Hi Willi. Great content. May we ask questions here? Was wondering why sigma should be lower if there are more shocks. The shocks are independent, right? In some books, they put a parameter in front of sigma, say, eta*sigma and they maintain sigma = 0.01.
If you do simulations with all shocks, you get a much higher variance of the variables which is not realistic. If you have only one shock all the variability stems from this shock. That's why typically, the more shocks you have, the smaller the standard error. Whether you scale the shock by putting a parameter in front is a matter of taste. Typically people than assume an identity matrix in the shocks block and scale the actual variance by this parameter in front. But this works only at a first-order approximation. So it is just a matter of taste. I like to keep it straight, so if I have a value for the standard error of my shock, I put it into the shocks block.
Thank you so much Prof. Mutschler, this is very helpful. I have a quick question, if you are working with production function with labor only and technology, no capital, can you still work with share of labor income to total income as 1 minus alpha, or that will effectively be 1?
Yes you can still calibrate alpha to the labor income share; note that typically in such models it is assumed that there is capital, but it is fixed and then normalized to 1.
Hey. I was wondering why we don’t use homotopy setup on the target variable and make the parameter endogenous to get the calibrated parameter?
You’re the best…thank you!
Many thanks ,very useful job
No problem
Sir, will you upload videos on quantitative macroeconomic modelling with Vector Autoregression Models?
Yes in the fall I am teaching a course on this and plan to release such videos.
Hi Willi. Great content. May we ask questions here? Was wondering why sigma should be lower if there are more shocks. The shocks are independent, right? In some books, they put a parameter in front of sigma, say, eta*sigma and they maintain sigma = 0.01.
If you do simulations with all shocks, you get a much higher variance of the variables which is not realistic. If you have only one shock all the variability stems from this shock. That's why typically, the more shocks you have, the smaller the standard error.
Whether you scale the shock by putting a parameter in front is a matter of taste. Typically people than assume an identity matrix in the shocks block and scale the actual variance by this parameter in front. But this works only at a first-order approximation. So it is just a matter of taste. I like to keep it straight, so if I have a value for the standard error of my shock, I put it into the shocks block.
Thank you so much Prof. Mutschler, this is very helpful. I have a quick question, if you are working with production function with labor only and technology, no capital, can you still work with share of labor income to total income as 1 minus alpha, or that will effectively be 1?
Yes you can still calibrate alpha to the labor income share; note that typically in such models it is assumed that there is capital, but it is fixed and then normalized to 1.