# Begin forecast from arbitray value of endogenous variable, and not the steady state

I have been able to construct forecasts when I impose shocks, but the forecasts appear to start from the steady state values. Is it possible to construct forecasts from an arbitrary starting value of the endogenous variable, and not from the steady state? For instance, I would like to forecast output from 3.484 defined below, and not from its steady state value of 0.

I tried:

histval;

output(0) = 3.48447; // 1 indicates the first period of simulation

output(-1) = 4.665;

end;

shocks;

var residual_output;
periods 1;
values 10.88;
end;

resid;
stoch_simul(order = 1, irf = 20, graph)output;

forecast(periods = 40);

But it seems that in a stochastic model, histval performs the same role as initval? i.e. takes the initial value of 3.484 as an approximation, then finds the exact steady state using a numerical approx method. To avoid that, I removed the line that says “steady”, but the problem still persists.

How can I get forecast of output to start from 3.484, and not from steady state of 0? There are also several other endogenous variables in the system. Do I need to provide values of other endog variables in histval for this to work?

The problem is that output is an endogenous jumper. You cannot select an initial condition for it. Rather, output will be determined by the predetermined state variables. You could e.g. select a value for capital and TFP and that would determine output.

Ok, I see. Let’s say that the IRFs imply that the output gap fall by 10 percent (from the steady state of 0) to -10 on the y-axis. Would it be correct to say that if the output gap at steady state were 2, then then the output gap falls to -8 percent? Here I have only shifted the numbers in the y-axis, keeping the magnitude same, but I’m not sure if this statement is mathematically correct.

That depends on your measurement. If the level of output in steady state is 2 and it drops by 10 percent, then you will end up at 2(1-0.1)=1.8