---
title: Empirical Welfare Maximization (EWM)
url: https://www.emergentmind.com/topics/empirical-welfare-maximization-ewm
type: topic
---

# Empirical Welfare Maximization (EWM)

Empirical Welfare Maximization (EWM) is a statistical policy learning paradigm designed to select among treatment assignment rules with the explicit objective of maximizing population social welfare, usually defined as mean potential outcome under a given policy. EWM formalizes policy learning as a combinatorial optimization problem over a structured class of rules, grounded in principles of empirical process theory and robust causal inference. EWM unifies aspects of experimental design, machine learning classification, and econometric treatment choice; it has become central in the literature on optimal treatment assignment, welfare targeting, and data-driven resource allocation.

## 1. Formalization of the EWM Principle

Let $\{(X_i, D_i, Y_i)\}_{i=1}^n$ be an observed i.i.d. or sequential sample, where $X_i$ are individual covariates, $D_i \in \{0,1\}$ is treatment, and $Y_i$ is the realized outcome, corresponding to potential outcomes $(Y_i(0), Y_i(1))$. A policy $\pi:\mathcal{X} \to \{0,1\}$, or equivalently a measurable "decision set" $G \subseteq \mathcal{X}$, prescribes treatment assignment based on $X$. The population social welfare under $\pi$ is
\[
W(\pi) = \mathbb{E}[Y_1 \cdot \pi(X) + Y_0 \cdot (1 - \pi(X))]
\]
where the expectation is over the joint distribution of $(X, Y_1, Y_0)$. In the presence of unconfoundedness, $W(\pi)$ can be expressed using doubly robust or inverse-probability-weighted (IPW) estimators.

EWM proceeds by constructing an empirical analogue $\widehat{W}_n(\pi)$—via plug-in, IPW, or doubly robust scoring—and selecting
\[
\widehat{\pi}_{EWM} \in \arg\max_{\pi \in \Pi} \widehat{W}_n(\pi)
\]
where $\Pi$ is a user-defined policy class, often of finite VC-dimension [2510.26723; 2011.04993; 2103.15298].

## 2. Estimation and Computational Methodology

Central to EWM is the empirical welfare criterion. For randomized or unconfounded settings, the canonical doubly robust estimator is
\[
\widehat{W}_n^{DR}(\pi) = \frac{1}{n} \sum_{i=1}^n \left\{ \widehat{m}_1(X_i) \pi(X_i) + \widehat{m}_0(X_i) (1-\pi(X_i)) + \frac{D_i - \widehat{e}(X_i)}{\widehat{e}(X_i)(1-\widehat{e}(X_i))} \left( Y_i - \widehat{m}_{D_i}(X_i) \right) \left(\pi(X_i) - \frac{1}{2}\right) \right\}
\]
where $\widehat{e}(\cdot)$ and $\widehat{m}_d(\cdot)$ are estimated propensity score and outcome regression, respectively [2510.26723]. Other variants include IPW or plug-in estimators depending on the sampling design and identification strategy [2103.15298; 2011.04993].

The EWM optimization is combinatorial: maximizing (sample means of) linear functionals of the form $\sum a_i 1\{ X_i \in G \}$ over $G \in \mathcal{G}$, where $\mathcal{G}$ is the feasible class (e.g., threshold sets, decision trees). This is NP-hard for general classes but tractable for low-complexity or one-dimensional threshold rules via grid search [2011.04993; 2404.11767]. Recent work establishes an exact equivalence between EWM and least-squares prediction over pseudo-outcomes in a suitable function class, enabling convex relaxation and scalable, regularized policy fitting [2510.26723].

## 3. Statistical Properties and Regret Guarantees

EWM's statistical validity relies on uniform deviation bounds for empirical processes indexed by the policy class. For $\Pi$ of VC-dimension $V<\infty$, the expected regret satisfies
\[
\mathbb{E}[W(\pi^*) - W(\widehat{\pi}_{EWM})] = O\left( \sqrt{V/n} \right)
\]
where $\pi^*$ is the welfare-maximizing (oracle) policy in $\Pi$ [2510.26723; 2011.04993]. For threshold rules in regular nonparametric models, regret sharpens to $O(n^{-2/3})$ under additional smoothness and margin conditions (cube-root asymptotics) [2404.11767]. In dynamic or time-series settings with mixing or martingale conditions, regret bounds of similar $O(\sqrt{V/n})$ form hold under appropriate invariance and exogeneity properties [2205.03970].

In instrumental variable (IV) models with endogeneity, social welfare is represented as a function of the marginal treatment effect (MTE) kernel:
\[
W(G) = \mathbb{E}[Y_0] + \mathbb{E}[1\{Z \in G\} \cdot \int_0^1 \mathrm{MTE}(u,X) du ]
\]
EWM is applied by maximizing the empirical analogue of this integral, with the regret rate governed by both the complexity $v$ of $\mathcal{G}$ and the uniform estimation rate of $\widehat{\mathrm{MTE}}$ [2012.07624; 2206.09883]. When MTE is estimated at $\sqrt{n}$-rate (parametric or low-dimensional IV), EWM recovers the $O(\sqrt{v/n})$ regret rate; otherwise, the convergence is limited by the slower MTE estimation rate.

## 4. Extensions: Constraints, Robustness, and Alternative Welfare Criteria

EWM can incorporate explicit constraints, such as budget or fairness restrictions. With a per-unit cost $c(X)$ and budget $B$, the population-constrained problem is
\[
\max_{G \in \mathcal{G}} W(G) \quad \text{s.t.} \quad C(G) \leq B
\]
where $C(G) = \mathbb{E}[c(X) 1\{X \in G\}]$. Empirical analogues directly replace $W, C$ with sample estimates. However, Naive sample-analogue constrained EWM exhibits failures in uniform feasibility and efficiency: no rule can achieve both asymptotically across all DGPs. Remedies include tightening constraints with critical values (size control) or penalizing constraint violation (trade-off rules) [2103.15298; 2206.09883].

EWM generalizes to alternative social welfare functionals, notably α-Expected Welfare Maximization (α-EWM), which targets the lower-tail mean (CVaR) of the post-treatment outcome distribution over the worst-off α-fraction:
\[
W_\alpha(\pi) = \frac{1}{\alpha} \int_0^\alpha F_\pi^{-1}(u) du
\]
where $F_\pi$ is the distribution function of $Y(\pi)$. Estimation leverages dual representations and cross-fitted, doubly robust scores. Regret analysis reveals a $O(\sqrt{V/n})$ rate, with constants inflating as $\alpha \to 0$ [2505.00256]. This framework covers Rawlsian and distributionally robust welfare optimization.

Time-series EWM (T-EWM) adapts the machinery to sequential or nonstationary data. It defines welfare objectives as conditional expectations over policy-induced paths and maximizes empirical IPS-weighted welfare along observed trajectories. Theoretical guarantees extend to martingale and Markov-type processes [2205.03970].

## 5. Policy Classes, Threshold Rules, and Implementation Protocols

The choice of policy class $\Pi$ fundamentally impacts EWM's empirical behavior and feasibility. Common classes include

- Threshold rules: Scalar or multivariate policies of the form $\pi_t(x) = 1\{x > t\}$ or Cartesian products of indicator thresholds over selected coordinates. Regret rates and asymptotics are well understood for this class [2404.11767; 2011.04993].
- Linear scores and finite-depth decision trees: Used for interpretability and tractability.
- Set-indicator policies over VC-classes: General framework covering most practical applications.

Implementation is feasible with grid-search (for low-dimensional threshold rules), mixed-integer programming (for more complex policies), or, via the equivalence with least-squares, convex optimization for large-scale settings [2510.26723]. Standard protocol entails (i) estimating individual-level causal effects (e.g., via regression-adjustment, doubly robust estimation, or IV), (ii) evaluating empirical policy-specific welfare over a defined grid or function class, and (iii) selecting the maximizer and reporting welfare and treatment group trade-offs [2011.04993].

## 6. Empirical Applications and Illustrations

EWM has been validated in a range of empirical settings:
- Threshold-based welfare program eligibility using job training (LaLonde) data [2011.04993], showing welfare gains over random assignment and enabling policy menus parameterized by interpretable thresholds.
- Medicaid expansion eligibility under budget constraints, where trade-off rules outperform naive constrained EWM in terms of welfare-efficiency and controlled budget violation [2103.15298].
- Optimal tuition subsidy assignment under endogeneity, using estimated MTE in the Indonesian Family Life Survey; EWM (FEWM/BEWM) rules target subpopulations with high predicted gains within budget [2206.09883].
- Dynamic pandemic response policies, where T-EWM estimated adaptive COVID-19 restriction rules with empirical regret improvements confirmed in both simulation and real-world weekly data [2205.03970].
- Distributionally robust targeting (α-EWM), shifting treatment to disadvantaged subpopulations, with formal inference on lower-tail welfare [2505.00256].

## 7. Connections to Plug-in Policy Learning and Regularization

EWM and the plug-in approach—assigning treatment to those with positive estimated CATE—are theoretically equivalent under suitable reparameterization [2510.26723]. Specifically, EWM can be formulated as least squares regression of a pseudo-outcome on the class $g(x) = 2\pi(x) - 1$, yielding an exact correspondence between maximizing empirical welfare and minimizing square error within the policy class. This equivalence enables the design of convex, regularized training algorithms, circumventing the NP-hardness of discrete optimization without loss of statistical guarantees. Regularization enhances stability, enables large-scale implementation, and accommodates additional convex constraints (budget, fairness) via joint convex optimization.

---

**References**

- "Welfare Analysis via Marginal Treatment Effects" [2012.07624]
- "Empirical Welfare Maximization with Constraints" [2103.15298]
- "Policy Learning under Endogeneity Using Instrumental Variables" [2206.09883]
- "Policy Learning with $α$-Expected Welfare" [2505.00256]
- "Policy Choice in Time Series by Empirical Welfare Maximization" [2205.03970]
- "Regret Analysis in Threshold Policy Design" [2404.11767]
- "Optimal Policy Learning: From Theory to Practice" [2011.04993]
- "Bridging the Gap between Empirical Welfare Maximization and Conditional Average Treatment Effect Estimation in Policy Learning" [2510.26723]

Source: https://www.emergentmind.com/topics/empirical-welfare-maximization-ewm