No, the NK model is neither elegant nor useful. The IS/LM model was useful — but incomplete — because its core equations summarize observable empirical relationships between aggregates. The attempt to graft a forward looking ‘Phillips curve’ onto the model was an unmitigated disaster. @jasonfurman @FrancoisGeerolf @monacelt 1. The Euler equation is an elegant description of the first order condition of a representative agent with perfect foresight. That agent does not exist. A much better foundation for understanding how consumption moves over time was Friedman’s permanent income hypothesis. And since the work of Truman Bewley, we have a firm foundation for that equation that is grounded in borrowing constraints and incomplete markets in a heterogeneous agent model. 2. The NK Phillips curve does not characterize any observable empirical regularity. It never has. The original Phillips curve, a connection between wage inflation and unemployment in UK data, disappeared soon after Phillips published his eponymous article. The reason it takes a semester to teach the NK Phillips curve is that it requires a set of ugly non refutable theoretical contortions that only a mother could love. 3. The NK model has no unemployment. People voluntarily change their labor supply behavior in response to ‘shocks’, the consequences of which are correctly forecast. That assumption does not fit with what we know about labor markets. I blogged about that here rogerfarmer.com/rogerfarmerb… 4. The NK model assumes that employment (not unemployment) fluctuates around a stationary ‘natural rate of unemployment’. I wrote about that here static1.squarespace.com/stat… Fed economists spend countless hours attempting to estimate this ‘natural rate’. This is a Sisyphean task since the unemployment rate in data is essentially a random walk. The correct strategy at this point is to abandon much of the rational expectations revolution and to reconstruct macroeconomics beginning with what we knew circa 1960. This, I believe, is already beginning. There is a large body of theorists working on Bewley models and that group has concluded that rational expectations was a misstep. See, for example, the recent working paper by @ben_moll benjaminmoll.com/wp-content/… I am not as convinced as Ben, that our models must be nonlinear with global solution methods since, if Ben’s approach is successful, even linearized models in aggregates will likely display very different properties from either NK or classical RBC economies. But dropping the full implications of rational expectations is a promising way forward.

I spent a year at the Bank of England as a Senior Houblon Norman Fellow during the transition from Mervyn King to Mark Carney. Research economists briefed the Monetary Policy Committee ahead of their meetings but the information they presented could easily have been assembled by an AI. @cmatthes_econ @t_holden A second important input was the reports of the Bank’s regional agents who presented anecdotal evidence. For example, an agent might report back on his/her conversations with business representatives in Sheffield. That information would be harder to replicate by AI. A third important input, was from a few trusted individuals on staff who had the ear of the Governor. Monetary policy was run by discretion — not in any sense by ‘rules’. Prior to my Bank of England experience I spent several summers visiting the Econometrics group at the ECB. I naively believed, on arrival, that the econometricians would examine the data and convey a set of ‘facts’ to the research economists. The research economists would analyze the facts and convey to the policy makers the best strategy that could be offered by economic theory. In practice, that process runs in reverse. The policy makers decide what they want to do and they convey their opinions to the research department. The research economists go to the econometricians and say: make it so. In other words, the forecasts were reverse engineered to justify the policy actions. This is in no way a criticism of monetary policy in practice. It is a recognition that policies adapt to circumstances in ways that are not easily captured by mechanical rules.

There are other reasons to worry about unit roots static1.squarespace.com/stat… And to understand cointegration

This 👇

Congrats and I wish you every success in your new position @Birdyword

No, the belief function has the same methodological status as a production function or a utility function. Once one specifies a stationary probability distribution over shocks to beliefs the model becomes refutable and can be compared with the NK alternative. static1.squarespace.com/stat…

Beliefs are modeled as primitives that act as an equilibrium selection device. I call this a belief function. Prices are competitively determined in every period. Here is a link to an early representation of that idea static1.squarespace.com/stat… My equivalence conjecture is that dynamic indeterminacy is endemic to HA models once one moves beyond the simplest sessions of these models. A fruitful route to explore is the combination of incomplete markets with borrowing constraints with a perpetual youth environment.

💯Yep. Been there, done that.

I’m pretty sure the Fed now puts out consistent stock-flow accounts — but it’s been a while since I checked

I have work in progress with heterogeneous agents but I wouldn’t call it NK because the hallmark of NK models is, IMO, either Calvo or Rotemberg price stickiness. Everything I have worked on replaces artificial restrictions on price adjustment with an indeterminate monetary steady state in which slow price adjustment occurs endogenously. This is a link to the most recent iteration but there will be an updated version soon static1.squarespace.com/stat… The basic insight is that there is an isomorphism between OLG models and HA models and that the multiplicities that characterize OLG models are also ubiquitous in HA models.

Ahhh!! I never understood what that was all about. Every Dynamic Stochastic General Equilibrium model is stock-flow consistent by construction. One can argue about the equilibrium concept, but material balance conditions hold in every period.

that’s not an enlightening response @oktayozdn

SFC?

almost?

If you mean this paper static1.squarespace.com/stat… I remain unconvinced

🙏will read …

I recommend Post-Keynesian Dynamic Stochastic General Equilibrium Theory 😂 static1.squarespace.com/stat…

How does monetary policy work under a floor system George? … is there an accessible short piece on your thoughts on this @georgeselgin Because I’m not convinced that the price level is determinate under a floor system … and I’m not a believer in the FTPL, at least in its forward looking rational expectations incarnation

Perhaps. But the research economists at central banks are part of the graduating classes from the same institutions as freshly minted assistant professors. And in recent decades, the policy committees at the Fed, the Bank of England and the ECB have been dominated by academics.

Neutrality is a property of every well specified model —- super neutrality (neutrality of changes in the money growth rate) is not. But even in a model where money is neutral, monetary shocks can have permanent effects by nudging the system from one steady state to another. Here is a paper that explores that idea static1.squarespace.com/stat…