Forecasting with VEC models (version 2 of question)

[stoddj]

I deleted an earlier post on this topic, which speculated that the Error Correction component of a VEC models might sometimes overwhelm the VAR portion, thus lending a sort of inertia that makes the model good for covering relatively stable periods, but not for forecasting substantial changes.

I still think that might sometimes be the case, but here's another, perhaps more basic question. Consider a VEC demand model based on lagged first-difference terms of price and quantity, along with a few exogenous variables like unemployment. The negative coefficients on all the lagged price difference terms are large enough (in absolute value) to imply substantial price elasticity; i.e., that the % Change in Quantity divided by the % Change in Price is substantially greater than 1 in absolute value. It is then obvious that small decreases in price should increase overall revenue. My forecast of a small price change does show quantity increasing, but not by more than price has decreased -- i.e., not enough to increase revenue.

What's going on? How can the output of the forecast model be at such variance with the estimated coefficients? In calculating the elasticities, I'm assuming that the linear coefficients on several lagged difference terms can be averaged together along with averages on the Price change and Quantity change terms. Is it perhaps the case that estimating the elasticities is not so straightforward?

By the way, I am using Eviews 6. My time series has about 40 quarterly observations, and I'm trying to forecast about 8 quarters into the future