Time-Varying Parameter Hausman Test
Also known as: TVP Hausman test, time-varying Hausman specification test, Hausman test with time-varying parameters, TVP endogeneity test
The time-varying parameter Hausman test extends Hausman's (1978) classic specification test to models whose coefficients are allowed to evolve over time. It compares an efficient estimator (e.g., OLS or GLS assuming constant parameters) with a consistent estimator from a time-varying parameter model, using the contrast between them to detect parameter instability or endogeneity in dynamic settings.
Key highlights
- Provides a formal chi-squared test for parameter instability without requiring the researcher to pre-specify break dates.
- Nests the classic Hausman test as a special case, making it familiar and well-grounded in established specification-testing theory.
- Applicable to a broad range of TVP formulations (random walk, AR(1) drift, state-space) depending on the alternative of interest.
- Yields a single scalar statistic summarising joint instability across all slope coefficients simultaneously.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use the TVP Hausman test when you suspect that regression coefficients shift gradually over time — common in macroeconomic, financial, or policy time series spanning multiple regimes. It is particularly appropriate when you have a long time series (T > 50 or more) and want a formal test before committing to a computationally intensive TVP model. It is less suitable for short panels, cross-sectional data with no temporal dimension, or settings where discrete structural breaks (rather than smooth drift) are the plausible alternative — in that case a Chow test or Bai-Perron breakpoint test is preferred.
Strengths & limitations
- Provides a formal chi-squared test for parameter instability without requiring the researcher to pre-specify break dates.
- Nests the classic Hausman test as a special case, making it familiar and well-grounded in established specification-testing theory.
- Applicable to a broad range of TVP formulations (random walk, AR(1) drift, state-space) depending on the alternative of interest.
- Yields a single scalar statistic summarising joint instability across all slope coefficients simultaneously.
- Requires consistent estimation of the TVP model (e.g., Kalman filter), which adds computational complexity and sensitivity to priors or variance assumptions.
- Power can be low in short samples; the chi-squared approximation may be poor when T is small or parameters vary infrequently.
- Does not identify which coefficient is unstable or when the instability began — a rejection triggers further investigation.
- The test conflates parameter instability with other misspecifications (omitted variables, endogeneity), so a significant result is not exclusively diagnostic of drifting parameters.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
How does this differ from the standard Hausman test?
The standard Hausman test contrasts a fully efficient estimator with a robust-but-consistent alternative to detect endogeneity or random-effects misspecification in a static or fixed-T setting. The TVP variant replaces the consistent estimator with a time-varying parameter model, shifting the focus from cross-sectional endogeneity to temporal parameter instability.
What estimator is used for the TVP side of the test?
Typically a Kalman-filter-based state-space estimator where coefficients follow a random walk. The resulting filtered or smoothed coefficient estimates form the 'consistent' side of the Hausman contrast.
What should I do after rejecting the null?
Rejection implies constant parameters are an inadequate assumption. Examine time plots of rolling or Kalman-smoothed coefficients to understand the nature of the drift, and consider estimating a full TVP model, or run Bai-Perron tests to locate discrete break dates.
Can this test be applied to panel data?
Yes, panel variants exist where TVP models allow slope heterogeneity across both units and time. However, the asymptotic theory requires both N and T to be sufficiently large, and implementation is considerably more involved than in the pure time-series case.
How many observations do I need for reliable results?
As a rough guideline, T of at least 50 is advisable for the chi-squared approximation to be reliable. With shorter series, bootstrap or simulation-based critical values are recommended to control size.
Sources
- 1.Hausman, J. A. (1978). Specification tests in econometrics. Econometrica, 46(6), 1251-1271.
- 2.Cooley, T. F., & Prescott, E. C. (1976). Estimation in the presence of stochastic parameter variation. Econometrica, 44(1), 167-184.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 3). Time-varying parameter Hausman test. ScholarGate. https://scholargate.app/econometrics/time-varying-parameter-hausman-test