Taylor Rule
Posted: Sat Jul 24, 2010 7:06 am
I am interested in specifying, estimating and generating one (possibly multi-step) ahead forecasts of base rates using (a variation) of the Taylor Rule, across US, UK, Europe, Japan, but starting with US.
There are numerous versions of the Taylor Rule but the classical version appears to be:
i = istar + [ 0.5(pi -pistar) + 0.5(y-ystar)]
where,
i = Taylor prescribed policy rate
istar = neutral policy rate (possibly defined as Neutral Fed funds rate + Fed’s inflation target)
pi = proxy for inflation
pistar = Fed’s inflation target
y = actual output level
ystar = economy’s potential level of output
I have found the following working paper by:
Qin and Enders (2006) A Comparison of the In-Sample and Out-of-Sample Properties of Linear and Nonlinear Taylor Rules Using Real-Time Data, Department of Economics, University of Alabama.
http://www.cba.ua.edu/~wenders/Qin_Enders_Rewrite.pdf. This has subsequentlly appeared in the Journal of Macroeconomics, in 2008.
A less thorough approach than Qin and Enders (2006), is via the following link, specifically p6, of the following
http://www.ssc.wisc.edu/~mchinn/Rosenberg_Taylor.pdf
Amongst the various issues in specifying the Taylor Rule, one needs to measure:
A. Inflation gap (pi – pistar)
B. Output gap (y – ystar)
For the Inflation gap one could possibly use: (Core personal consumption expenditure rate – Fed’s inflation target)
For the Output gap one could be to use (un)employment numbers instead i.e. Output gap = [okum factor * (natural rate of unemployment – actual unemployment rate)], where okum factor is arbitrarily set at 2.
Both of these approach’s are stated in the second article.
Further, to generate monthly step ahead forecasts the Inflation gap & Output gap need to be lagged.
As is standard practice:
1. I’d like to specify in-sample (with usual tests on residuals for IID etc).
2. Run rolling regressions using various windows out-of-sample.
3. Then be able to update and generate one (possibly multi-step) ahead forecasts every month.
Any views/guidance on any of the above methodologies are appreciated along with suggestions for relevant recent literature and any work already in RATS.
There are numerous versions of the Taylor Rule but the classical version appears to be:
i = istar + [ 0.5(pi -pistar) + 0.5(y-ystar)]
where,
i = Taylor prescribed policy rate
istar = neutral policy rate (possibly defined as Neutral Fed funds rate + Fed’s inflation target)
pi = proxy for inflation
pistar = Fed’s inflation target
y = actual output level
ystar = economy’s potential level of output
I have found the following working paper by:
Qin and Enders (2006) A Comparison of the In-Sample and Out-of-Sample Properties of Linear and Nonlinear Taylor Rules Using Real-Time Data, Department of Economics, University of Alabama.
http://www.cba.ua.edu/~wenders/Qin_Enders_Rewrite.pdf. This has subsequentlly appeared in the Journal of Macroeconomics, in 2008.
A less thorough approach than Qin and Enders (2006), is via the following link, specifically p6, of the following
http://www.ssc.wisc.edu/~mchinn/Rosenberg_Taylor.pdf
Amongst the various issues in specifying the Taylor Rule, one needs to measure:
A. Inflation gap (pi – pistar)
B. Output gap (y – ystar)
For the Inflation gap one could possibly use: (Core personal consumption expenditure rate – Fed’s inflation target)
For the Output gap one could be to use (un)employment numbers instead i.e. Output gap = [okum factor * (natural rate of unemployment – actual unemployment rate)], where okum factor is arbitrarily set at 2.
Both of these approach’s are stated in the second article.
Further, to generate monthly step ahead forecasts the Inflation gap & Output gap need to be lagged.
As is standard practice:
1. I’d like to specify in-sample (with usual tests on residuals for IID etc).
2. Run rolling regressions using various windows out-of-sample.
3. Then be able to update and generate one (possibly multi-step) ahead forecasts every month.
Any views/guidance on any of the above methodologies are appreciated along with suggestions for relevant recent literature and any work already in RATS.