Specifying a parameter as 2nd order

[kesman]

Hi all,

I am trying to build a DSGE model with a portfolio choice problem as in Devereux and Sutherland (2007, 2011) and Tille and Van Wincoop (JIE, 2010). I include a proportional holding cost for foreign equity. As in Tille and Van Wincoop (2010), it is convenient to assume this cost parameter is second order, that is proportional to the variance of shocks. Hence it will not affect the steady state nor the 1st order approximation of the model, but will affect the 2nd order approximation and higher. Is there a way to specify a parameter as 2nd order?

It is easy to obtain the correct steady state (I use DS method for portfolio) and 1st order solution by simply ignoring the cost. However, the problem comes when using the stochsimul command to obtain 2nd order solution. Supposedly the algorithm performs both a 1st order approximation and 2nd order approximation to obtain the 2nd order solution. Here I don’t see a way to tell Dynare to ignore the cost in the 1st order approximation and simultaneously include it in the 2nd order approximation.

Any thoughts on how to get around this problem? Thanks in advance.

[jpfeifer]

Sorry, but the Devereux/Sutherland approach is not directly supported by Dynare. You have to rely and their codes to make this work. I have never worked with it so I cannot provide support.

[kesman]

Thank you for your answer. I was possibly unclear but answering the question requires technical knowledge about the properties and use of Dynare but not necessarily about the DS-method. I think specifying a parameter as 2nd order is not directly supported by Dynare but there might be a trick to get past this.

[jpfeifer]

Isn't that a standard property of quadratic adjustment costs? At first order, they evaluate to 0 in steady state. However, the first derivative is not 0, making them influence the dynamics.