In middle and high school economics you learn that we must decide how to spend limited money. Deaton tracked detailed spending patterns to reveal how consumption behavior links to poverty. Together with John Muellbauer he created the Almost Ideal Demand System (AIDS), which predicts how spending shares on each good change when prices or income move. Policymakers can use it to calculate, for example, which households are most affected if a government raises a food tax. Deaton also showed that the standard theory that consumption is smoother than average income conflicts with data, posing what is now called the “Deaton paradox.” He argued that we must consider household-level variation and borrowing limits, sparking a shift toward using micro data in macroeconomics. His survey methods let researchers estimate local prices from “unit values,” making it possible to measure real living standards even where markets are poorly recorded. These ideas are directly applied in poverty-reduction policies and international aid design. In short, his work builds a bridge from mathematical models to the everyday rules that shape society.

Deaton’s oeuvre is distinguished by its integration of microeconomics, macroeconomics, and development economics. The 1980 Almost Ideal Demand System (AIDS) expresses expenditure shares as linear functions of log price indices, preserving theoretical consistency (budget exhaustion, first-order conditions, Slutsky symmetry) while remaining tractable for estimation. The model is now standard for evaluating VAT reforms and constructing international price indices. In papers around 1990, Deaton decomposed variances to show the mismatch between aggregate and individual data, challenging the Permanent Income and Life-Cycle hypotheses that relied on a representative agent. This “Deaton paradox” stimulated the rise of heterogeneous-agent DSGE models and partial-insurance frameworks. He also proposed treating repeated cross sections as pseudo-panels and estimating dynamic consumption functions using cohort fixed effects. For developing countries lacking price data, Deaton treated unit values as noisy price proxies and extracted market prices through regressions, enabling region-specific real poverty lines. The method underpins the World Bank’s USD 2.15/day international poverty standard. His within-household allocation work infers gender bias by observing how spending on adult goods contracts when boys versus girls are born. Quantitative techniques developed by Deaton advanced Age-Period-Cohort decompositions linking cohort parameters to aggregate welfare trends and are embedded in the World Bank’s LSMS program. In sum, Deaton provides an integrated economic framework connecting theory with the real world, micro behavior with macro outcomes, and data collection with policy application. His contributions are regarded as a cornerstone of the contemporary micro-data revolution in economics.

Deaton’s contributions statistically implement theoretical restrictions on demand systems via dual cost-function methods and gamma-class nonlinear estimation. The AIDS model emanates from a second-order translog cost expansion; its log price index P* absorbs curvature, facilitating estimation. In “Saving and Liquidity Constraints” (Econometrica 1991) he used Euler-equation moment tests to show the failure of excess-smoothness under liquidity constraints and CARA utility. By exploiting cross-sectional variance, he identified idiosyncratic shocks that vanish in aggregates and applied DIF-GMM to wipe out cohort effects, striking a blow to representative-agent models. Addressing the Gibson–Rozelle critique, he proposed hedonic quality controls in unit-value regressions, refining price indices for developing countries. His adult-goods method detects intra-household gender bias, while the fusion of EKS-superlative indices with PPP adjustments yields welfare-dominance rankings robust to shocks. A Bayesian hierarchical survey-to-survey reconciliation further aligns heterogeneous samples. These techniques are embedded in ICP, LSMS, and DHS datasets, dramatically shrinking poverty-estimate variance. Deaton’s methodological legacy underpins HANK macro models and distributional national accounts, establishing a two-way feedback loop between measurement and theory.