Skip to contentScholarGate
LibraryBookshelfDeskReview StudioAssistant
Sign in
On this page
IntuitionHow it worksWhen to use itStrengths & limitationsCommon pitfallsApplicationsFrequently asked🔒 Read the full methodSourcesRelated methods
Cite this pageSpotted an issue on this page? Report or suggest a fix →
Home›Econometrics›Pesaran-Timmermann Test of Directional Predictive Accuracy
Hypothesis testForecast evaluation

Pesaran-Timmermann Test of Directional Predictive Accuracy

Also known as: PT Test, Directional Accuracy Test, Nonparametric Predictive Performance Test, Pesaran-Timmermann Yön Testi

Introduced by Pesaran and Timmermann (1992), the PT test is a nonparametric procedure that evaluates whether a forecasting model correctly predicts the direction (sign) of a target variable more often than would be expected by chance. It is widely used in financial econometrics and macroeconomic forecasting to assess the practical utility of a model beyond simple error metrics, particularly when the economic cost of getting the direction wrong is high.

ScholarGate
  1. Hypothesis test
  2. v1
  3. 1 Sources
  4. PUBLISHED
Cite this page →
Tools & resources
Download slides
Learn & explore

Read the full method

Members only

Sign in with a free account to read this section.

Sign in

Method map

The neighbourhood of related methods — select a node to explore.

Pesaran-Timmermann Test
Diebold-Mariano TestRuns TestSign Test

When to use it

Use the PT test when the primary concern is directional accuracy rather than magnitude of forecast errors, for example in predicting stock return signs, exchange rate movements, recession indicators, or business cycle turning points. The test assumes that observations are independently and identically distributed under the null; applying it to serially correlated data or rolling-window forecasts may require bootstrap modifications. It is not suitable when zero-valued outcomes are frequent, as sign classification is ambiguous at zero. For magnitude-based comparisons, prefer the Diebold-Mariano test; for turning-point analysis, consider dedicated turning-point tests.

Strengths & limitations

Strengths
  • Nonparametric and distribution-free under the null, requiring no assumptions about the error distribution of the forecasting model.
  • Directly measures economically relevant directional skill, which can matter more than MSE in trading or policy decision contexts.
  • Simple closed-form statistic with a standard normal limiting distribution, making inference computationally trivial.
  • Robust to model misspecification because it uses only the signs of forecasts and realizations, not their magnitudes.
Limitations
  • Ignores the magnitude of forecast errors, so a model that is directionally accurate but badly wrong in size will pass the test despite poor predictive quality.
  • The asymptotic normal approximation can be unreliable in small samples; simulation or bootstrap critical values are advisable when T is less than roughly 30–50.
  • Does not account for serial correlation in the hit sequence; autocorrelated forecast errors may inflate the test statistic under the null.
  • Cannot handle ties (zero realizations or zero forecasts) without an ad hoc convention, potentially biasing results in low-frequency macroeconomic data.

Frequently asked

How does the PT test differ from the Diebold-Mariano test?

The Diebold-Mariano test compares the mean forecast loss (e.g., MSE or MAE) of two competing models, focusing on error magnitude. The PT test instead evaluates whether the sign of the forecast matches the sign of the realization more often than chance, making it complementary to loss-based tests and particularly relevant when directional accuracy carries direct economic value.

Can the PT test be used with overlapping or rolling-window forecasts?

The original asymptotic theory assumes i.i.d. observations under the null. Overlapping or rolling-window forecasts introduce serial correlation in the hit sequence, invalidating standard critical values. In those settings, block bootstrap or heteroskedasticity-and-autocorrelation-consistent variance estimators should replace the analytic variance formula.

What should be done when the realization or forecast is exactly zero?

A zero outcome has no unambiguous sign. Common conventions include excluding zero observations from the count, treating zero as a missed prediction, or using a three-category sign function that maps zero to its own class. The choice should be reported transparently, as it can materially affect the hit rate and test conclusion in low-frequency data.

Sources

  1. Pesaran, M. H., & Timmermann, A. (1992). A simple nonparametric test of predictive performance. Journal of Business & Economic Statistics, 10(4), 461–465. DOI: 10.1080/07350015.1992.10509922 ↗

How to cite this page

ScholarGate. (2026, June 2). Pesaran-Timmermann Test of Directional Predictive Accuracy. ScholarGate. https://scholargate.app/en/econometrics/pesaran-timmermann-test

Related methods

Diebold-Mariano TestRuns TestSign Test

Which method?

Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.

  • Diebold-Mariano TestEconometrics↔ compare
  • Runs TestStatistics↔ compare
  • Sign TestStatistics↔ compare
Compare side by side →

Referenced by

Diebold-Mariano Test

Similar methods

Diebold-Mariano TestGiacomini-White TestHatemi-J Asymmetric CausalityModel Confidence SetGranger CausalityGranger Causality TestNonlinear Granger CausalityPanel Toda-Yamamoto Causality

Related reference concepts

Permutation TestsStatistical Hypothesis TestingEconometric ModelingNonparametric StatisticsEconometricsRank-Based Methods

Spotted an issue on this page? Report or suggest a fix →

ScholarGate — Pesaran-Timmermann Test (Pesaran-Timmermann Test of Directional Predictive Accuracy). Retrieved 2026-07-21 from https://scholargate.app/en/econometrics/pesaran-timmermann-test · Dataset: https://doi.org/10.5281/zenodo.20539026
Quick facts
Originator
M. Hashem Pesaran & Allan Timmermann
Year
1992
Type
Nonparametric one-sided test
Subfamily
Forecast evaluation
Null Hypothesis
Forecasts and realizations are independently distributed (no directional predictability)
Asymptotic Distribution
Standard normal under the null
Related methods
Diebold-Mariano TestRuns TestSign Test
ScholarGate

A content-first reference library for research methods — what each one is, how it works, and where it comes from.

Open data (CC-BY)

Explore

  • Library
  • Search the library…
  • Browse by field
  • Fields
  • Journey
  • Compare
  • Which method?

Reference

  • Subjects
  • Atlas
  • Glossary
  • Methodology
  • Philosophy

Your tools

  • Bookshelf
  • Desk
  • Chat

Company

  • About
  • Pricing
  • Contact
  • Suggest a method

Entries are compiled from published sources for reference. Verifying the accuracy and suitability of any information for your own use remains your responsibility.

© 2026 ScholarGate · A research-method reference library
  • Privacy
  • Cookies
  • Terms
  • Delete account