The Endowment Effect

In the AI Story

The endowment effect was formalized through a series of experiments by Tversky, Kahneman, and Richard Thaler in the late 1980s and early 1990s. The 1990 paper in the Journal of Political Economy and the 1991 Journal of Economic Perspectives article 'Anomalies: The Endowment Effect, Loss Aversion, and Status Quo Bias' established the effect as a robust and consequential violation of the Coase theorem and related neoclassical predictions.

The mechanism of the endowment effect is typically explained through loss aversion: relinquishing an endowed object is coded as a loss and therefore weighted more heavily than the equivalent gain from acquiring it. Possession changes the reference point, and reference-point dependence does the rest. Recent work has proposed alternative mechanisms, including reference-point ambiguity and psychological ownership, but loss aversion remains the dominant explanation.

In the AI transition, the endowment effect produces a paradox: the most experienced professionals, who are best positioned to benefit from AI amplification of their judgment, are often the most resistant to AI adoption. The objective gain (amplification of judgment) is genuinely larger than the objective loss (devaluation of implementation skill), but the subjective comparison is dominated by the endowment effect. The implementation skill, being deeply endowed, is overvalued; the amplified judgment, not yet endowed, is undervalued. The expert evaluates a bad trade because the endowment has distorted the comparison.

The process-level dimension compounds the bias. The expert does not only value the output of expertise (working code, correct diagnosis) but the process of producing it — the hours of concentrated effort, the struggle with difficulty, the embodied satisfaction of having earned the result. This process is itself an endowment, separate from the skill and separate from the output, and its elimination by AI is evaluated as an additional loss. The Orange Pill's geological metaphor — layers of understanding deposited through debugging — captures exactly this: the process is the deposition, and AI skips the deposition while delivering the surface.

Origin

The endowment effect was first hypothesized by Thaler in 1980, drawing on his earlier work on mental accounting. The systematic experimental program that established its robustness was conducted jointly with Tversky and Kahneman through the late 1980s.

The Coase theorem had predicted that initial allocation of property rights would not affect final allocation under conditions of costless bargaining. The endowment effect showed this prediction to be empirically false: initial allocation matters, because it determines reference points, and reference points determine valuations.

Key Ideas

Possession inflates valuation. Ownership produces a gap between willingness-to-accept and willingness-to-pay that classical economics cannot explain.

Investment depth scales endowment. The deeper the investment in a skill, identity, or relationship, the greater the endowment inflation — and the more resistant the holder to trades that seem objectively favorable.

Process as endowment. The process of producing expertise is itself endowed, separate from the expertise and separate from its output; AI automation of the process produces a loss that output-level analysis does not capture.

The paradox of experience. The most experienced professionals, best positioned to benefit from AI, are often the most resistant — because their deep endowments produce the largest distortions.

Coase theorem violation. Initial allocation of rights matters because reference-point dependence converts allocation into valuation.

Debates & Critiques

Whether the endowment effect reflects loss aversion, ownership psychology, ambiguity aversion, or multiple overlapping mechanisms remains debated. Recent experimental work has isolated conditions under which the effect weakens or disappears, suggesting it may be less universal than initially thought. But in the high-stakes, identity-relevant context of professional expertise, the effect appears robust.