📈 economics

An Irrelevance Theorem for Risk Aversion and Time-Varying Risk

This paper establishes a theorem demonstrating that under specific conditions regarding preference separation, primitive drivers, and constraint linearity, risk aversion and time-varying risk are irrelevant for the elasticity of endogenous variables with respect to state variables that do not affect higher moments, thereby clarifying the mechanisms required in macroeconomic models to generate a prominent role for risk.

Original authors: Andrew Chen, Francisco Palomino

Published 2026-06-05

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Andrew Chen, Francisco Palomino

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to predict how a car will drive on a road. You have two main things to worry about:

The Engine and the Road: How much gas you put in (productivity) and how steep the hill is (capital). These determine how fast the car usually goes.
The Weather and the Driver's Nerves: Is it raining? Is the road bumpy? Is the driver scared of accidents? These represent risk and risk aversion.

For decades, economists have been trying to figure out if the "Weather and Nerves" (Risk) change how the car responds to the "Engine and Road" (Productivity). Some models say yes, risk changes everything. Others say no.

This paper by Chen and Palomino provides a powerful "Irrelevance Theorem." In simple terms, it argues that in many standard economic models, the driver's fear of risk and the changing weather do not change how the car reacts to the engine or the road.

Here is the breakdown using everyday analogies:

1. The Three Rules of the Game

The authors say this "irrelevance" only happens if the economic model follows three specific rules (like the rules of a board game):

Rule 1: The Driver is Split. The model assumes the driver cares about two separate things: how smooth the ride is over time (intertemporal preference) and how scared they are of bumps (risk aversion). They don't mix these feelings together.
Rule 2: The Road is Split. The model assumes the things that determine the average speed (like the engine) are completely separate from the things that determine the bumpiness (volatility). The engine doesn't get stronger just because the road is bumpy.
Rule 3: The Road is Mostly Straight. The math used to describe the road is roughly a straight line. It doesn't have sudden, wild curves or loops.

2. The Big Discovery: The "Split Screen" Effect

If those three rules are in place, the authors prove a surprising thing: Risk is irrelevant for the car's reaction to the engine.

The Analogy: Imagine you are driving a car on a straight, flat highway. If you press the gas pedal (a change in productivity), the car speeds up.
The Result: It doesn't matter if the driver is a nervous wreck (high risk aversion) or a daredevil (low risk aversion). It doesn't matter if the weather is slightly stormy or perfectly calm. The car will speed up by the exact same amount.
Why? Because the math separates the "gas pedal" logic from the "fear of crashing" logic. Since the road is straight and the two systems don't talk to each other, the fear of risk never gets a chance to change how the car accelerates when you hit the gas.

3. What Does Risk Change?

If risk doesn't change how the car reacts to the gas pedal, what does it change?

The Average Speed (The Intercept): Risk aversion determines the baseline speed. A nervous driver might drive slower on average, even if they react to the gas pedal the same way a brave driver does.
The Bumpy Road: If the shock comes from the weather itself (a change in volatility), then risk matters a lot. But if the shock comes from the engine (productivity), risk doesn't change the reaction.

4. Why Do Some Models Say Risk Matters?

You might ask, "But I've read models where risk causes huge economic swings!" The authors explain that those models are "breaking the rules" of their theorem. To make risk matter, a model must do one of three things:

Mix the Driver's Feelings: Make the driver's fear of risk change how they feel about time (e.g., if I'm scared, I want to spend money now rather than later).
Mix the Road and the Weather: Make the engine work differently depending on how bumpy the road is (e.g., a bumpy road makes the engine less efficient).
Make the Road Curvy: Use math that isn't a straight line. If the road has a sudden, sharp cliff (a non-linear constraint), then the driver's fear of falling off the cliff changes how they drive up the hill.

5. The "Adaptation" Strategy

The paper also suggests that if you want to use a standard model (where risk is irrelevant for productivity shocks) but still want to explain economic booms and busts, you have to change your strategy. Instead of blaming "productivity shocks" (the engine), you have to blame "volatility shocks" (the weather).

The Analogy: If you want to explain why traffic jams happen using a model where the engine doesn't care about the weather, you can't say "the engine broke." You have to say "the weather got really bad, and that scared everyone."

Summary

The paper is a "reality check" for economists. It says:

"If you build your model with a straight road and a driver who keeps their fear and their planning separate, then fear of risk will not change how the economy reacts to productivity shocks."

This explains why many standard models struggle to match real-world data where risk seems to drive everything. To fix the models, economists either need to build "curvy roads" (non-linearities), mix the driver's emotions, or stop trying to explain business cycles with productivity shocks and start explaining them with volatility shocks.

Technical Summary: An Irrelevance Theorem for Risk Aversion and Time-Varying Risk

Problem Statement
Despite advances in specifying risk attitudes (e.g., recursive preferences, long-run risk, disaster risk) and risk dynamics, standard real business cycle (production) economies struggle to generate significant roles for risk in explaining observed macroeconomic and financial dynamics. While such preferences successfully match asset price facts in endowment economies, they often fail to produce sufficient volatility or realistic comovement in production economies unless augmented with specific frictions (e.g., nominal rigidities or market incompleteness). The authors seek to clarify why these "exotic" risk models often fail to alter the impulse responses of standard production models to fundamental shocks, and under what conditions risk aversion and time-varying risk become irrelevant for economic dynamics.

Methodology
The paper employs a representative agent framework with Epstein and Zin (1989) (EZ) recursive preferences, which separate intertemporal substitution from risk aversion. The analysis proceeds through three primary methodological stages:

Two-Period Illustration: A simplified two-period model is used to derive the intuition behind the irrelevance result. The authors decompose the Euler equation into an "Intertemporal Term" (driven by expected values and the elasticity of intertemporal substitution) and a "Risk Term" (driven by risk aversion and higher moments). They demonstrate that if the production technology separates intertemporal transformation from state-contingent transformation, and if constraints are approximately linear, the Risk Term decouples from the Intertemporal Term regarding elasticities with respect to first-moment states.
General Theoretical Framework: The authors formalize a general abstract structure for rational-expectations models. They define "first moment states" (variables controlling expected values, e.g., productivity, capital) and "higher moment states" (variables controlling volatility and skewness). The proof relies on risk-adjusted affine approximations. This involves:
- Assuming constraints are conditionally affine (log-linearized).
- Assuming the cumulant generating functions (CGFs) of shocks are conditionally affine in the higher moment states.
- Using the method of undetermined coefficients to solve for the laws of motion of endogenous variables.
Numerical Validation: To address concerns regarding the approximate linearity assumption, the authors employ global projection methods (Judd, 1992; Caldara et al., 2012). They simulate two classes of models:
- Class 1: Frictionless models with normal shocks (extending Tallarini, 2000).
- Class 2: Models with large investment frictions, low elasticity of intertemporal substitution (EIS), and skewed, fat-tailed gamma shocks.
  These simulations test the irrelevance result under extreme risk aversion (up to 100) and high volatility of volatility (up to 80% of baseline).

Key Contributions and Theoretical Results
The paper establishes a formal Irrelevance Theorem stating that under three specific conditions, risk aversion and time-varying risk are irrelevant for the elasticity of any endogenous variable with respect to first-moment states:

Separation of Preferences: Intertemporal and risk attitudes are separated (as in EZ preferences).
Separation of Drivers: The model's primitives separate the drivers of first moments (means) from higher moments (volatility, skewness).
Approximate Linearity: The constraints are approximately linear.

Main Findings:

Elasticity Irrelevance: Under the theorem's conditions, the elasticity of endogenous variables (including asset prices) with respect to first-moment states (e.g., productivity shocks, government policy) is independent of risk aversion ( $\gamma$ ) and parameters governing time-varying risk. The impulse responses to these shocks are identical to those in a certainty-equivalent (no risk) model.
Distinction from Tallarini (2000): The paper clarifies that the Tallarini separation is not between macroeconomic quantities and financial prices, but between means and volatilities. Risk aversion affects the intercepts and steady states (e.g., the average equity premium) but not the dynamics (volatilities and impulse responses) of variables driven by first-moment shocks.
Heteroskedasticity: If shocks are heteroskedastic, risk aversion remains irrelevant for responses to first-moment states but becomes critical for responses to higher-moment states (e.g., volatility shocks).
Robustness: The irrelevance result holds even in models with significant non-linearities (low EIS, high adjustment costs) and non-Gaussian shocks (fat tails, skewness), provided the constraints do not create strong non-linear interactions between first and higher moments.
Extensions: The theorem extends to broader preference classes, including smooth ambiguity (Klibanoff et al., 2005) and multiplier preferences (Hansen and Sargent, 2001), as long as they maintain the separation of intertemporal substitution from other preference dimensions.

Significance and Implications
The paper provides a benchmark for understanding why standard production economies often fail to incorporate risk as a primary driver of business cycle dynamics. It suggests that for risk modeling to be meaningful in such economies, modelers must either:

"Break" Irrelevance: Remove one of the three key assumptions. This involves modeling non-separable risk-return tradeoffs in production (linking first and higher moments), introducing strong non-linearities (e.g., occasionally binding constraints like the zero lower bound), or using preferences where intertemporal and risk attitudes are not separated (e.g., habit formation).
"Adapt" to Irrelevance: Drive business cycles using shocks to higher-moment states (e.g., volatility shocks) rather than first-moment states. However, this requires overcoming the Barro and King (1984) comovement problem, often necessitating additional frictions like nominal rigidities or investment reallocation channels.

The authors conclude that the desire to separate economic channels (intertemporal vs. risk) in standard modeling leads to a stark separation in equilibrium conditions, rendering risk irrelevant for specific dynamics. This theorem clarifies the limitations of current models and outlines the specific structural modifications required to generate a prominent role for risk in production economies.