The BIS Quarterly Review had a recent paper on r* (the preferred term for the “natural rate of interest”) by Benigno, Gianluca, Boris Hofmann, Galo Nuño Barrau, and Damiano Sandri.
thanks for reviewing our paper! I read with interest your posts indeed and have learned a lot from it along with other contributors that are not necessarily mainstream. In all fairness though, our piece is quite critical of the standard approach. As you note it aims at providing an overview (short because of space constraint) of the r* concept. At the same time in the monetary policy section we point out at less mainstream perspective on medium rum real rates and their determinants. If anything I would say that there is a bit of a critical perspective on r* as a guidepost for monetary policy. In my next posts I will provide an alternative concept that points out at the importance of financial factors in dictating the business cycle.
Thanks for a courteous reply to my rather snarky article. My excuse is that I have written the same r* article about a dozen times over the past decade, and I am quite busy now (hence the late reply to the comment). The comments about the uncertainty about policy guidance are why I looked at the paper in the first place (and I obviously agree with). I noted that as best I could, but I wanted to note my concerns about the technically meaty part of the article.
Although my conclusions are theoretical nihilism, I think there is an angle that can be explored (which admittedly leads to toxic conclusions if I am correct). Create a model that generates the economic data needed for a r* fitting, but interest rates have a negligible effect. Then, take the same "real economy" scenario parameters, but policymakers use different reaction functions (creating multiple scenarios that mainly have different policy rates, with mild differences in the real economy variables). We then run the model data through one (or more) r* estimators. Do the r* estimates look "reasonable" (when compared to how r* estimates relate to the observed actual economy), but end up with different levels?
(My intuition is that they should, but I have been too lazy to do that work - most descriptions of r* estimators are cryptic when compared to what I was used to in control systems. Since such a project was tangential to my writing interests, I did not pursue it. I am supposed to have finished an introductory book on inflation in December 2023, so things are obviously not going well on that front.)
hi Brian: many thanks. yes if I understand your guess there are few ways in which it is possible to expose the problems of estimating r*. One approach as you suggest would consist in having faith in model and treat is a a true data generating process. Then run estimates using some common approaches on the generated data and compare estimates with the "correct" r*. People might object saying that you would then need to estimate the model and then see what is the difference between the implied r* from the estimated model on out of sample data with the "correct" one. I suspect that there would be errors that could go in both direction but a "good" model would track the direction of the "correct" r* one. Another way would be to run some of the models that are out there using different inputs for interest rates and then see what the implied estimates comes out. I suspect that some of the econometric model tends to track the input in terms of interest rate that you feed into the model. In any case I would argue that it is difficult to know what is the measure of an object that is not observable. More fundamentally our review article was suggesting (in the last part) that are elements that are missing in the characterization of the way the economy works within the mainstream approach. I am looking forward to reading your book on inflation (I have myself few blogs on that from my time at the NY Fed and a new one that should come out soon that you might find of interest) and your future posts.
On a roll. Show me where 10Y real yields are and I will tell you where all the illuminated will gravitate towards with their measure of r*. And real yields are mostly a reflection of real policy rate expectations. All rather un-luminating, so to speak.
It seems that what you're saying is that DSGE models are guilty of the philosophical sin described by Bertrand Russell: "The method of 'postulating' what we want has many advantages; they are the same as the advantages of theft over honest toil."
Hi Brian:
thanks for reviewing our paper! I read with interest your posts indeed and have learned a lot from it along with other contributors that are not necessarily mainstream. In all fairness though, our piece is quite critical of the standard approach. As you note it aims at providing an overview (short because of space constraint) of the r* concept. At the same time in the monetary policy section we point out at less mainstream perspective on medium rum real rates and their determinants. If anything I would say that there is a bit of a critical perspective on r* as a guidepost for monetary policy. In my next posts I will provide an alternative concept that points out at the importance of financial factors in dictating the business cycle.
Thanks for a courteous reply to my rather snarky article. My excuse is that I have written the same r* article about a dozen times over the past decade, and I am quite busy now (hence the late reply to the comment). The comments about the uncertainty about policy guidance are why I looked at the paper in the first place (and I obviously agree with). I noted that as best I could, but I wanted to note my concerns about the technically meaty part of the article.
Although my conclusions are theoretical nihilism, I think there is an angle that can be explored (which admittedly leads to toxic conclusions if I am correct). Create a model that generates the economic data needed for a r* fitting, but interest rates have a negligible effect. Then, take the same "real economy" scenario parameters, but policymakers use different reaction functions (creating multiple scenarios that mainly have different policy rates, with mild differences in the real economy variables). We then run the model data through one (or more) r* estimators. Do the r* estimates look "reasonable" (when compared to how r* estimates relate to the observed actual economy), but end up with different levels?
(My intuition is that they should, but I have been too lazy to do that work - most descriptions of r* estimators are cryptic when compared to what I was used to in control systems. Since such a project was tangential to my writing interests, I did not pursue it. I am supposed to have finished an introductory book on inflation in December 2023, so things are obviously not going well on that front.)
hi Brian: many thanks. yes if I understand your guess there are few ways in which it is possible to expose the problems of estimating r*. One approach as you suggest would consist in having faith in model and treat is a a true data generating process. Then run estimates using some common approaches on the generated data and compare estimates with the "correct" r*. People might object saying that you would then need to estimate the model and then see what is the difference between the implied r* from the estimated model on out of sample data with the "correct" one. I suspect that there would be errors that could go in both direction but a "good" model would track the direction of the "correct" r* one. Another way would be to run some of the models that are out there using different inputs for interest rates and then see what the implied estimates comes out. I suspect that some of the econometric model tends to track the input in terms of interest rate that you feed into the model. In any case I would argue that it is difficult to know what is the measure of an object that is not observable. More fundamentally our review article was suggesting (in the last part) that are elements that are missing in the characterization of the way the economy works within the mainstream approach. I am looking forward to reading your book on inflation (I have myself few blogs on that from my time at the NY Fed and a new one that should come out soon that you might find of interest) and your future posts.
1) The inflation book is extremely basic, as it covers all the silly stuff you run into on the internet in financial commentary.
2) I may do another piece early next week discussing my r* concerns. It would be a rework of something I wrote years ago, but hopefully clearer.
On a roll. Show me where 10Y real yields are and I will tell you where all the illuminated will gravitate towards with their measure of r*. And real yields are mostly a reflection of real policy rate expectations. All rather un-luminating, so to speak.
It seems that what you're saying is that DSGE models are guilty of the philosophical sin described by Bertrand Russell: "The method of 'postulating' what we want has many advantages; they are the same as the advantages of theft over honest toil."