Nonparametric testing for long-horizon predictability with persistent covariates

Research output: Contribution to journalArticlepeer-review


We propose a testing procedure for long-horizon predictability via kernel-based nonparametric estimators of long-run covariances between multiperiod returns and persistent covariates. Asymptotic properties of the proposed tests are studied. As for implementation of the test, sieve bootstrap methods are employed to obtain reasonable approximation to the sample distribution of the test statistics. Monte Carlo simulations are conducted to verify the theoretical conjecture. Empirical analysis, using US monthly data from 1929 to 2011, are presented for testing stock return predictability of some forecasting financial variables. Long-term interest rates, unlike default spreads or price-earning ration, are found to show some forecasting power.

Original languageEnglish
Pages (from-to)359-372
Number of pages14
JournalJournal of Nonparametric Statistics
Issue number2
StatePublished - 2014

Bibliographical note

Publisher Copyright:
© American Statistical Association and Taylor & Francis 2013.


  • Long-horizon predictability
  • Nonparametric estimator
  • Sieve bootstrap


Dive into the research topics of 'Nonparametric testing for long-horizon predictability with persistent covariates'. Together they form a unique fingerprint.

Cite this