Negative values of simulated variable

[lovefukanglin]

Dear professor jpfeifer,

Sorry for bothering you again. I estimate a DSGE model using Bayesian method and the mod all run well. But I find the simulated values of output(y) using posterior mean have some negative values!! We know the level of output obvious can not be negative. so there must be something serious wrong but I can not find. I updata my mod file and hope you can help me, always thank you, professor.

eleven11.mod (19.0 KB)

[jpfeifer]

See

Negative numbers

That is because a first-order approximation approximates exp(x) with a linear function exp(x_ss)*(x-x_ss), where x_ss is the steady state. Of course, a linear function will always hit negative values with some probability (as long as its slope is not 0). Using a second-order approximation means fitting a parabola to exp(x), thus preserving some of the curvature. This may avoid having negative values if the vertex is above 0 (but for extreme shocks does not preserve monotonicity of exp(x)). Th…

Looking at

THEORETICAL MOMENTS
VARIABLE         MEAN  STD. DEV.   VARIANCE
Y             11.0632     5.5452    30.7491

shows that output is strangely variable in your model. That indeed suggests a problem with the model itself.

[lovefukanglin]

1m

Thank you so much, professor. The steady state value of capital is more than 150, while steady state value of capital is only about 11, I am wandering if this is normal or strange. I do not know if the strangely variable output is induced by the calibrated parameters or by the large steady state value of capital, can you tell me your intuition？thanks.

[jpfeifer]

Steady state values themselves are a function of the parameters.

[lovefukanglin]

Thank you professor, I will spend time modify the estimated model in order to fit the real data well. Best wishes and always appreciate for your quick and kind helps.