Cointegration, error-connection models and forecasting using realigned foreign exchange rates

Nathan Lael Joseph

Research output: Contribution to journalArticlepeer-review


This study employs error-correction models (ECMs) to forecast foreign exchange (FX) rates where the data-sampling procedures are consistent with the rules governing the settlement (delivery) of FX contracts in the FX market. The procedure involves thatching (aligning) the forward rate to the 'actual' realized (future) spot rate at the settlement (delivery) date. This approach facilitates the generation of five different sets of sub samples of FX rate series for each currency. For comparative purposes, non-aligned month-end rates are also examined. The results indicate that the moments of the realized forecast errors for the same currency are not similar. Further, the ECMs derived are unstable, and their forecasting performance vary. The forecasting performance of the ECMs appear to be affected by the choice of the interval in which the sets of sub samples are observed. These results are attributed to the observed seasonal variation in FX rates.
Original languageEnglish
Pages (from-to)499-522
Number of pages24
JournalJournal of Forecasting
Issue number6
Publication statusPublished - Nov 1995


  • unbiased ness hypothesis
  • unit roots
  • co integration
  • error-correction models
  • forecasting


Dive into the research topics of 'Cointegration, error-connection models and forecasting using realigned foreign exchange rates'. Together they form a unique fingerprint.

Cite this