Reference Dependence and the Structure of the WTA/WTP Gap

Technical Summary: Reference Dependence and the Structure of the WTA/WTP Gap

Problem Statement
The paper addresses the robust empirical regularity known as the endowment effect: the systematic disparity between Willingness to Accept (WTA) and Willingness to Pay (WTP) for the same good. While standard equilibrium theory predicts these values should converge when income and transaction costs are controlled, behavioral literature (e.g., Kahneman et al., 1990; Horowitz and McConnell, 2002) consistently documents large WTA-WTP ratios.

The paper identifies two primary limitations in the standard explanation provided by Cumulative Prospect Theory (CPT):

1. Lack of Structural Variation: The standard model relies on an exogenous, fixed loss-aversion parameter ($\lambda > 1$), offering no structural mechanism to explain why the gap should vary with transaction scale or familiarity.
2. Decoupling of Phenomena: Recent evidence (Chapman et al., 2024) suggests that the endowment effect and loss aversion regarding risk are decoupled, challenging the view that a single fixed parameter drives both.

Methodology
The author develops a decision-theoretic framework called Weak Rank-Dependent Utility (WRDU) to model preferences over finite lotteries under objective probabilities. The approach is axiomatic and non-parametric, avoiding assumptions of Constant Relative Risk Aversion (CRRA) or probability weighting.

The methodology rests on six key axioms:

1. Completeness, Transitivity, and Continuity: Standard rationality conditions.
2. Weak Independence: A relaxation of the von Neumann-Morgenstern independence axiom. Independence is required only for mixtures that preserve the reference point and do not cross the induced gain-loss partition. Cross-partition aggregation is mediated by a range-dependent coefficient.
3. Reference Partition: The outcome space is partitioned into gains, losses, and a reference point ($x_r$).
4. Range Dependence: The loss-aversion index is not fixed but varies with the transaction range (magnitude of gains and losses).

The core representation, WRDU, evaluates lotteries by separating gains and losses into distinct subutilities ($v_g$ and $v_\ell$) anchored at the reference point. These components are recombined via a Lagrangian penalty coefficient ($\rho$) applied to the loss-side component. The standard loss-aversion index is defined as the reciprocal: $\lambda = 1/\rho$.

The reference point $x_r$ is endogenously selected as the maximizer of a Lagrangian penalty functional:
$$\mathcal{L}(r; \rho) = v_g(z(r)) - \rho(r)v_\ell(y(r))$$
where $z(r)$ and $y(r)$ are the gain and loss magnitudes induced by a candidate reference point $r$.

Key Results

1. Structural Origin of the Gap:
   The paper derives the WTA and WTP indifference equations based on the WRDU representation.

   • WTP (Buying): $v_\ell(\text{WTP}) = \frac{1}{\lambda} v_g(m)$
   • WTA (Selling): $v_g(\text{WTA}) = \lambda v_\ell(m)$

   The paper demonstrates that if $\lambda > 1$ (equivalently $\rho < 1$), the structural asymmetry of these equations necessarily implies $\text{WTA} > \text{WTP}$, generating the endowment effect wedge. This result holds without assuming specific parametric forms for utility, provided the subutilities are normalized to be comparable across dimensions.

2. Impossibility of Attenuation with Fixed Indices:
   A key theoretical result (Proposition 1) establishes that if the reciprocal loss-aversion index $\lambda$ (or penalty $\rho$) is fixed across transaction scales, the WTA-WTP gap cannot converge to zero as transaction size increases. The wedge persists regardless of scale unless additional, external restrictions are imposed on the utility functions.

3. Range-Dependent Attenuation:
   The paper's primary mechanism for resolving the scale issue is range dependence. The coefficient $\rho$ (and thus $\lambda$) is a function of the transaction range.

   • For small or unfamiliar transactions, the model allows $\lambda(m) > 1$ (and $\rho(m) < 1$), sustaining the endowment effect.
   • As the transaction scale ($m$) increases or becomes more familiar/market-like, the model posits a path where $\lambda(m) \to 1$ and $\rho(m) \to 1$.
   • As these indices approach the neutral value of 1, the WTA and WTP equations converge, causing the gap to attenuate.

Significance and Claims
The paper claims to provide a decision-theoretic account of the WTA-WTP gap that is distinct from existing literature:

• Distinct from Rabin Calibration: Unlike the Rabin (2000) calibration result, which concerns favorable risky gambles and implies extreme risk aversion for small stakes, this paper addresses the buy-sell wedge. The attenuation here arises from the range-dependence of the Lagrangian penalty, not from the curvature of utility functions alone.
• Structural vs. Exogenous: The model replaces the exogenous, fixed loss-aversion parameter of CPT with an endogenous, range-dependent penalty coefficient. This provides a structural reason for why the endowment effect might be strong for small stakes but diminish for large or familiar transactions.
• Scope and Modesty: The author explicitly limits the scope of the claim. The result is a theorem for a "normalized admissible transaction class" satisfying the six stated axioms. It does not claim that all endowment effects stem from this mechanism, nor does it rely on CRRA utility or probability weighting. The contribution is the demonstration that a weakened independence axiom, combined with reference anchoring and range-dependent penalization, is sufficient to generate and structurally explain the WTA-WTP wedge and its attenuation.

Conclusion
The paper concludes that the WTA-WTP gap is a structural consequence of the asymmetry in buying and selling indifference conditions under reference-dependent preferences. The persistence of the gap at small scales and its attenuation at large scales are explained by the behavior of the Lagrangian penalty coefficient $\rho$ and its reciprocal $\lambda$ as they move toward their neutral value of one along specific transaction paths.