Partial Adjustment Model: Derivation & Estimation

When you’re juggling graduate classes, it’s easy for homework and exam prep to become a bottleneck. Online tutoring can help you focus on the exact skills your course is testing and avoid avoidable mistakes.

This article is written at a graduate level and emphasizes definitions, assumptions, and a transparent reasoning process. Rather than listing results, it explains why the method works and how to apply it step by step. The goal is a clear, defensible understanding suitable for assignments, exams, and projects.

The partial adjustment model describes how an actual variable adjusts gradually toward a desired level. It’s widely used in inventory, investment, and consumption dynamics. Let

yₜ* = γ₀ + γ₁ xₜ + eₜ
yₜ − yₜ₋₁ = λ(yₜ* − yₜ₋₁) + aₜ

Here, yₜ* is the desired level, yₜ is the actual level, xₜ is a contemporaneous regressor, λ ∈ (0, 1) is the speed of adjustment, and eₜ, aₜ are error terms.

(i) Deriving the Dynamic Equation

Substitute yₜ* into the second equation:

yₜ − yₜ₋₁ = λ(γ₀ + γ₁ xₜ + eₜ − yₜ₋₁) + aₜ

Expand and rearrange:

yₜ = (1 − λ) yₜ₋₁ + λ γ₀ + λ γ₁ xₜ + λ eₜ + aₜ

Define:

β₀ = λ γ₀
β₁ = 1 − λ
β₂ = λ γ₁
uₜ = λ eₜ + aₜ

Then the model becomes:

yₜ = β₀ + β₁ yₜ₋₁ + β₂ xₜ + uₜ

This is a dynamic linear model with a lagged dependent variable and contemporaneous regressor.

(ii) Estimation Strategy

If E(eₜ | xₜ, yₜ₋₁, xₜ₋₁, …) = E(aₜ | xₜ, yₜ₋₁, xₜ₋₁, …) = 0 and all series are weakly dependent, then uₜ is uncorrelated with regressors.

We can estimate β₀, β₁, β₂ using OLS on:

yₜ = β₀ + β₁ yₜ₋₁ + β₂ xₜ + uₜ

OLS will be consistent and asymptotically normal under standard regularity conditions.

(iii) Recovering γ₁ and λ from Estimates

Suppose we estimate:

β̂₁ = 0.7
β̂₂ = 0.2

Then:

λ̂ = 1 − β̂₁ = 1 − 0.7 = 0.3
γ̂₁ = β̂₂ / λ̂ = 0.2 / 0.3 ≈ 0.667

So the estimated speed of adjustment is 30%, and the effect of xₜ on desired yₜ* is about 0.667.

Summary

  • The partial adjustment model leads to a dynamic regression with lagged y and contemporaneous x
  • OLS is valid under exogeneity and weak dependence
  • Structural parameters γ₁ and λ can be recovered from β̂₁ and β̂₂

This framework is widely used in empirical economics to model gradual behavioral responses over time.

Graduate-level courses frequently include this topic in problem sets, exams, or project requirements. Careful attention to definitions and steps can improve results. Online tutoring support is available for graduate quantitative studies.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *