# What are the most popular technical models used for forecasting exchange rates in international economics?

Are there any de facto standards when publishing, or is there substantial latitude in deriving one's own methodology?

I understand in practice, even the St. Louis Fed is gently skeptical of the validity of most technical methods, but I was wondering how academia viewed them.

While Dooley and Shafer (1976, 1984) were largely dismissed, Neely and Weller (1999) seemed to fare fairly well using genetic algorithms, and Markov and ARIMA methods are reviewed more or less favorably in terms of predictive value.

Is accuracy of prediction (and consequently profitability) even necessarily a factor in modelling for academia, or is the simplicity and explanatory value more heavily favored?$^1$

$^1$*Note: Not sure if this merits a separate question, or can be answered as an elaboration of the primary*

-
Maybe too general, but try this: scholar.harvard.edu/files/rogoff/files/51_jie1983.pdf –  Anton Tarasenko Dec 20 '14 at 11:05
Certainly helpful, @Anton. Thank you. Do you know if there are any updates to this paper? Though I suppose I could run it through a citation engine to see who built off it. –  Jason Nichols Dec 21 '14 at 5:45
Try papers referring this one. And this paper: ssc.wisc.edu/~mchinn/CCG-P_JIMF.pdf –  Anton Tarasenko Dec 21 '14 at 9:40

Writing as an academic modeller who would use the outputs of such models, what's required (and what I look for as a reviewer) is:

• reproducibility: the algorithm is described with sufficient detail to enable someone else to reimplement it;
• validation, using data that is completely distinct from the calibration dataset; and
• that the output is a distribution of forecasts, not just point values.

Such a distribution can then be fed into ensemble modelling. So instead of using a single forecast of the future value of a currency pair, I could feed in the distribution to a monte carlo simulator or similar.

It doesn't need to have the right expected value, nor a correct value for volatility (or any other moment).

All that makes it essentially useless for trading. And that's a good thing, because no one of sound mind would give away a profitable forex forecaster: such a forecaster would immediately render itself obsolete as it entered common use.

-
I agree with the fact that forecasters only work if no one know about them, but I'm surprised to hear that there isn't a standard forecasting model (like Black-Scholes for derivatives). That's helpful though. Thank You! –  Jason Nichols Dec 21 '14 at 5:43