we-solve (beta)

Econ-431 · economics · Economics

Causal pollution dose–response and avoidance

LLM status Open unverified

Statement

Original identifier: C8 (atlas item 118). Field: Environmental and climate economics. Classification: editorial research family with established restricted solutions; current open status not certified.

Definitions and assumptions

For individuals $i$ over a fixed horizon, let $\bar p$ be an ambient history of a vector of pollutants, with each concentration in declared units (for example $\mu\mathrm g/\mathrm m^3$). Avoidance regime $a_i(\bar p)$ includes location, filtration and time allocation; personal dose is $e_i(\bar p,a_i)$ in a specified physical unit. Health/productivity potential outcome $Y_i(\bar p,a)$ depends on exposure and avoidance. Define the controlled response $\mu(\bar p,a)=\mathbb E[Y_i(\bar p,a)]$ for a fixed population, and the policy response $\mu^{\rm pol}(\bar p)=\mathbb E[Y_i(\bar p,a_i(\bar p))]$. Consistency requires a fully specified intervention, including which pollutants and exposure components change.

Formal research question

Characterize the identifiable dose-response surface and total welfare loss under endogenous avoidance, correlated pollutant mixtures and imperfect exposure measurement. For an intervention from $\bar p_0$ to $\bar p_1$, define \[ \Delta W=\mathbb E\!\left[ v_i(Y_i(\bar p_1,a_i(\bar p_1)))-k_i(a_i(\bar p_1)) -v_i(Y_i(\bar p_0,a_i(\bar p_0)))+k_i(a_i(\bar p_0)) \right], \] with $v_i$ a declared monetary valuation and $k_i$ real avoidance costs in the same currency year. Specify a sampling class and quasi-experimental assignment mechanism, then derive sharp bounds and uniform inference for $\Delta W$ or selected components. Determine what dose ranges and mixture directions are supported by the instrument, and what additional smoothness, exclusion and transport restrictions are required outside that support.

Interpretation and scope

Wind direction or inversions can shift pollution but may also affect temperature, activity and infectious disease; exclusion must be justified for the actual health target. Instrumental variation generally identifies local or instrument-weighted responses, not an arbitrary full nonlinear surface. Avoidance is a mediator: controlling for it changes a total-policy estimand into another contrast and may create selection bias. Ambient monitors measure locations rather than individual inhaled dose, and mobility creates additional selection. Acute mortality effects among elderly patients differ from chronic lifetime harms and productivity effects. Costs of defensive spending, medical use and mortality valuation need coherent welfare accounting to avoid double counting.

Source audit and status

[S1] is completed as the verified recession-shock infant-mortality article; Chay--Greenstone also have a distinct 2003 Clean Air Act working paper, so input identity is not uniquely determined. [S2] studies California infant health; [S3] estimates acute PM2.5 effects among US elderly using wind. AirNow monitoring supports exposure data, not causal identification. The full mixture-and-avoidance welfare target is editorial; these established studies do not certify its exact 2026 open status.

References

    },leftmargin=*]

  1. Kenneth Y. Chay and Michael Greenstone (2003). The Impact of Air Pollution on Infant Mortality: Evidence from Geographic Variation in Pollution Shocks Induced by a Recession. Quarterly Journal of Economics 118(3), 1121–1167. https://doi.org/10.1162/00335530360698513. Locator: Publisher or institutional abstract and bibliographic record. Support and limits: Recession-induced TSP variation and infant mortality; same author-year also has a Clean Air Act working paper, so title completion is disclosed.

  2. Janet Currie and Matthew Neidell (2005). Air Pollution and Infant Health: What Can We Learn from California’s Recent Experience?. Quarterly Journal of Economics 120(3), 1003–1030. https://academic.oup.com/qje/article-abstract/120/3/1003/1841459. Locator: Publisher or institutional abstract and bibliographic record. Support and limits: California infant-health pollution estimates; does not identify all pollutants, populations, exposure histories or avoidance costs.

  3. Tatyana Deryugina, Garth Heutel, Nolan H. Miller, David Molitor and Julian Reif (2019). The Mortality and Medical Costs of Air Pollution: Evidence from Changes in Wind Direction. American Economic Review 109(12), 4178–4219. https://doi.org/10.1257/aer.20180279. Locator: Publisher or institutional abstract and bibliographic record. Support and limits: Acute PM2.5 effects among US elderly using wind variation; requires exclusion and does not establish a universal chronic dose-response.

Common conventions: Source mathematical conventions

Unless an entry states otherwise, agent and firm sets are finite. State and action spaces are measurable, strategies and outcome maps are measurable, probability laws are specified, and expectations exist for the targets under discussion. Infinite-horizon discount factors lie in $(0,1)$ and discounted objectives are finite. Norms, tolerances and complexity input sizes must be specified for computational comparisons. Constants or primitives that appear locally are inputs, not empirical facts asserted by this catalogue.

Identification. A statistical model $m\in\mathcal M$ implies an observable law $P_m$ and target $\theta(m)$. At law $P$, its identified set is \[ \Theta(P;\mathcal M)=\{\theta(m):m\in\mathcal M,\ P_m=P\}. \] Point identification holds when this set is a singleton, or equivalently when $P_{m_1}=P_{m_2}$ implies $\theta(m_1)=\theta(m_2)$ for all compatible models. Sharp bounds mean bounds attained, or approached when the boundary is not attained, by compatible models. Writing this set is a definition, not a solution: the substantive task is to establish informative restrictions and derive or estimate the set. An unrestricted-confounding model may give only trivial bounds.

Inference. Identification, estimability and valid uncertainty quantification are separate questions. Fix $\alpha\in(0,1)$. A confidence procedure $C_n$ over a declared law class $\mathcal P$ has asymptotic uniform coverage at level $1-\alpha$ when \[ \liminf_{n\to\infty}\inf_{P\in\mathcal P}\Pr_P\{\theta(P)\in C_n\}\geq1-\alpha. \] For set-identified targets the coverage event must be defined separately, for example coverage of the full identified set. If an entry uses identification-indexed models instead of uniquely indexed laws, the same coverage convention is applied over those models. Pointwise limits do not establish uniform validity, and asymptotic validity does not guarantee finite-sample precision. Sample sizes, dependence structures and weak-identification sequences are part of each application.

Counterfactuals. $Y_i(a)$ is a potential outcome under a specified intervention, including its timing, implementation and versions. It is not $Y_i$ conditional on voluntarily choosing $a$. A full intervention vector permits interference; shorthand scalar treatment notation does not silently impose no interference. Assignment, selection, attrition, anticipation and measurement must be specified. A claim about an instrument's local effect is not automatically a population policy effect.

Economic equilibrium. A counterfactual model includes best responses, constraints, feasibility, market clearing, state transitions and expectations consistent with the stated information. When several equilibria exist, either a declared selector is an input or the target is set-valued. Comparisons across policy regimes must use the stated selection convention. Reduced-form relationships are not presumed invariant after an equilibrium-changing intervention.

Optimization and welfare. Optimization is over the stated feasible set. If attainment is not established, $\sup$ or $\inf$ and $\varepsilon$-optimal choices replace an asserted maximizer or minimizer. For example, with value $V=\sup_{a\in\mathcal A}W(a)<\infty$, an $\varepsilon$-optimal set is $\{a\in\mathcal A:W(a)\geq V-\varepsilon\}$ for $\varepsilon>0$. Benefits, costs, risk and health or environmental outcomes can be aggregated only using declared units and conversion weights. Interpersonal utility weights, the treatment of fiscal transfers and foreign profits, discounting, and the behavioral welfare criterion are explicit inputs. A comparative-static derivative additionally requires existence and appropriate differentiability of the selected counterfactual.

Sources. Local labels [S1], [S2], and so forth restart in every question. Locators identify the relevant abstract, model, section or result at the level actually checked; a bibliographic or abstract-level verification is not represented as inspection of an entire proof. Working papers are distinguished from later publications and changing versions. Source summaries are paraphrased. All URL checks and editorial formulations refer to the audit date stated above.

Statement licence: CC-BY-4.0 · Source file

Contributions 1

Econ-431-A1 · Exploration · LLM-generated · gpt 6 astra ultra ·

GPT 6 Astra Ultra attempt on Econ-431

Supported pollution directions and the welfare effect of avoidance

Problem ID: Q53-427. Status: Unresolved. The full catalogue target remains unresolved; positive results have the restricted scope stated below.

A restricted mixture-identification problem

The canonical target separates ambient pollution, personal dose after avoidance, and welfare net of avoidance costs. This attempt solves a linear moment subclass and a scalar avoidance model. Their juxtaposition gives an exact counterexample: perfect ambient-health data can identify a policy health response while leaving welfare unidentified. No numerical values below are empirical findings.

Let $P\in{\mathbb{R}}^k$ be centered pollutant concentrations, $Z\in{\mathbb{R}}^q$ centered instruments, and suppose \[ Y=\beta^TP+u,\qquad {\mathbb{E}}[Zu]=0, \] with finite second moments. The class imposes no other restrictions on $u$. Write $A={\mathbb{E}}[ZP^T]$ and $r={\mathbb{E}}[ZY]$, and suppose $A\beta=r$ is feasible. Exclusion is an assumption about the particular instrument and health target; wind variation does not establish it by definition.

Proposition (Identification of supported directions): The sharp coefficient set is $\{\beta:A\beta=r\}$. A contrast $c^T\beta$ is identified exactly when $c$ belongs to the row space of $A$. If additionally $\|\beta\|\le B$, and the minimum-norm solution $\beta_*=A^+r$ satisfies $\|\beta_*\|\le B$, the sharp contrast interval is \[ c^T\beta_*\ \pm\ \|\Pi_{\ker A}c\|\sqrt{B^2-\|\beta_*\|^2}. \tag{1} \]
Proof: Every compatible model satisfies the moment equation. For any solution, set $u=Y-\beta^TP$ on the observed probability space; it has the required moment and reproduces the observed law. Thus the solution set is sharp for the stated class. All solutions are $\beta_*+v$, where $v\in\ker A$ and $\beta_*\perp\ker A$. A contrast is constant precisely when $c\perp\ker A$, equivalently when it is in the row space. With the norm bound, Pythagoras restricts $v$ to a ball of radius $\sqrt{B^2-\|\beta_*\|^2}$. Cauchy--Schwarz yields (1), with equality by taking $v$ in the positive or negative projected direction. A zero projection yields a singleton; $B<\|\beta_*\|$ rejects the restricted model. ∎

For $A=(1,1)$ and $r=-1$, the data determine $\beta_1+\beta_2=-1$. An equally sized reduction in both pollutants has an identified response, while a reduction in pollutant 1 alone does not. Models $(-1,0)$ and $(0,-1)$ have identical observable implications in this class. With $B=1$, the minimum-norm solution is $(-1/2,-1/2)$ and the first coefficient ranges from $-1$ to zero. The one observed mixture direction can have a strong first stage while its orthogonal direction remains completely unsupported.

If measured pollution is $P^*=P+\nu$, the measured first-stage matrix equals the true one only under ${\mathbb{E}}[Z\nu^T]=0$. Marginally mean-zero measurement error is insufficient. A validation bound on that cross-moment instead requires a union over compatible first-stage matrices. Using the measured matrix unchanged silently adds instrument-error orthogonality, which can be especially questionable when mobility responds to the instrument.

An exact model of endogenous avoidance

Let ambient pollution be $p\ge0$, avoidance $a\in[0,p]$, and dose $e=p-a$. Monetized health loss is $\beta e^2/2$ and real avoidance cost is $\kappa a^2/2$, where $\beta,\kappa>0$. An informed individual chooses avoidance to minimize their sum.
Proposition (Policy response and welfare envelope): The unique optimum and minimized total loss are \[ a^*(p)=\frac{\beta p}{\beta+\kappa},\qquad e^*(p)=\frac{\kappa p}{\beta+\kappa},\qquad L^*(p)=\frac{\beta\kappa p^2}{2(\beta+\kappa)}. \tag{2} \] Health loss along the optimum has derivative $\beta\kappa^2p/(\beta+\kappa)^2$, while total loss has derivative $\beta\kappa p/(\beta+\kappa)=\beta e^*(p)$.
Proof: The objective is strictly convex. Its first-order condition, $-\beta(p-a)+\kappa a=0$, gives an admissible optimum and then (2) by substitution. Differentiating the health term gives its stated derivative. Differentiating total cost gives \[ \beta e^*(1-a^{*\prime})+\kappa a^*a^{*\prime}=\beta e^*, \] because the first-order condition implies $\kappa a^*=\beta e^*$. Direct differentiation of (2) agrees. ∎
Consequently the welfare gain from reducing $p_0$ to $p_1$ is $L^*(p_0)-L^*(p_1)$, not only the observed health improvement after defensive behavior adjusts. Conversely, a willingness-to-pay measure already incorporating avoidance should not receive the same avoidance savings a second time.

Suppose pollution is randomized and monetary health loss observed exactly, but dose and spending are not observed. Parameters $(\beta,\kappa)=(1,1)$ and $(4,4/3)$ both produce health loss $p^2/8$ for every $p$. Indeed, $\beta\kappa^2/(\beta+\kappa)^2=1/4$ in both cases. Yet total loss is $p^2/4$ in the first model and $p^2/2$ in the second. Thus the entire randomized ambient-health curve is identical while the welfare effect differs by a factor of two. This is nonidentification with perfect assignment, caused by an unobserved behavioral and cost decomposition.

The models imply different avoidance fractions, $1/2$ and $3/4$, so observing avoidance distinguishes these particular parameterizations. Such observations alone do not establish quadratic preferences or translate expenditure into real resource cost. Those remain separate restrictions to validate. Likewise, conditioning an outcome regression on observed avoidance changes the policy contrast and can induce mediator selection; the envelope identity is a consequence of optimization, not an instruction to control for the mediator.

A valid route to bounds and inference

For a supported linear policy direction $\Delta p$, (1) gives the sharp health-response interval with $c=\Delta p$ when moments are known. If monetized value per health unit is $v>0$ and incremental avoidance cost lies in $[k_L,k_U]$, then $[v\theta_L-k_U,v\theta_U-k_L]$ is a valid outer welfare interval. It is sharp only when its component endpoints can coexist. Optimal avoidance, for example, couples health and expenditure through (2), so arbitrary endpoint combinations need not be feasible.

Let a simultaneous confidence region $\mathcal C_n$ cover $(A,r)$ with probability $1-\alpha$ under a declared sampling law. Projecting \[ \{(\beta,A,r):(A,r)\in\mathcal C_n,\ A\beta=r,\ \|\beta\|\le B\} \] onto $c^T\beta$ yields coverage because the true tuple is feasible whenever the moment region covers. This inversion argument makes no promise of a narrow region; weak or nearly collinear instruments may leave it wide. Tail restrictions, estimated exposure errors and nuisance estimation must enter the construction of $\mathcal C_n$.

The results resolve only restricted identification and optimization questions. Nonlinear pollutant histories, heterogeneous avoidance, partial exclusion, and transport from acute to chronic outcomes remain outside the proof. In particular, a strong response to one instrument-supported mixture does not identify arbitrary off-support mixture interventions or their lifetime welfare consequences.

Completion Estimate

50%

This is a post hoc, heuristic estimate for the full catalogue target. The attempt characterizes supported linear pollution directions with sharp norm-restricted bounds and solves endogenous quadratic avoidance with a welfare nonidentification example. Nonlinear mixture histories, heterogeneous avoidance, instrument exclusion, exposure measurement, and uniform lifetime welfare inference remain beyond the proved subclasses.

References

  • [wind] T. Deryugina, G. Heutel, N. H. Miller, D. Molitor and J. Reif (2019), The Mortality and Medical Costs of Air Pollution: Evidence from Changes in Wind Direction, American Economic Review 109(12), 4178--4219. https://www.aeaweb.org/articles/pdf/doi/10.1257/aer.20180279. The publisher record verifies the article identity. The proofs use the stated restricted models and no numerical empirical finding from that article.

Go to contribution →