Skip to content
All library documents

Forecasting FX Spot Rates Beyond the Random Walk Benchmark

Article Quant Q&A · Author: beeba
Read the original

Summary

This discussion asks how to forecast a currency spot rate over a three-month horizon and whether the matching forward rate is the best estimate. It presents the random walk without drift as a difficult benchmark to beat and notes that forward rates can be biased predictors. The cited literature review discusses potential explanatory approaches, including Taylor rules and net foreign assets; commodity prices may also matter for the Canadian dollar because it is a commodity-linked currency.

A second suggestion is to compare forward points with realized exchange-rate changes to assess the predictive value of interest-rate differentials. An observation from a one-year horizon suggests the differential matters, while the best forecast may fall between zero drift and the forward-implied rate. These are discussion-level observations rather than a validated, consistently superior forecasting model. Results may depend on currency pair, horizon, and sample, and the document gives no detailed empirical procedure or quantitative comparison for the three-month USD/CAD case.

Key ideas

  • The random walk without drift is presented as a challenging benchmark for exchange-rate forecasts.
  • Forward rates may be biased and are not necessarily the most accurate forecasts of future spot rates.
  • Taylor rules and net foreign assets are cited as possible sources of exchange-rate predictability.
  • Commodity prices may be informative for the Canadian dollar.
  • Comparing forward points with realized exchange-rate changes can help evaluate interest-rate differentials.

Tags

ForexStatisticsCarry
Full text
# Predicting the Future FX Spot Rates


# Predicting the Future FX Spot Rates












Say I need to predict what the spot rate between USD and CAD will be in 3 months. What will be the most accurate measure or model that I could possibly use? Does the 3 month forward rate necessarily represent the best guess for the future rate, or is there a model or measure that consistently estimates a superior guess for the future price?

## Answer by Alex C (score 3)

https://quant.stackexchange.com/a/25486

It is very difficult to outperform the "random walk without drift" benchmark. The forward rate is not a particularly good predictor as it is often biased.

Nevertheless some economists claim it is possible. Here is a literature review (Barbara Rossi: Exchange Rate Predictability, Journal of Economic Literature vol. 51, no. 4, December 2013):

https://repositori.upf.edu/bitstream/handle/10230/20816/1369.pdf

From reading this it would seem that Taylor Rules and Net Foreign Assets might hold some promise. Since the CAD is a Commodity Currency, commodity prices may be worth investigating.

But it looks like a tough job.

## Answer by James65 (score 2)

https://quant.stackexchange.com/a/36965

You could compare 3 month fx forward points versus realised 3 month fx differentials to see if interest rate differentials are a good predictor. I looked at a 1 year horizon and concluded that you cannot ignore the interest rate differential, but on the other hand the best prediction might lie between zero drift and the fx forward predicted rate.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.