Treatment Effects with Targeting Instruments

Bernard Salani\'e; Sokbae Lee

arxiv: 2007.10432 · v7 · submitted 2020-07-20 · 💰 econ.EM · stat.ME

Treatment Effects with Targeting Instruments

Sokbae Lee , Bernard Salani\'e This is my paper

Pith reviewed 2026-05-24 14:31 UTC · model grok-4.3

classification 💰 econ.EM stat.ME

keywords instrumental variablestreatment effectsmultivalued treatmentstargetingcomplier groupspartial identificationHead Startselection bias

0 comments

The pith

Targeting relations between instruments and treatments allow identification of treatment effects for composite complier groups.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a framework using the concept of targeting to handle selection bias when treatments take multiple discrete values. Targeting specifies which instruments affect which treatment levels, enabling conditions for identifying counterfactual averages and treatment effects among groups that respond to particular instrument-treatment pairs. This is useful for applications with multivalued treatments where standard binary instrument methods fall short. The authors show how a positive selection assumption can tighten the resulting bounds. Revisiting the Head Start Impact Study, they obtain bounds suggesting smaller benefits from program expansions compared to earlier parametric findings.

Core claim

Conditions based on targeting allow counterfactual averages and treatment effects to be point-identified or partially identified for composite complier groups. An additional positive selection assumption provides further identifying power. In the Head Start application, the resulting bounds indicate less beneficial effects of expansions than parametric estimates suggest.

What carries the argument

Targeting, the mapping from instruments to the treatments they affect, which separates composite complier groups sufficiently for identification.

If this is right

Counterfactual averages are identified for composite complier groups under the targeting conditions.
Treatment effects for these groups are point- or partially-identified.
Positive selection strengthens identification, leading to informative bounds.
Bounds on Head Start effects are less positive than parametric estimates.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

This method could extend to designing instruments in other multivalued treatment settings like job training programs.
Policy evaluations might use targeting to obtain more robust bounds rather than relying solely on point estimates.
Empirical work could test targeting relations directly from data to apply these identification results.

Load-bearing premise

The targeting relation between instruments and treatments is known or can be established to separate the composite complier groups as required for identification.

What would settle it

Observing that the estimated bounds for Head Start treatment effects include the parametric estimates or finding that the targeting conditions do not hold in the data would challenge the paper's identification results.

read the original abstract

Multivalued treatments are commonplace in applications. We explore the use of discrete-valued instruments to control for selection bias in this setting. Our discussion revolves around the concept of targeting: which instruments target which treatments. It allows us to establish conditions under which counterfactual averages and treatment effects are point- or partially-identified for composite complier groups. We explore the additional identifying power of a positive selection assumption. We illustrate its usefulness by revisiting the findings of Kline and Walters (2016) on the Head Start Impact Study. We derive informative bounds that suggest less beneficial effects of Head Start expansions than their parametric estimates.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Targeting organizes multivalued IV identification around instrument-treatment links and yields tighter bounds on Head Start than prior parametric work, but the abstract leaves the key conditions unexamined.

read the letter

The targeting concept is the main new element. By specifying which instruments target which treatments, the authors can define composite complier groups and derive conditions for point or partial identification of counterfactual averages and treatment effects. This builds directly on standard instrumental variables assumptions without apparent circularity. What the paper does well is to take this organization and add a positive selection assumption to sharpen the results. The revisit of the Head Start Impact Study then yields bounds suggesting smaller effects from expansions than the original parametric estimates in Kline and Walters. That provides a concrete illustration of how the framework can change conclusions in a policy-relevant setting. The soft spots are mostly around implementation. The weakest assumption appears to be that the targeting relation is known or can be established in a way that sufficiently separates the groups. If establishing targeting requires strong prior information, the method may not apply as broadly as hoped. Since only the abstract is available, I cannot check the formal derivations or see how sensitive the bounds are to the exact positive selection setup. The abstract claims the bounds are informative, but without the steps it is hard to judge their tightness or credibility. This work is for economists doing applied work with multivalued treatments and discrete instruments, such as in education, health, or labor programs. Readers interested in extending IV methods to these cases without new data collection will get the most out of it. I would recommend sending it for peer review. The idea is clear from the abstract, the application is well-chosen, and the claims are specific enough that referees can evaluate the technical contributions properly.

Referee Report

1 major / 0 minor

Summary. The paper introduces the concept of targeting to organize the relationship between discrete instruments and multivalued treatments. It claims to derive conditions under which counterfactual averages and treatment effects are point- or partially-identified for composite complier groups, shows that a positive selection assumption adds identifying power, and applies the framework to the Head Start Impact Study to obtain informative bounds indicating smaller effects than the parametric estimates in Kline and Walters (2016).

Significance. If the targeting-based identification results hold, the framework would extend standard IV methods to multivalued treatments in a structured way, with direct relevance to policy evaluations involving program expansions. The empirical bounds on Head Start effects would be noteworthy if they survive scrutiny of the underlying assumptions.

major comments (1)

[Abstract] Abstract: the central claim that targeting separates composite complier groups sufficiently for point or partial identification cannot be assessed because the abstract supplies none of the formal definitions, explicit assumptions, or derivations required to check internal consistency or the absence of hidden restrictions.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their comments on our paper. We address the single major comment below.

read point-by-point responses

Referee: [Abstract] Abstract: the central claim that targeting separates composite complier groups sufficiently for point or partial identification cannot be assessed because the abstract supplies none of the formal definitions, explicit assumptions, or derivations required to check internal consistency or the absence of hidden restrictions.

Authors: We agree that the abstract, as currently written, is high-level and does not contain the formal definitions of targeting, the explicit identifying assumptions, or any derivations. These elements are developed in the body of the paper (Sections 2–4). While abstracts are conventionally concise, we accept that a modest expansion could allow readers to assess the central claim more directly from the abstract itself. We will therefore revise the abstract to include brief statements of the key targeting concept, the main identification conditions, and the role of the positive selection assumption. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

Only the abstract is provided, which introduces the targeting concept as an organizing device for multivalued IV settings and states that it yields point or partial identification of counterfactuals for composite compliers under standard selection assumptions plus an optional positive selection restriction. No equations, fitted parameters, self-citations, or derivation steps are supplied that could reduce any claimed result to its own inputs by construction. The abstract presents the targeting relation and positive selection as additional structure that separates groups, without evidence that these reduce to tautologies or prior self-referential results. This is the normal case of a paper whose central claims remain externally falsifiable once the full text is examined.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract alone supplies no explicit list of free parameters, axioms, or invented entities; the work appears to rest on standard instrumental variable assumptions plus the newly introduced targeting relation and positive selection assumption.

pith-pipeline@v0.9.0 · 5587 in / 1083 out tokens · 46749 ms · 2026-05-24T14:31:27.184197+00:00 · methodology

Treatment Effects with Targeting Instruments

Core claim

What carries the argument

If this is right

Where Pith is reading between the lines

Load-bearing premise

What would settle it

discussion (0)