Stock market efficiency in Iran: unit root testing with smooth structural breaks and non-trading days

Seyyed Ali Paytakhti Oskooe, Daly Vince

Research output: Working paperDiscussion paper

Abstract

A 'flexible Fourier trend' unit root test, permitting smooth structural breaks of unknown form and dates, is used to test weak-form market efficiency in the Tehran stock market's TEPIX index. Monte Carlo experiments show that this test has low power when non-trading-day gaps in the daily data are filled with missing value codes. The test's properties for weekly returns and for data as published, with non-trading-day gaps suppressed, are better and similar to each other. Analysis of the full sample of TEPIX data as published supports a unit root null but indicates the presence of additional autocorrelation - questioning weak-form efficiency. Sub-sample analysis again finds evidence of a unit root, but also of complex autocorrelation. Support for the unit root increased in the years (2000-2004) following regulatory reform and has decreased since 2008. A Diebold and Mariano (1995) test is used to assess whether the revealed autocorrelation provides an effective basis for predicting price deviations from trend on the basis of their own history. Predictive effectiveness is found at a horizon of one trading day. We conclude that this market has not shown weak-form efficiency.
Original languageEnglish
Place of PublicationKingston upon Thames, U.K.
PublisherFaculty of Arts and Social Sciences, Kingston University
Number of pages26
Publication statusPublished - Sept 2015
Externally publishedYes

Publication series

NameEconomics Discussion Paper
PublisherFaculty of Arts and Social Sciences, Kingston University
No.2015-06

Keywords

  • Economics and econometrics
  • Market efficiency
  • Non-trading days
  • Structural breaks
  • Unit root tests

Fingerprint

Dive into the research topics of 'Stock market efficiency in Iran: unit root testing with smooth structural breaks and non-trading days'. Together they form a unique fingerprint.

Cite this