David and Goliath: The sunspot agenda meets the MIT machine

Two interesting developments occurred on the DSGE front this week. CEPR, the London based research organization, published an ebook on the future of Dynamic Stochastic General Equilibrium Models and Aurélian Saïdi and Beatrice Cherrier released an updated version of their history of the Penn/CEPREMAP research agenda on indeterminacy and sunspots. There could have been, and should have been, considerable overlap in these two events. Sadly there was not. 

For forty years, macroeconomics has been dominated by a battle between classical ideas emanating from the universities of Minnesota and Chicago, and New Keynesian ideas, emanating from MIT. In the past decade or so, MIT has been dominant and MIT trained economists have steamrolled an agenda that developed from Samuelson's interpretation of Keynes. According to this interpretation, the economy is Keynesian in the short run, when prices are sticky, and classical in the long-run, when all prices have had time to adjust. The MIT machine is the Goliath of my story.

A parallel agenda developed in the 1980s at the University of Pennsylvania in the U.S. and at CEPREMAP in France. This separate agenda accepted some but not all of the ideas of Minnesota and Chicago and married them with notions of  indeterminacy, sunspots and multiple equilibria as a way of understanding the macroeconomy. The sunspot/indeterminacy program is the David of my tale.

There is much to like and to agree with in the CEPR ebook, edited by Refet Gürkaynak and Cédric Tille. As with other projects in this series, it consists of a series of short articles from practitioners in the field, each giving a different perspective on the current state of the art in macroeconomic modeling. I particularly liked the piece by Del Negro and Giannoni explaining progress made at the Federal Reserve Bank of New York (FRBNY) on developing a suite of models for forecasting and policy evaluation. The FRBNY DSGE model has been around for a while and is now available as free open source code written in the programming language Julia. Julia, like MatLab, is a matrix based language compiled in real time. It is up to ten times faster than MatLab and free. Time (for me) to make the switch.

The CEPR ebook ends with speculation by Jordi Galí and Olivier Blanchard on the future of DSGE modeling. Here, I was disappointed by the inability of either Jordí or Olivier to acknowledge the body of knowledge developed over the past three decades by researchers in the sunspot/multiple equilibrium program. Neither Jordí nor Olivier are unaware of these ideas. Jordi has written sunspot papers and Olivier was the discussant on the Animal Spirits paper I wrote for the new Phelps volume where I showed how models with indeterminate steady state equilibria can explain hysteresis, a topic after Olivier’s heart.

Saïdi and Cherrier speculate that sunspot ideas did not capture the hearts of the profession because the researchers involved in the program failed to develop a common language and engaged in territorial disputes over property rights. There was no single leader to push the agenda. There is certainly an element of truth to that. Sunspots, self-fulling prophecies, animal spirits and confidence were all used interchangeably by different groups of authors to mean the same thing. Now there is yet another emergent power play by George-Marios Angeletos and Jennifer La’o to take the mantle. Rather than use sunspots, self-fulfilling prophecies or animal spirits, Angletos and La’o substitute yet another term ‘sentiments’ to mean the ability of non-fundamental shocks to influence economic activity.

There will always be turf battles over intellectual ideas. The stakes are high, particularly when these ideas are once more gaining prominence. It is a personal view, expanded on here, that the entire MIT inspired New Keynesian edifice is built on sand. It is hard to read the General Theory without recognizing the role of multiple equilibria. Not just multiple paths, as in the first generation sunspot models, but multiple indeterminate steady states, as in the second generation models based on search theory that I survey here.  By marginalizing that idea and building their models on ‘frictions’, the New Keynesians sold their collective soul.

In some circles, DSGE is a four-letter term of abuse. That is a pity and here, I am on the side of Gürkaynak and Tille. To quote from my 1993 book, The Macroeconomics of Self-Fulfilling Prophecies, [page 1]  “… the future of macroeconomics is as a branch of applied general equilibrium theory.”  For forty years or more, Post-Keynesian economists clung to the idea that, whatever Keynesian economics is about, it is not about sticky prices. I agree.   But unlike the Post-Keynesians, I am also a fan of DSGE modelling. As I argue here, by adopting the ideas of second generation models of indeterminacy and sunspots, it becomes possible to be a Post-Keynesian, as opposed to a New-Keynesian, and a DSGE fan, at the same time. 

I am optimistic, perhaps in vain, that David's sling-shot will one day, bring down Goliath.



Cách Nên Kinh Tê Vân Hành

For all my Vietnamese readers: the Vietnamese edition of my book, How the Economy Works, is now available.  You can order it here. My apologies to all those of you who read Vietnamese: Apple has not yet figured out how to add all of the accents used in Vietnamese.

You can now order How the Economy Works in hardback, paperback,  audio and on Kindle. Its also available in Hungarian and Vietnamese

Here is what one reviewer said about the book on Amazon

This is a timely book to spark debate & reflection following the September 2008 financial crisis that is still with us in 2010...   In this book ...   leading economist professor Roger E. A. Farmer dissects the current economic meltdown ; suggesting how the economics must change to get out of the doldrums. Professor Roger E. A. Farmer traces the swings between classical & Keynesian economics since the early twentieth century, explaining the elements of both theories. ... He adopts the principle that markets do not always work well & that capitalism needs some strategic guidance to improve. ...A highly recommended reading, clear & accessible to the general public to hopefully spark debate on this current important issue.  M. Mariba


Let's All Be Keynesians Now

Konstantin Platonov and I have updated our working paper on reinventing the IS-LM model. You can download it from the NBER here, or from an open link on this website. If you haven't read our paper, or our summary on VOX, here are a few highlights.

The Hicks-Hansen representation of The General Theory is summarized by the  IS-LM diagram, a tool that was designed to represent temporary equilibrium in the sense that Hicks used that term in his book, Value and Capital. The price level is  pre-determined in Hicks version of Keynesian economics and it wasn't until Samuelson introduced the idea of the neo-classical synthesis in the third edition of his undergraduate textbook that the current view of sticky-price Keynesian economics was born. Samuelson was the original 'bastard Keynesian' (in the immortal words of Joan Robinson.) 

In our working paper, we reinvent the IS-LM model and add a new curve, the NAC condition, which integrates the value of the stock market into the  model and explains, not just the interest rate and GDP, but also how the price level is determined. In our IS-LM-NAC model, high involuntary unemployment is not just a temporary phenomenon; it is a potentially permanent feature of a market economy. Larry Summers has resuscitated Alvin Hansen's idea of secular stagnation. Our work provides a coherent theory of what that might mean. Building on earlier work  that integrates The General Theory with general equilibrium theory in a new way, our paper offers fresh insights about the role of the asset markets on economic activity that cannot be understood using a more conventional approach.

A second feature of our paper is the introduction of animal spirits as an independent driver of economic activity. We formalize that idea  with a belief function, an idea that was first introduced in my 1993 graduate text, the Macroeconomics of Self-Fulfilling Prophecies. And as in my previous use of that concept, the idea that beliefs are fundamental is completely consistent with rational expectations. Our IS-LM-NAC model provides a coherent way for policy makers to think about the effects of monetary and fiscal policy on output, employment and inflation and we propose that our model be adopted as a viable and superior alternative to the current textbook approach.



Post Keynesian Dynamic Stochastic General Equilibrium Theory

Last year, I was invited to present a keynote address to the 20th annual conference of the FMM Research Network on Macroeconomics and Macroeconomic Policies, “Towards Pluralism in Macroeconomics”, held in Berlin on October 20th – 22nd 2016. Due to unforeseen circumstances, I was  unable to attend in person but the organizer, Torsten Niechoj, kindly invited me to submit a paper for the conference volume. That paper is now available as an NBER working paper here, and as a CEPR discussion paper here. In it, I make the case that Post-Keynesians and New-Keynesians have much to learn from each other and I invite the next generation of macroeconomists to help develop a fresh approach to our subject: Post-Keynesian Dynamic Stochastic General Equilibrium Theory.

Image from: The Rocky Horror Picture Show

Image from: The Rocky Horror Picture Show

Here is an extract from the paper, that reports a conversation overheard between Janet, a North American economist from the Great Lakes region, and Brad, a Post Keynesian graduate student who has read my book, Prosperity for AllJanet seems a bit confused over Gene Fama's use of the word 'efficiency' in the efficient markets hypothesis. Brad explains that Fama is talking about informational efficiency and that this is distinct from Pareto Efficiency. Janet thinks that these are the same thing. Brad disagrees.

Janet:      The distinction between informational efficiency and Pareto efficiency is a distraction. If people trading in the financial markets cannot make money, nor can the government. There is no such thing as a free lunch. 

Brad:    I disagree. The fact that a market is in equilibrium does not mean that it is efficient.

Janet:    Well OK, I know that you Post-Keynesian types are willing to make all sorts of assumptions about market frictions. Those assumptions don’t seem credible to me. As a first approximation, I am willing to assume that prices are perfectly flexible.

Brad:    You seem to be confusing me with a New-Keynesian. I’m happy with the flexible price assumption.  I’ll give you that one.

Janet:    Well perhaps you think that there are missing financial markets. I know that Kenneth Arrow has argued that transactions costs preclude the existence of all the financial instruments necessary to generate market efficiency. But I don’t buy that. If there are big potential gains from trade, there are big incentives for private agents to create new markets. Just look at all the derivatives that were created over the last few decades. They came about because of lower transactions costs. I’m willing to assume, to a first approximation, that there is a complete set of financial markets.

Brad:    While that seems like a stretch; I’ll give you that one too. Let’s assume that there is a complete set of financial markets and that everyone participating in the financial markets can make any conceivable trade in the futures markets at zero cost.

Janet:    Is it the assumption of perfect knowledge you’re uncomfortable with? I know that Frank Knight  drew a distinction between risk and uncertainty and that you Post Keynesian types keep harping on about radical uncertainty. Although you may have a point there, I just don’t see how we can make any progress if we assume that nobody knows anything about the future. I’m willing to go with the assumption that we live in a stationary world because it’s the best chance we have of saying something about the behavior of short-run macroeconomic variables. And that means that we should also assume that people have rational expectations.

Brad:    Well I’m not entirely on board with that. But, for the sake of argument, let’s agree to model the way that human beings would act if they did live in a world where all uncertainty is generated by a known stationary probability distribution and where the people in our model have rational expectations. If we can agree on the answer to what an equilibrium would look like in that world, then perhaps we can extend our equilibrium concept to a more complicated world where the future is characterized by radical uncertainty.

Janet:    Hmmm… If you accept all of these assumptions, I’m having trouble understanding why you don’t understand, that if private agents can’t make money in markets, then, neither can the government. Maybe you’re one of those Marxist types who thinks that product markets are characterized by monopolists and we need unions to defend workers’ rights. In my view, that’s a load of poppycock. Product markets are contestable and, to the extent they’re not, the solution is limited regulation. I don’t see what that has to do with the distinction between informational efficiency and Pareto efficiency. 

Brad:    Wow: you really have drunk the Kool-Aid. But no; that’s not my point either. Let’s suppose you’re right that the labor market and the product markets are well approximated by the assumption of perfect competition. I’m willing to give you flexible prices, complete markets, rational expectations, and perfect competition. But there’s one little fact you can’t get around.

Janet:    What’s that Brad?

Brad:    The Grim Reaper. People die and new people are born. Even if everyone present today could make trades with each other contingent on every conceivable future event, the unborn cannot participate in markets that open before they are born. Government, on the other hand, is present in every period and it can intervene on behalf of our children and our grandchildren.

Janet:    Nonsense. We don’t need government for that; we need the family. Robert Barro showed that all we need to correct the inefficiency in an overlapping generations model is for parents to love their children. The representative household is a useful fiction because each of us is connected by a chain of operative bequests. 

Brad:    You’re wrong Janet. Overlapping generations models have not one, but two kinds of inefficiencies. Robert Barro’s argument applies to dynamic inefficiency, which is the fact that, in overlapping generations models, interest rates can be too low. There is a second kind of inefficiency that was pointed out by David Cass and Karl Shell . And Costas Azariadis (Azariadis, 1981) showed that overlapping generations models can lead to volatile fluctuations in asset markets that have nothing to do with fundamentals. In his book, Prosperity for All Farmer argues that a large fraction of the fluctuations we see in stock markets are caused by this second kind of inefficiency. And this second kind of inefficiency cannot be eliminated by the family because it would require that our parents leave positive bequests in some states of nature and negative bequests in other states of nature.

My imagined conversation between Janet and Brad is meant to illustrate the idea that we can accept the tenets of a version of general equilibrium theory while rejecting the first welfare theorem.

It is my hope that the shock of the Great Recession will catalyze interbreeding between new-Keynesian and heterodox economists. If I am right, more of my neo-classical contemporaries will need to listen to the drum beat that post-Keynesians have been sounding for sixty years. And Post-Keynesians will need to explain to neo-classical and new-Keynesian economists, in their own language, what they are doing wrong. General equilibrium theory, broadly interpreted, like mathematics, is a language.

If you are young enough to have not yet been corrupted by establishment elites of either subspecies, I urge you to think hard about joining me in establishing Post-Keynesian DSGE theory as the future of macroeconomics. 

Keynesian Economics Without the Consumption Function

In Prosperity for All,  (PFA) I describe a theory of Keynesian economics, developed in my recent body of work,  in which the transmission mechanism from demand to employment is through wealth, not through income. I call this, a theory of Keynesian economics without the consumption function.

I reject the Keynesian theory of aggregate demand, and I reject the theory of the multiplier that goes along with it. My reason is that the empirical evidence does not support a theory in which government expenditure  multipliers work by increasing consumption, as they must if the Keynesian consumption theory is correct. As Valerie Ramey has shown, when government expenditure goes up; consumption goes down.

Here is how I describe my reluctance to embrace the traditional Keynesian approach to aggregate demand in Chapter 10 of PFA,

Keynesian economics has two parts: a theory of aggregate supply and a theory of aggregate demand. Traditionally, Keynesians have focused on the theory of aggregate demand. The central part of that theory is the consumption function, and an implication of the theory of the consumption function is that an increase in government expenditure will cause GDP to increase by a multiple of the initial increase in spending. That theory is wrong. Consumption depends on wealth, not on income. PFA, pages 177-178

If consumption depends on wealth, and not on income, we should be concerned that movements in wealth may lead to socially inefficient fluctuations in the unemployment rate. Those movements are not simply temporary movements away from a social planing optimum, they are permanent movements in the unemployment rate that can lead to decades of misery for those without jobs.

Figure 1: Roger E. A. Farmer. (c) Oxford University Press.

Figure 1: Roger E. A. Farmer. (c) Oxford University Press.

Quoting again from PFA,

Figure 1 illustrates the implications of a theory of Keynesian economics without the consumption function. The aggregate demand curve does not slope upward with income; it is a horizontal straight line. The position of this line depends on the beliefs of market participants about the value of their financial assets. As the value of financial assets fluctuate, driven by self-fulfilling beliefs, so the aggregate demand curve moves up and down between the solid horizontal line and the dashed horizontal line. As people feel more or less wealthy, they buy more or fewer goods. Firms hire more or fewer workers and real GDP fluctuates between point YA* and YB*.  (PFA, page 178.)

But although, I reject the simple Keynesian version of Aggregate Demand, I do not reject Keynesian economics. The key idea in the General Theory is that high involuntary unemployment may persist as an equilibrium of a market economy. How can that be? In Prosperity for All,

I provide a foundation—Keynesian search theory—to the Keynesian theory of aggregate supply. This new theory is rooted firmly in the microeconomic theory of behavior. According to Keynesian search theory, everything demanded will be supplied and any unemployment rate can be an equilibrium unemployment rate.  (PFA pages 178-9.)  

The implications of these ideas, taken as a package, are profound. If demand works through wealth, the right policy to maintain full employment is an intervention in the asset markets, not in the goods markets.