Define the unit of observation as an identified search spell, a contact, an individual offer, a bargaining sequence, a contract, or a renewal; the correct choice depends on the question. Each record should contain the time stamp, market and institution, participant-side identifier under appropriate privacy restrictions, product characteristics, quoted terms, availability of outside options, recorded costs, sequence outcome, and a reason for censoring or exit when observed. Prices should be expressed on the same basis: currency and date, taxes, total maturity-adjusted cost, contract quantity, service quality, and any rebates or mandatory fees.
Hazard analysis estimates the rate of acceptance conditional on remaining at risk. If every contact follows a Poisson arrival process with rate lambda and an independent fraction a of contacts is accepted, the completed-trade hazard is lambda a. Observing only transaction times identifies that product under the benchmark, not lambda and a separately. A completed-transaction hazard of 0.2 per month could arise from one contact per month with acceptance probability 0.2 or 0.25 contacts per month with acceptance probability 0.8. The expected time to completion is five months in both cases. Contact logs or an independently identified shift in acceptance are needed to distinguish the mechanisms. When hazard rates vary with elapsed time, the exponential formula should not be imposed without testing.
The U.S. Bureau of Labor Statistics' JOLTS data definitions illustrate why measurement conventions matter. An opening requires an available specific position, the possibility of starting within 30 days, and active external recruiting. Hires count monthly additions to payroll, including flows that do not correspond one-for-one to the snapshot of vacancies. JOLTS openings are a stock at the end of the month; hires are a flow over the month. Dividing one by the other does not, without timing and coverage assumptions, yield an exact vacancy-filling probability. Direct estimation of microeconomic acceptance rates further requires applications, offers, and rejections, which these aggregate series do not supply.
Experimental identification can manipulate one dimension of the opportunity set while holding other visible attributes constant. Examples include randomized disclosure of market prices, offers sent by a comparison service, or scripted statements of outside alternatives. The primary reduced-form estimand is the intention-to-treat difference between intervention and comparison groups. Treatment take-up, price exposure, compliance, and equilibrium spillovers distinguish that effect from the effect of actually searching. If households communicate information across experimental groups, outcomes depend on other units' treatment as well as their own. Randomizing at a village or market level, measuring network links, and reporting exposure definitions can help address this issue.
Observational methods require a defensible exclusion or comparison. A distance measure may capture geographic search costs, but it can also affect transport expenditure and product quality. A new marketplace can lower matching costs while selecting traders and changing seller composition. A licensing reform can induce both entry and quality changes. An instrumental-variable estimate is causal only under the relevant independence, exclusion, and other identification conditions; a difference-in-differences design additionally requires a credible counterfactual trend and attention to anticipation and staggered treatment.
Structural analysis specifies primitives such as valuation distributions, the meeting technology, the cost of effort, and the bargaining protocol; computes equilibrium and maps it into offers, delays, and transaction quantities; then chooses parameters to fit data. Its advantage is coherent welfare and policy counterfactuals. Its limitation is that multiple primitives may produce similar observables, especially when only successful matches are recorded. Identification should be evaluated using variation in contacts, offers, durations and exit, auxiliary restrictions, sensitivity to alternative protocols, and out-of-sample prediction. A fit to final prices alone cannot identify both reservation values and bargaining weights because they enter the same negotiated-price formula.
A credible empirical audit retains failed contacts, reports right censoring and competing risks, checks whether negotiations are independent or clustered, and tests comparability across product definitions. The scope of uncertainty must include sampling variability, structural specification, and external validity. The present monograph reviews evidence available in the identified papers; it does not reproduce their microdata or likelihood functions.
Comparative Evidence Across Markets
The empirical literature provides complementary perspectives because institutional environments expose different stages of the matching and bargaining process. A historical review must record what each source observes, what it estimates, and under which benchmark the result is interpreted.
Employment in the United States
Hall and Krueger's worker survey of wage formation (2012) reports that approximately one third of respondents bargained over pay when accepting their current job, nearly one third knew exact pay at their initial meeting, and about 40 percent were able to remain in their earlier job when accepting the new one. The survey was designed around a representative sample of U.S. workers and current or most recent employment. These figures describe reported institutional experience in that survey; they are not rates of successful bargaining among all job applications or contemporary global employment statistics. An employer's negotiation procedure and whether a worker has an existing job are related but distinct dimensions. The study supports modelling coexistence of posted and negotiated employment terms.
Financial and durable-asset markets
Duffie, Gârleanu, and Pedersen's over-the-counter market analysis (2004 working paper; 2005 journal article) describes asset-price and dealer-spread determination when investors locate counterparties and negotiate rather than all trading against a centralized continuous market. An investor's ability to transact promptly depends on search access; a dealer's inventory and capacity to offer immediacy can itself be a source of surplus. A spread can compensate for productive matching services and also reflect the intermediary's bargaining position.
In the business-aircraft application, Gavazza's 2015 structural asset-market study estimates allocation and transaction effects using information about aircraft offered for sale, trading outcomes, dealer inventory, and retail and wholesale prices. The discussion-paper abstract reports estimates of 18.3 percent misallocation, prices 19.2 percent below the Walrasian benchmark, and a total welfare loss of 23.9 percent relative to that benchmark. These are model-based results from the identified 2015 version, not raw sample frequencies or universal asset-market magnitudes. The comparison of dealer and no-dealer counterfactuals also shows why greater immediacy does not guarantee a net aggregate welfare improvement when operating costs and diversion of direct matches are included. The final article appeared in 2016; the numerical attribution remains deliberately attached to the accessed discussion paper rather than assuming an unchanged final specification.
Canadian mortgage negotiation
Allen, Clark, and Houde's negotiated-price mortgage research (2019) combines mortgage market data and a search-and-negotiation model to study search costs, loyalty, and lenders' market power. It provides a setting where competition in an aggregate banking market is filtered through the borrower's ability to obtain and mobilize competing offers. The paper's identified abstract supports the qualitative mechanism; this review does not assert its inaccessible table-specific estimates.
Allen and Li's dynamic mortgage pricing research (2025) analyzes Canadian contract data, documenting an acquisition-versus-renewal pricing pattern and estimating a dynamic model with search and switching frictions. Its value for comparative analysis is the explicit continuation state created by a previous contract. A policy affecting initial comparison costs can change the profitability and pricing of later renewals, so a one-contract study may miss an important response. The Canadian mortgage institutions, contract structure, and sample cannot be transplanted into U.S. or other countries' refinancing markets without further evidence.
Wholesale used vehicles
Larsen's wholesale auto bargaining study (2021 issue publication, online 2020) uses approximately 265,000 alternating-offer sequences after ascending auctions to estimate or bound latent valuations and compare realized outcomes with feasible efficiency frontiers. The reported efficiency loss is 12–23 percent of available gains from trade, and more than half of failed negotiations are estimated to involve positive gains. This study differs from analyses based on transaction prices alone: failed negotiations and the auction history help inform the bounds. The ex ante private-information frontier lies relatively close to the ex post efficient frontier in the studied setting, while actual bargaining falls short of both. It is consequently inappropriate to attribute the entire estimated loss to the theoretical impossibility theorem. The results concern wholesale used-car transactions under that auction and negotiation procedure.
Retail electricity in Australia
Byrne, Martin, and Nah's retail electricity field experiment (2022) generates price-offer sequences using scripted agents contacting retailers. Random variation in signals about customers' ability and willingness to search and switch helps isolate price discrimination based on perceived alternatives, holding other scripted characteristics constant. The study distinguishes incumbent and entrant behavior and investigates differences in outcomes for low-income subsidy recipients. Its reported interpretation links the regressive pattern in that market to differences in search and negotiation instead of direct targeting of subsidy recipients. The outcome is context-specific; the audit identifies the effects of experimentally manipulated signals in the retail institution studied, not a universal effect of income or an independently estimated purchasing behavior of every participant.
Transport in India and agricultural sales in Ghana
Keniston's Jaipur autorickshaw working paper (January 2017 draft) studies repeated negotiations with hidden valuations in a local transport market. Its author describes the theoretical welfare ambiguity of replacing haggling with fixed prices: a cheaper procedure may reduce costly negotiation, while negotiation may permit exchange between heterogeneous participants excluded by a fixed price. This working paper is considered a research design and mechanism example, rather than a final peer-reviewed estimate. Its scope is local travel in Jaipur, not national transport pricing.
Soldani and colleagues' Ghana market information experiment (2023) uses randomized provision of commodity price information and inter-village communication data. For yam, the first-year price received by the treated farmers was reported to increase by 9 percent, while other crops did not exhibit the same impact. The possible indirect effects in the second year are more uncertain, and the authors advance bargaining spillovers as an interpretation. The comparison across crops demonstrates that the market structure and prevalence of negotiation condition the value of price information. The result should not be generalized to all commodities or equated with a 9 percent increase in agricultural welfare: output volumes, costs, counterparties, and information spillovers also matter.
The cases identify mechanisms using different evidence. A survey describes the incidence of institutions; transaction histories inform bargaining efficiency; an audit manipulates perceived search options; randomized information changes actual access; structural estimates trace equilibrium counterfactuals under a model. Their numerical results have different denominators and causal interpretations. Pooling them would create an artificial average with no well-defined target population.
Institutional Design and Distribution
Institutions change the economic value of waiting and the ability to turn a possible alternative into a credible one. A searchable database can improve awareness without reducing the administrative burden of a transfer. Standardized disclosure can facilitate comparison without making offers binding. A switching service can reduce procedural costs but may create new intermediation fees. Each intervention must be mapped to the specific decision node it changes.
Mandatory publication of final terms can make dispersion visible and facilitate comparison. In a market with publicly comparable goods and enforceable offers, this may improve the leverage of buyers who currently face high search costs. In another market, prices may depend on attributes that cannot be communicated cheaply or on confidential contract services. Disclosure may require standardization of coverage, fees and eligibility. An intervention can also change how competing firms observe each other; the net effect on competition must be investigated under the actual rules rather than inferred from transparency alone.
An intermediary can provide information, screen counterparts, guarantee performance, consolidate offers, or hold inventory. Each activity has a distinct cost and informational advantage. A fee for matching is not automatically a measure of monopoly power; it may cover services that are otherwise provided through inefficient bilateral search. Conversely, control over information and access can enable an intermediary to collect rents or steer counterparties. Evaluate search duration, contract comparability, completion, allocation, fee incidence and real cost jointly. Net benefit depends on the counterfactual search method available to different groups.
For some buyers, time and liquidity constraints make waiting disproportionately expensive. A low-income customer may have a lower ability to expend effort, tolerate service interruption, or supply documentation. The direction and magnitude of distributional effects should be measured rather than inferred from income alone. Negotiated discounts can widen gaps when informed customers secure preferential terms or narrow gaps when firms offer different prices in order to sustain trade. The social evaluation needs both access and transfers. A lower average negotiated price could coexist with an increase in rejected low-value contracts.
Labor-market institutions similarly shape credible alternatives: unemployment benefits, notice periods, restrictions on mobility, credential recognition and portability of benefits affect continuation values. A policy can increase workers' reservation utility while changing firms' vacancy creation decisions. The distinction between the gain to a current employed worker and the change in the probability that an unemployed worker finds an acceptable job is decisive for aggregate welfare.
Price caps, posted-price rules, auctions, and bargaining are alternative contracting procedures, not interchangeable policy instruments. Each changes the allocation of information rents and the incentives to exert search effort. A posted-price rule can reduce the time spent negotiating, but may prevent mutually acceptable contracts at other prices. Auctions can aggregate competition but require multiple participants who can credibly submit bids and may carry participation costs. A bargaining protocol can preserve flexibility but introduces incentives for strategic delay. The empirical evidence supports market-specific diagnosis and a clearly stated counterfactual rather than a universal institutional ranking.
Integrative Analytical Synthesis
The central conclusion of the theoretical and empirical comparison is that the value of an outside opportunity depends jointly on its quality, availability, exercise cost, timing, and credibility. The same gross alternative price can imply different reservation utilities when one buyer can switch immediately and another must navigate documentation, transport, or a liquidity constraint. Consequently, the observable competitive environment is not fully summarized by the count of firms, the existence of a market price, or the number of online listings.
Consider the generalized Nash example again. If an intervention raises only d_B by five units, with alpha=0.6, the negotiated price falls by two units. If it raises both d_B and d_S by five units, the net price effect is +1 because the seller's reservation utility contributes +3 and the buyer's contributes -2. This calculation holds gross values and the division rule fixed; the interval of feasible prices contracts by ten. It demonstrates that a price increase after lowering search friction is compatible with improvements in alternatives for both sides. Welfare cannot be inferred from the direction of the negotiated price. When the changes in outside utility derive from real reductions in search time, aggregate resource savings must be calculated in the market as a whole rather than by adding disagreement utilities as if they were newly produced goods.
A second implication is that search reform should be evaluated on reach and bargaining response separately. A comparison platform can increase the number of observed offers without changing negotiation conditional on contact if those offers are not verifiable or exercise is costly. In a different setting, even a single clearly credible outside offer can change the final agreement. The analyst can measure the first margin with contacts, quote counts, and switching success; the second with randomized changes in the content of alternatives and sequences of counteroffers. When possible, estimate both the intention-to-treat effect on participation and the conditional effect on terms with explicit selection considerations.
A third implication concerns the relationship between individual rationality and market performance. An agreement can be individually rational at the moment it is reached while the system generates inefficient waiting, excessive entry on one side, or poorly allocated matches. Equally, refusal can be individually and socially efficient if the feasible exchange creates insufficient total surplus. Private information and congestion matter at different scales, and the appropriate benchmark changes between them. These distinctions explain why evidence concerning failure in one market is not a direct test of the bargaining-weight restriction in another.
The fourth implication concerns comparisons across income groups and markets. Receipt of a low price is an incomplete proxy for welfare; failure to buy and the expenditure necessary to secure favorable terms are missing. A complete distributional assessment follows the initial opportunity set, attempted contracts, realized prices, access to service, and time costs, while allowing for the possibility that increased choice changes the composition of purchasers. Report the welfare target and denominator before comparing percentages.
These conclusions organize established results and supply independently reconstructed examples and consistency checks. They constitute an authorial interpretation of the literature. No new named mechanism, formal theory, general identification theorem, or empirical discovery is claimed. A study intending to establish such a contribution would require a separate proof or research design and independent comparison with the existing literature.
Open Scientific Questions and Research Agenda
The interaction between matching and strategic bargaining remains an active research area because the feasibility of economic alternatives is itself shaped by information design, institutions, and endogenous market behavior. A prioritized research agenda begins with the data structure rather than a preferred model.
First, future studies should distinguish contacts that cannot produce an eligible agreement from contacts that fail because parties cannot settle the terms. In a model with separate screening and bargaining stages, these events generate different hazards. Matched data from a search interface, quote system, and final contract register could establish how much variation comes from arrival, screening, acceptance, or post-agreement completion. Appropriate consent and protection of sensitive records are prerequisites to such data linkage.
Second, researchers should identify the cost of exercising, as opposed to merely learning about, an alternative. An intervention that supplies credible quotes can be compared with one that additionally reduces the cost of switching. A factorial design would separate complementary channels under random assignment, while market-level measurement could reveal spillovers. The treatment definition needs to specify the days for which a quote remains valid, terms of eligibility, and the outcome horizon.
Third, model comparisons should test the informational assumptions underlying strategic delay. Counteroffers can reflect learning, private valuations, inability to authorize a decision, or changing alternatives. Experimental variation in deadlines or verification procedures could test predictions of distinct models with observable time stamps and offer histories. A fitted distribution of delays alone does not establish a particular strategic game.
Fourth, equilibrium analyses of entry and intermediation should include heterogeneous match quality and distributional outcomes. A faster encounter technology may increase matching while changing which buyers and sellers meet. Researchers should report the joint distribution of valuations or credible proxies, real costs, and occupational or geographical group composition. Robustness should include alternative matching elasticities and bargaining rules; if the outcome changes sign across plausible specifications, report that dependence directly.
Fifth, international comparisons require harmonized units and contract definitions. The studies reviewed here cover U.S. workers, Canada mortgages, Australian retail electricity, business aircraft, wholesale automobiles, India transport, and Ghana agriculture. They establish a range of feasible empirical designs but do not provide a common global survey of negotiation. Future comparative research can harmonize observable stages of search and bargaining while preserving local institutional detail.
A final task is replication. Source-data and code access, under the relevant licensing and privacy restrictions, would permit re-estimation of reported structural parameters and observational measures. The present investigation derives benchmark identities and checks source claims within the available publication records. It leaves original empirical replication to appropriately equipped studies and explicitly avoids treating source-level verification as a substitute for that work.
Conclusion
Search and bilateral bargaining form an interdependent sequence of decentralized exchange. Search determines the distribution and timing of realizable alternatives; bargaining divides the gains of particular encounters; the anticipated terms of agreement influence participation, search effort, and entry. This feedback makes the market environment an endogenous part of the reservation utility. The division of surplus can be reconstructed in explicit benchmarks, while the welfare of the market requires additional information about allocation, waiting, resource costs and matching externalities.
The theoretical literature provides several complementary explanations. Costly information affects the decision to continue searching; alternating offers explain the strategic value of patience; private information imposes limits on efficient exchange; two-sided matching connects marginal choices with congestion and the availability of partners. Their predictions depend on the information structure and institutional protocol. Neither a claim of complete competitive convergence nor a universal presumption of bargaining inefficiency follows from the mere presence of search costs.
The empirical evidence provides a corresponding set of research designs. Wage surveys reveal heterogeneity in the institution of pay formation. Negotiation histories and latent-value estimates illuminate the efficiency of failed trade. Retail audits identify responses to controlled signals of switching and search, while information treatments can change actual negotiating conditions in local markets. Structural studies link allocations, prices, and market-wide counterfactuals. The differences in their methods and populations are economically meaningful and do not license aggregation of their numerical estimates.
The scientific conclusion is that market performance must be analyzed through both the opportunity process and the agreement process. A credible outside option is an economic object with an expected value, an exercise technology, and an institutional status. Observed price dispersion and negotiating delay are outcomes that require explanation, not quantities that independently diagnose the magnitude of frictions. Welfare evaluation must identify the feasible counterfactual, distinguish transfers from real resource costs, and account for unsuccessful contacts as well as completed transactions. These requirements establish the proper evidentiary basis for interpreting search and bargaining across markets.
References
- Allen, Jason, and Shaoteng Li. 2025. "Dynamic Competition in Negotiated Price Markets." Journal of Finance 80 (1): 561–614. DOI: 10.1111/jofi.13408. Source record
- Allen, Jason, Robert Clark, and Jean-François Houde. 2019. "Search Frictions and Market Power in Negotiated-Price Markets." Journal of Political Economy 127 (4): 1550–1598. DOI: 10.1086/701684. Source record
- Burdett, Kenneth, and Kenneth L. Judd. 1983. "Equilibrium Price Dispersion." Econometrica 51 (4): 955–969. DOI: 10.2307/1912045. Source record
- Byrne, David P., Leslie A. Martin, and Jia Sheen Nah. 2022. "Price Discrimination by Negotiation: A Field Experiment in Retail Electricity." Quarterly Journal of Economics 137 (4): 2499–2537. DOI: 10.1093/qje/qjac021. Source record
- Diamond, Peter A. 1971. "A Model of Price Adjustment." Journal of Economic Theory 3 (2): 156–168. Source record
- Duffie, Darrell, Nicolae Gârleanu, and Lasse Heje Pedersen. 2005. "Over-the-Counter Markets." Econometrica 73 (6): 1815–1847. DOI: 10.1111/j.1468-0262.2005.00639.x. Source record
- Gavazza, Alessandro. 2015. "An Empirical Equilibrium Model of a Decentralized Asset Market." CEPR Discussion Paper 10546. Paris and London: CEPR Press. Subsequently published in Econometrica 84 (5), 2016, 1755–1798, DOI: 10.3982/ECTA10847. Source record
- Hall, Robert E., and Alan B. Krueger. 2012. "Evidence on the Incidence of Wage Posting, Wage Bargaining, and On-the-Job Search." American Economic Journal: Macroeconomics 4 (4): 56–67. DOI: 10.1257/mac.4.4.56. Source record
- Hosios, Arthur J. 1990. "On the Efficiency of Matching and Related Models of Search and Unemployment." Review of Economic Studies 57 (2): 279–298. DOI: 10.2307/2297382. Source record
- Keniston, Daniel. 2017. "Bargaining and Welfare: A Dynamic Structural Analysis of the Autorickshaw Market." January 2017 working paper, author research listing, Louisiana State University. Source record
- Larsen, Bradley J. 2021. "The Efficiency of Real-World Bargaining: Evidence from Wholesale Used-Auto Auctions." Review of Economic Studies 88 (2): 851–882. DOI: 10.1093/restud/rdaa007. Source record
- Mangin, Sephorah, and Benoît Julien. 2021. "Efficiency in Search and Matching Models: A Generalized Hosios Condition." Journal of Economic Theory 193: 105208. DOI: 10.1016/j.jet.2021.105208. Source record
- McCall, J. J. 1970. "Economics of Information and Job Search." Quarterly Journal of Economics 84 (1): 113–126. DOI: 10.2307/1879403. Source record
- Myerson, Roger B., and Mark A. Satterthwaite. 1983. "Efficient Mechanisms for Bilateral Trading." Journal of Economic Theory 29 (2): 265–281. Source record
- Nash, John F., Jr. 1950. "The Bargaining Problem." Econometrica 18 (2): 155–162. DOI: 10.2307/1907266. Source record
- Pissarides, Christopher A. 2000. Equilibrium Unemployment Theory. Second edition. Cambridge, MA: MIT Press. Source record
- Rubinstein, Ariel, and Asher Wolinsky. 1985. "Equilibrium in a Market with Sequential Bargaining." Econometrica 53 (5): 1133–1150. Source record
- Rubinstein, Ariel. 1982. "Perfect Equilibrium in a Bargaining Model." Econometrica 50 (1): 97–109. Source record
- Soldani, Emilia, Nicole Hildebrandt, Yaw Nyarko, and Giorgia Romagnoli. 2023. "Price Information, Inter-Village Networks, and ‘Bargaining Spillovers’: Experimental Evidence from Ghana." Journal of Development Economics 164: 103100. DOI: 10.1016/j.jdeveco.2023.103100. Source record
- Stigler, George J. 1961. "The Economics of Information." Journal of Political Economy 69 (3): 213–225. DOI: 10.1086/258464. Source record
- U.S. Bureau of Labor Statistics. n.d. "Data Definitions." Job Openings and Labor Turnover Survey. Accessed November 17, 2026. Source record