# Classical IV foundations and refinement

> One thread of 6 from the "instrumental variable" map, covering the 12 papers in it. The other threads are not represented here.
>
> Every quotation was copied word for word from the paper's own text, and
> checked against that text. Quotes marked *inferred* failed that check and
> must be re-checked before use. Quotes marked *not re-checked* have not been
> matched against the paper's text as it now stands, so they carry no current
> verification either. Lines labelled *the tool's reading* are
> model judgment, not quotation, and carry no verification.
>
> **This is a scaffold, not prose.** The citations, quotes and structure are
> real; the argument is yours to write.

> Connects to: High-dimensional and sparse IV methods, Nonparametric and semiparametric IV models, IV validity, violations, and sensitivity analysis, Specialized IV applications and domain extensions.

Beginning with treatments of classical instrumental variables under nonstationary data and asymptotic theory, this line evolved toward understanding the behavior of 2SLS under weak instruments, first-stage design, and formal statistical tests. Work progressed from proving consistency under departures from stationarity through developing tests for instrument strength and robustness to weak identification.


**Nonparametric methods for inference in the presence of instrumental variables** (2005) \cite{hall2005nonparametric}

What it did: Establish optimal rates for nonparametric IV estimation  *(the tool's reading)*

The 2005 paper defines instrumental variables through the orthogonality condition E(U|W)=0 and derives optimal convergence rates for the first time in nonparametric settings, establishing foundational benchmarks for this estimand.  *(the tool's reading)*

> “Suppose, however, that for each i we have available another observed data value, Wi, say (an instrumental variable), for which E(Ui|Wi) = 0 (1.2) and there is a “sufficiently strong” relationship between Xi and Wi.”
>
> ✓ verified: found word for word in the paper's own text


**Cowles commission structural equation approach in light of nonstationary time series analysis** (2006) \cite{hsiao2006cowles}

What it did: Extend classical 2SLS consistency to nonstationary settings  *(the tool's reading)*

Building on classical IV definitions, the 2006 paper shows 2SLS consistency persists under nonstationarity but requires modifications like Phillips-type adjustments or lag augmentation to restore validity of inference.  *(the tool's reading)*

> “The same rank condition for identification holds for stationary and nonstationary data and some sort of instrumental variable estimators will have to be employed to yield consistency.”
>
> ✓ verified: found word for word in the paper's own text


**A maximum likelihood method for the incidental parameter problem** (2009) \cite{moreira2009maximum}

What it did: Unify strong and weak instrument asymptotics via likelihood  *(the tool's reading)*

The 2009 paper treats IV as simultaneous equations and develops invariant likelihood theory for LIML that unifies asymptotic behavior across both strong and many weak instruments regimes.  *(the tool's reading)*

> “Consider a simple simultaneous equations model with two endogenous variables, multiple instrumental variables (IVs) and errors that are normal with known covariance matrix.”
>
> ✓ verified: found word for word in the paper's own text


**Instrumental Variables: An Econometrician's Perspective** (2014) \cite{imbens2014instrumental}

What it did: Bridge simultaneous equations tradition to modern causal inference  *(the tool's reading)*

The 2014 perspective paper connects the classical 1920s supply-demand simultaneous equations literature with contemporary statistical applications, situating IV as a unified method for causal inference.  *(the tool's reading)*

> “Instrumental Variables (IV) refers to a set of methods developed in econonometrics starting in the 1920s to draw causal inferences in settings where the treatment of interest cannot be credibly viewed as randomly assigned, even after conditioning on additional covariates, that is, settings where the assumption of no unmeasured confounders does not hold.”
>
> *(inferred: the check ran against the paper's text and could not find this passage, so re-check it before citing)*


**Inference with Many Weak Instruments** (2020) \cite{mikusheva2020inference}

What it did: Formalize weak identification via concentration parameter conditions  *(the tool's reading)*

The 2020 paper defines weak identification through necessary and sufficient conditions based on concentration parameters relative to instrument count, providing a rigorous characterization of the many-instruments regime.  *(the tool's reading)*

> “We develop a concept of weak identification in linear IV models in which the number of instruments can grow at the same rate or slower than the sample size.”
>
> ✓ verified: found word for word in the paper's own text


**Wild Bootstrap for Instrumental Variables Regressions with Weak and Few\n Clusters** (2021) \cite{wang2021wild}

> “Furthermore, we let Z_i,j∈R^d_z denote the IVs for X_i,j.”
>
> ✓ verified: found word for word in the paper's own text


**Weak Instrumental Variables: Limitations of Traditional 2SLS and Exploring Alternative Instrumental Variable Estimators** (2021) \cite{huang2021weak}

What it did: Analyze 2SLS under weak instruments and develop alternatives  *(the tool's reading)*

The 2021 paper provides detailed theoretical analysis of 2SLS properties under weak instruments and introduces alternative estimators, explicitly addressing limitations of standard methods.  *(the tool's reading)*

> “A valid instrument Z_i must satisfy two conditions: * Instrument Relevance: corr (Z_i, X_i) ≠ 0.”
>
> ✓ verified: found word for word in the paper's own text


**Weak Instruments, First-Stage Heteroskedasticity, the Robust F-Test and a GMM Estimator with the Weight Matrix Based on First-Stage Residuals** (2022) \cite{windmeijer2022weak}

What it did: Account for first-stage heteroskedasticity in weak settings  *(the tool's reading)*

The 2022 paper introduces GMMf, a robust estimator using first-stage residuals in the weight matrix to correctly account for concentration differences across instruments when first-stage heteroskedasticity exists.  *(the tool's reading)*

> “This could either be a non-robust or robust version of the test, with robustness to for example heteroskedasticity, serial correlation and/or clustering. Under maintained assumptions, these are valid tests for the null H_0:π=0 in the first-stage linear specification x=Zπ+v, where x is the endogenous explanatory variable in the model of interest y=xβ+u, and Z are the instruments.”
>
> ✓ verified: found word for word in the paper's own text


**On the Role of the Zero Conditional Mean Assumption for Causal Inference in Linear Models** (2022) \cite{crudu2022role}

What it did: Challenge causality interpretation despite exogeneity condition  *(the tool's reading)*

The 2022 paper shows that 2SLS can fail to identify causal effects even when E[ε|Z]=0 without requiring nonlinearity, revealing limitations of the classical exogeneity assumption for causal validity.  *(the tool's reading)*

> “even if D is endogenous i.e., Cov[D,ε]≠ 0, estimating λ by 2SLS using an instrument Z that is uncorrelated with ε does not guarantee a causal interpretation.”
>
> ✓ verified: found word for word in the paper's own text


**Nickell Bias in Panel Local Projection: Financial Crises Are Worse Than You Think** (2023) \cite{mei2023nickell}

> “For general dynamic panel models, the Nickell bias is usually addressed with the instrumental variable (IV) method […] or analytical formulas […]. The IV method is subject to the lack of efficiency and performs poorly with finite samples when the regressor is persistent […].”
>
> ✓ verified: found word for word in the paper's own text


**The First-stage F Test with Many Weak Instruments** (2023) \cite{huang2023first}

What it did: Establish first-stage F distribution under many-instruments asymptotics  *(the tool's reading)*

The 2023 paper derives the asymptotic normal (not noncentral chi-squared) distribution of the first-stage F statistic under simultaneous-limit many-instruments asymptotics, refining inference for weak instrument detection.  *(the tool's reading)*

> “However, this approach was originally developed for a fixed number of instrumental variables (IVs), and does not address the case of a large number of instruments, which is commonly encountered in practice […].”
>
> ✓ verified: found word for word in the paper's own text


**Philip G. Wright, directed acyclic graphs, and instrumental variables** (2025) \cite{abbring2025philip}

What it did: Credit Wright with formalizing IV via DAGs and moments  *(the tool's reading)*

The 2025 paper traces IV estimation to Wright's Appendix B, formalizing it through structural equations, directed acyclic graphs, and method of moments as the original classical foundation.  *(the tool's reading)*

> “In particular, it is generally credited with introducing instrumental variables (IV) estimation.”
>
> ✓ verified: found word for word in the paper's own text


_[Your synthesis: what it enabled, what it left unsolved.]_


## Where this thread connects

Each crossing is where one line of work fed another. These are the tool's reading of
the corpus, not quotations.

- **Classical IV foundations and refinement** → **High-dimensional and sparse IV methods** (2010): Many instruments, LASSO first-stage
- **Classical IV foundations and refinement** → **Nonparametric and semiparametric IV models** (2011): Local identification, nonlinear IV
- **Classical IV foundations and refinement** → **IV validity, violations, and sensitivity analysis** (2018): Testing instrument validity
- **Classical IV foundations and refinement** → **Specialized IV applications and domain extensions** (2016): IVQR, spatial models