The claim that marginal utility theory should be abandoned rests on a set of real limitations that need to be examined individually.
Case for the claim — the theory's limitations. First, utility is not cardinally measurable. MU theory assumes that a consumer can compare, say, 15 utils against 24 utils across goods and across units. In reality no utils meter exists. Second, the theory assumes rational maximising behaviour — that consumers compute MU for every available bundle, know their preferences precisely and never err. Behavioural economics has documented systematic violations: loss aversion (prospect theory shows losses loom roughly twice as large as gains), framing effects (identical choices are made differently depending on how they are presented), anchoring, present bias (hyperbolic discounting, which the standard model excludes), reference-dependence, mental accounting, limited attention and status-quo bias. Third, the theory assumes perfect information, which is clearly unrealistic for experience, credence or complex financial goods. Fourth, the model is static — no habit formation, no learning, no reference-point updating. Fifth, interpersonal comparisons of utility are not permitted methodologically, which limits the direct normative application of the theory. On this list, the theory looks like a stylised, unrealistic account of a messy human reality — which is the basis for the claim that it has been superseded.
Case against the claim — the theory's enduring strengths. First, MU theory correctly predicts the direction of demand responses to price changes in nearly every empirically estimated market. Downward-sloping demand curves are an empirical regularity, and MU theory derives that regularity from axioms — not many theories in economics have that structural achievement. Second, the equi-marginal condition MUx/Px=MUy/Py remains the correct marginal rule for allocating a budget: a consumer who violates it is leaving utility on the table, and this logical result does not depend on cardinal measurement — the ordinal reformulation (indifference curves) delivers the same tangency condition between the marginal rate of substitution and the price ratio. Third, most behavioural anomalies are reformulations of utility maximisation, not replacements for it: prospect theory IS a utility function, with a kink at a reference point and loss aversion built in. Hyperbolic discounting IS maximisation of a utility function with a different discounting rule. The behavioural literature extends the framework; it does not reject the idea of a maximised value function. Fourth, the theory is analytically indispensable for welfare economics — consumer surplus, deadweight loss, the welfare effects of taxes and subsidies — all are grounded in some version of MU. Fifth, MU theory has very high pedagogical and expositional value: it is the simplest and clearest route to the downward-sloping demand curve and the first exposition of utility maximisation that students meet.
Reasoned judgement. The appropriate verdict is not abandonment but supplement. The limitations named by critics are real, but they do not demolish the theory — they identify the conditions under which it is weakest. The theory should continue to be used WHERE its assumptions are approximately met and SUPPLEMENTED where they are not.
Conditions under which MU theory remains the right tool. First, for AGGREGATE market analysis — where individual departures from rationality tend to wash out across many consumers and the law-of-demand prediction holds. Second, for ORDINARY GOODS over NORMAL price ranges — where behavioural anomalies are small and well-known MU-driven effects dominate. Third, for the DERIVATION of the demand curve and for CONSUMER SURPLUS analysis — where no equally parsimonious alternative exists. Fourth, for POLICY ANALYSIS of taxes, subsidies and price controls in standard markets — where the welfare calculus runs on MU-based consumer surplus.
Conditions under which MU theory should be SUPPLEMENTED. First, choices under UNCERTAINTY — where prospect theory outperforms expected-utility maximisation. Second, decisions dominated by DEFAULTS, FRAMING or PRESENT BIAS — where behavioural interventions (nudges, defaults, commitment devices) work on channels the standard model does not see. Third, EXPERIENCE and CREDENCE goods where information is a first-order concern. Fourth, decisions over LONG HORIZONS where habit formation, learning and reference-point updating matter. In these cases the standard MU model should be used alongside behavioural models, not instead of them.
Teaching and the discipline. Discarding MU theory would require teaching consumer theory without an account of why demand slopes down, without a derivation of consumer surplus, and without the equi-marginal rule. No existing alternative delivers all of these in equally transparent form. The theory should therefore remain the baseline of consumer theory, and behavioural economics should be taught as the extension that handles the cases where the baseline breaks down.
Verdict. The claim that MU theory's limitations are so serious that it should no longer be used is rejected on balance, under the condition that the theory is used as a baseline model for ordinary markets and ordinary goods and is supplemented by behavioural, informational and dynamic extensions where its assumptions clearly fail. The correct formulation is 'MU theory is a limited but durable foundation — not a complete explanation, but still the right starting point'.