Control VAR: a counterfactual based approach to inference in macroeconomics

summary

The gist

This paper addresses the challenges of giving a causal interpretation to vector autoregressions (VARs) and proposes an alternative method called control-VAR to estimate causal effects when…

In short

Traditional recursive VARs are often misinterpreted as causal when policy is ordered first, but this ordering fails if the policy reacts to the economy it influences. The Control VAR (CVAR) method addresses this by comparing a treated unit with untreated controls sharing its trends. This removes biases from common trends and specific unit reactions, allowing for robust causal inference even when independence assumptions are violated.

Key concepts

Recursive VAR (VAR)
A statistical model that describes how a set of time series variables are related to each other over time. When used in policy analysis, the order in which variables are placed matters; if policy is ordered first, the VAR can be interpreted as causal, but this is not always true.
Control VAR (CVAR)
An alternative method that improves causal inference by comparing a treated unit with untreated controls that share similar stochastic trends. This comparison helps remove biases arising from common economic trends and specific reactions of the treated unit to those trends.
Impulse Response Function
A tool used to measure the effect of a shock (like a policy change) on future economic variables. The paper shows that when a policy reacts to the economy, this function combines the direct causal effect with biases from common trends and biases specific only to the treated unit.

Terminology used across episodes

This episode discusses

The paper

Control VAR: a counterfactual based approach to inference in macroeconomics · Read on arXiv

Raimondo Pala

University of Rome Tor Vergata

A recursive VAR is read as causal whenever the policy is ordered first. I show that this ordering is not sufficient when the policy reacts to the economy it is meant to affect: the impulse response then combines the causal effect with a bias from the common trends that the policy reacts to and a bias specific to the treated unit. I develop the Control VAR (CVAR), which compares the treated unit with untreated controls that share its stochastic trends, and removes the first bias and, under exogeneity with respect to the counterfactual cycle, the second. The method allows repeated, continuous and multiple policies, separates the effect of a single policy change from that of its later values, measures the bias of a VAR without controls through the response of the controls, and bounds the remaining bias. I estimate the weights, derive bootstrap inference, show that the CVAR in changes remains valid when the treated unit has a stochastic trend of its own, test whether the more efficient CVAR in levels is supported, and characterise the bias of weights estimated in changes. I apply the CVAR to the Alaska Permanent Fund Dividend, extended unemployment benefits and California's Low Carbon Fuel Standard.

Transcript

Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.

Ruth: We're looking at the paper: "Control VAR: a counterfactual based approach to inference in macroeconomics".

Warren: Here's the gist of the paper. This paper addresses the challenges of giving a causal interpretation to vector autoregressions (VARs) and proposes an alternative method called control-VAR to estimate causal effects when independence assumptions cannot be relied upon.

Ruth: Causal Interpretation of VARs The paper clarifies the causal interpretation of VAR with macroeconomic data and highlights the different causal effects that can be identified using VAR with different policy distributions under the same type of independence assumptions.

Paper summary: Ruth: We've touched on the idea that this paper, "Control VAR: a counterfactual based approach to inference in macroeconomics," is about finding ways to make causal claims in VARs when independence assumptions fail. They introduce control-VAR as an alternative method that uses control variables specifically for estimating those effects.

Warren: They claim that control-VAR can estimate average treatment effects on the treated for dummy policies, or even average causal responses over time for continuous policies >

Iris: The paper demonstrates the advantage of these approaches by looking at the impact of natural disasters on the US economy and using Germany as a control variable. Contrary to some prior literature, their results show that natural disasters have a negative economic impact without any cyclical positive effect >

Ruth: So, the core thesis is that control-VARs provide a more credible way to get causal estimates when you can't assume independence conditions are met >

Warren: And they show how this method helps us address situations where we need to define counterfactual worlds, like what would have happened if central banks around the world had chosen inaction instead of unconventional monetary policy on the offset of the Great Financial Crisis >

Iris: It seems to be about providing economists with more practical tools for defining those counterfactuals that help establish a clear causal link between a policy and an outcome >

Ruth: And they are looking at how this method works in practice, comparing it to methods that rely on impulse response functions, which is what we often use >

Warren: So, the paper sets up this control-based approach as a way to handle situations where those standard identification tools just don't work >

Conclusion: Iris: Looking at "Control VAR: a counterfactual based approach to inference in macroeconomics," the authors are essentially arguing for using control variables to make causal claims in VARs when independence assumptions are shaky. They're providing a new framework for identification >

Ruth: The implication is that this method gives economists more reliable estimates, especially when dealing with complex policy situations where we can't just assume simple independence holds across all variables >

Warren: It shifts the focus from just finding a single impulse response function to building an entire control-based structure to get a better view of the causal effect over time >

Iris: So, what this means for us is that we have an additional tool in our econometric toolkit for when we think about how policies affect things and we need to be careful about the assumptions underlying our measurements >

More episodes

← Home