Ok, firstly, why is the exchange rate fluctuating?
.
Well, the exchange rates respond very quickly to new information in
the markets. So the majority of short term variation is due to either
new information reaching the markets (which can't be forecasted), or
speculative activity (which also can't be forecasted).
.
Now, I COULD sit in excel and develop some kind of sophisticated
autoregressive, intergrated distributed lag econometric model, but I'd
rather tell you how to do it :) I'm not opening excel :)
.
Firstly, because the exchange rate is a ratio, I'd use a log-log
model. Take logs of all the data you implement in the model.
.
Next, consider using a difference model. Although you might get more
accurate short-term forecasts with difference models, we know that
short run forecasting for ex rates sucks anyway, so just leave it. If
it is in a non-differenced format, the longer term rates tend to be
much better.
.
Now consider implementing a few economic variables that might be
important. Suggestions include inflation rates, interest rates, the
money supply, imports, exports, GDP. Also decide which of these need
to be differenced and which don't. (eg, log the interest rates unless
they go negative, maybe difference the money supply for it's rate of
growth, etc).
.
Choose a functional form, that is, how you think the model fits
together, how each variable should be combined with the other
variables to determine the logUSD/EUR rate (or whatever)
.
When you estimate the model, leave off the last few observations, and
let your model produce forecasts against them. Then see how your
'forecast' fits against the actual data. This is a pretty good
indicator of whether your model is worth anything.
.
Finally, remember that even the experts can't forecast the exchange
rate with any degree of certainty. Most certainty occurs over time
frames of a year (or longer!)
.
I hope my answer was somewhat satisfactory :) |
