Beliefs, Networks, History and the Housing Premium Puzzle

This is week five of my posts featuring research presented at the conference on Applications of Behavioural Economics and Multiple Equilibrium Models to Macroeconomics Policy Conference held at the Bank of England on July 3rd and 4th 2017. Today’s post features the coauthored work of Héctor Calvo-Pardo, and a series of coauthored papers by Alan Taylor, a co-organizer of the conference. 

network.png

Hector Calvo-Pardo, from the University of Southampton, presented his paper on social networks, coauthored with Luc Arrondel Research Director at CNRS in Paris, Chryssi Giannitsaro of Cambridge University and Michael Haliassos of Goethe University. Alan Taylor, a Professor at UC Davis, helped organize the conference. In a linked video, he discusses an amazing new data set developed jointly with Òscar Jordà, Vice President of the Federal Reserve Bank of San Francisco and Moritz Schularick, Professor of Economics at the University of Bonn.


Arrondel, Calvo-Pardo, Giannitsaro and Haliassos (ACGH) conduct work on social networks that fits beautifully into the theme of this conference.  In the opening Macro Memo  I explained how economic models with multiple equilibria could be combined with psychological models of belief propagation to advance our understanding of financial panics. The ACGH paper “Informative Social Interactions” is a very nice example of the use of network analysis to explain the propagation of ideas.

Quoting from the paper [page 2]

“…we design, field and exploit novel survey data that provide measures of stock market participation (relative to individuals’ financial wealth), connectedness, but also of subjective expectations and perceptions of stock market returns via probabilistic elicitation techniques.  Our empirical analysis exploits cross-sectional variation for a representative sample by age, asset classes and wealth of the population of France, collected in two stages, in December 2014 and May 2015.  
… the questionnaire [also] contains a rich set of covariates for socioeconomic and demographic controls, preferences, constraints and access and frequency of consultation of information sources, typically absent from social network empirical studies.

And this is where networks come in:

[our data set] … contains specific questions designed to obtain quantitative measures of relevant network characteristics that enable identification of information network effects on financial decisions from individual answers…

ACGH break up a person’s network into a small inner circle of contacts who are financially knowledgeable, and a separate larger set of contacts who are not.  They find significant peer effects from contacts in the financial network to a person’s beliefs about future financial variables. The paper provides evidence of how beliefs spread through social networks, a mechanism that may help to promote the spread of asset prices bubbles. I recommend listening to Hector’s explanation in his own words, linked above.

This brings me to a fascinating series of papers by Òscar Jordà, Moritz Schularick and Alan Taylor (JST). To my knowledge, they have so far produced five papers, “When Credit Bites Back: Leverage, Business Cycles, And Crises, (2011), “Sovereigns versus Banks: Credit, Crises and Consequences" (2013), “The Great Mortgaging: Housing Finance, Crises, and Business Cycles”, (2014), “Betting the House”, (2015), and the paper I know best “The Rate of Return on Everything” (2017), which I saw presented at the NBER Summer Institute in July of 2017. This paper has two additional co-authors Katharina Knoll and Dmitry Kuvshinov who are/were both students of Moritz at the University of Bonn. 


The unifying theme in all of these papers is a new data set that the authors have painstakingly assembled. Here I quote from JKKST (2017):

Our paper introduces, for the first time, a large dataset on the rates of return on all major asset classes in advanced economies, annually since 1870. Our data provide new empirical foundations of long-run macro-financial research. Along the way, we uncover new and somewhat unexpected stylized facts.
… Notably, housing wealth is on average roughly one half of national wealth in a typical economy, and can fluctuate significantly over time… But there is no previous rate of return database which contains any information on housing returns.

Screen Shot 2017-09-02 at 1.42.03 PM.png

Table 1 shows the extent of the coverage which includes sixteen advanced economies for 145 years. 

There are many nuggets to be mined here, some of which have already been dug out by the authors. For example, it is well known that the return to equity has been 5% higher than the return to T-bills in a century of US data, a fact that Rajnish Mehra and Ed Prescott labeled the equity premium puzzle.  JKKST provide evidence that the equity premium puzzle is universal across all sixteen economies and, in addition, there is a second ‘housing premium puzzle’, documented in Table 2. This second puzzle is all the more intriguing since the variance of returns to housing are significantly lower than the returns to equities.

Screen Shot 2017-09-02 at 1.50.24 PM.png

JKKST focus on the relationship between average stock market returns, “r” and average growth rates “g”. The connection between r and g has gained widespread notoriety since Thomas Piketty revived the Marxist claim that capitalism will self-destruct as the rich grow richer in his acclaimed tome, Capitalism in the 21st Century. I suggested at the recent 2017 NBER Summer Institute, that Alan and his coauthors look instead, at the connection between the safe rate of return and the growth rate. If, as I suspect, the safe rate of return is roughly equal to the growth rate across these data, it suggests (to me) that Samuelson’s biological rate of interest is at play. It’s time to start teaching our graduate students more about the Overlapping Generations Model, just one of the topics I’ll be covering in a series of ten lectures on Indeterminacy and Sunspots in Macroeconomics, as part of the Swiss Doctoral Programme at Gerzensee this week.

Short Sharp Shocks

This is week four of my posts featuring research presented at the conference on Applications of Behavioural Economics, and Multiple Equilibrium Models to Macroeconomics Policy Conference held at the Bank of England on July 3rd and 4th 2017.

money.jpg

Today’s memo features two economists working on models of multiple equilibria from different perspectives. George Evans is a pioneer in models of adaptive learning, a topic he has worked on for more than thirty years. George presented his joint work with Seppo Honkapohja, Deputy Governor of the Bank of Finland, and Kaushik Mitra, Professor of Economics at the University of Birmingham. Patrick Pintus, a Researcher at the Banque de France, presented a co-authored paper with Yi Wen, an Assistant Vice-President at the Federal Reserve Bank of St. Louis and Xiaochuan Xing from Yale University.  The papers presented by both of these authors make small changes to a relatively conventional monetary Dynamic Stochastic General Equilibrium (DSGE) Model. Both of them reach non-mainstream conclusions by exploiting the fact that monetary equilibrium models always contain multiple equilibria.


George Evans began his work on adaptive learning in his Ph.D. dissertation at Berkeley in the early 1980s. When the rest of the profession was swept up by the rational expectations revolution, George persevered with the important idea that perfectly correct beliefs about the future cannot be plucked from the air, they must be learned. For an introduction to George’s work, I highly recommend the book co-authored with his long-time co-author, Seppo Honkapohja.


The paper of Evans, Honkapohja and Mitra (EHM), begins with a theme we met in post two where I discussed the fact that the standard New Keynesian model, in which the central bank follows a Taylor Rule, has two steady state equilibria. The intellectual foundation for that idea comes from the work of Jess Benhabib, Stephanie Schmitt Grohé and Martín Uribe, (BSU). BSU pointed out that the money interest rate cannot be negative. It follows from that observation that the Taylor Rule must be non-linear. 

ESM picture.png

The evidence for multiple steady states is presented in Figure 1 which is taken from George's paper. The dashed line is called the Fisher equation, after the American economist Irving Fisher. This graphs the relationship we would expect to hold between the money interest rate and the inflation rate if the real interest rate is constant. The solid line is an estimated Taylor Rule that takes account of the non-linearity in the central bank’s response to inflation that arises from the existence of the lower bound.  A steady state is an inflation rate and an interest rate that satisfies both of these equations. Notice that these two curves intersect twice, one at an interest rate of roughly 2.5% and one with an interest rate close to zero.

Evans, Honkapohja and Mitra (EHM) build on this idea by adding a theory of adaptive learning. I am often asked how my own work on the belief function is related to George’s work on adaptive learning. They are very closely linked. I agree with George that expectations, aka beliefs, are not plucked from the air. They must be learned. In models where there is a continuum of equilibria, like the ones I work with, it is beliefs that select which equilibrium will prevail. George’s work on adaptive learning provides a micro-foundation for what I have called the belief function.

Previous work has shown that, in the basic New-Keynesian model, the upper steady state is stable under adaptive learning but the lower steady state is not. They modify the basic model by adding the assumption that the rate at which prices and output can fall has a lower bound. They show that this assumption implies that there exists a third steady state in which recessions can be persistent and deep.

Figure 2

Figure 2

Figure 2 illustrates the dynamics that arise from adaptive learning in their model. The three steady states, A B and C are respectively the target 2% inflation steady state, the zero-lower bound steady state and the deflation steady state that arises from EHM’s assumption that deflation is bounded below. Importantly, steady states A and C are stable; steady state B is not. Most of the time, the economy is hit by shocks that keep it in region A. But occasionally a large shock, like the Great Recession, knock it over into region C and, when that happens, it may be very difficult to escape.

George and his co-authors use their analysis to argue that a large fiscal intervention, a short-sharp shock, can knock the economy out of region C and back into region A. Readers of this blog will know that I have expressed scepticism of that idea in the past, largely because I am not a big fan of the basic NK model. However, this is the most convincing rationale in favour of a large fiscal stimulus that I have yet seen. The mechanism that ESM propose works by permanently shifting expectations; that mechanism is likely to be at work in many economic models and I am pleased to see the George’s agenda is once again getting exposure. If you are a young researcher who is thinking of working in macroeconomic theory and policy, consider working on models of expectations formation. 


Next, I will turn to the work of Patrick Pintus, Yi Wen and Xiaochuan Xing (PWX). I have been a fan of Patrick’s work since he was a graduate student in Paris working with Jean-Michel Grandmont and I have followed Yi Wen’s papers closely since we first met at a conference in New York many years ago. Yi wrote the state of the art paper on why models of increasing returns to scale should be taken seriously and it is no surprise that a collaboration that involves the two of them would produce ideas worth listening to. This is my first exposure to Xiaochuan Xing, and I am sure we will hear much more of him in the coming years.


PWX take up a puzzle that has long been known to plague the equilibrium real business cycle (RBC) model that has dominated macroeconomic theory for more than thirty years. That model predicts that when interest rates are high, the economy will soon enter an expansion. The reality is different. High interest rates are an omen that a recession is coming down the road. What are the features of the real world that are missed by the classical RBC paradigm?

PWX relax the RBC model in two ways. First, they introduce the realistic assumption that it is difficult or impossible to borrow in large amounts without providing collateral. Second, they recognize that many loan contracts are arranged with variable-rates as opposed to fixed-rates. By combining these two assumptions with an otherwise standard business cycle model, they arrive at a model where many different outcomes can occur in equilibrium. The alert reader will by now, have picked up the theme: this is a model with multiple equilibria where outcomes are driven by beliefs.

The fact that PWX are able to construct a theoretical model with multiple equilibria is a first step: But does this model help to explain data? Until recently, much of the work on multiple equilibria consisted of esoteric calibrated theoretical models. The reason for that was two-fold. First, most graduate students were not exposed to the potential for multiple equilibrium models to explain data. And second, the techniques that were available to confront those models with data had not been developed. That began to change when Thomas Lubik and Frank Schorfheide showed in 2004 how to estimate a model with a set of indeterminate equilibria. That agenda was advanced further when Farmer, Khramov and Nicolò (FKN) developed a simple method for implementing their idea using standard software packages.  FWX use the FKN technique to estimate their model using US data and they find that roughly 25% of the variance in GDP is caused by animal spirits. You can hear Patrick discuss these ideas in the video linked above.

Agent-Based Models and Loss Aversion in the UK Housing Market

This is my third post featuring research presented at the conference on Applications of Behavioural Economics, and Multiple Equilibrium Models to Macroeconomics Policy Conference held at the Bank of England on July 3rd and 4th 2017.

houses.jpg

Last week I featured two US central bankers, Jim Bullard, President of the St Louis Fed and Kevin Lansing, a Research Advisor at the Federal Reserve Bank of San Francisco. This week’s post features two papers on the housing market, presented by two researchers at the Bank of England; Arzu Uluc and Philippe Bracke. One of these papers uses an Agent Based-Modelling approach to study the effects of macroprudential policy in the housing market. The other is a large-scale empirical study which finds significant evidence against the hypothesis that we are all perfectly rational. Homo Economicus and Femina Economica display more subtle behaviour than simple neo-classical models admit. 


Let’s begin with a fascinating paper co-authored by Philippe Bracke and Silvana Tenreyo. You can find a video link where Philippe discusses their research by clicking on the image here.


Philippe is a senior economist at the Bank of England and Silvana is a Professor at the London School of Economics and she is also the latest addition to the Bank of England’s Monetary Policy Committee.

Philippe and Silvana’s paper studies history dependence in the UK housing market. They ask a simple and fascinating question. Imagine that there are two comparable houses for sale. They are both in the same neighbourhood.  They are both three-bedroom semi-detached houses with two bathrooms. Both houses have a front garden and off-street parking. Both houses are five-minute’s walk from an underground station. But they differ in one important respect. House A is owned by a person who bought the house when house prices were at an all-time high. House B is owned by someone who bought it when house prices were at an all-time low. By studying a comprehensive data set of house sales in the UK, Bracke and Tenreyo find that the owner of house A will typically hold out for a higher price than the owner of house B.

That is an interesting finding because rational Femina Economica should care only about the price. History should not matter. If that was the end of the story, the paper would be an interesting contribution to the literature and would likely be published in a top journal. But Philippe and Silvana go further. They distinguish two possible explanations for history dependence. 

The first explanation is that people are not rational in the sense of Femina Economica and Homo Economicus. Instead, they behave in a way that Tversky and Kahneman call loss aversion. If you are buying a house from a person who bought it a long time ago and who has made a big profit, that person will be willing to sell it for less than someone who is selling a comparable house, but is taking a loss on the sale. Bracke and Tenreyo call that anchoring.

A second possible explanation for the same observation is that the person who suffered a loss has a large mortgage and needs a higher price to move on to her next purchase.  To separate those two explanations Bracke and Tenreyo separate their data into those who have large mortgages and those who own their homes outright. If the second explanation were correct, there should be no evidence of history dependence among those who are debt free. The data does not support that hypothesis. Bracke and Tenreyo find that sellers demand a higher price if they bought the house when prices were high; and the effect is uniform across those with mortgages and those without.  Tversky and Kahneman studied loss aversion in experiments on human subjects. Bracke and Tenreyo find evidence that supports the loss-aversion hypothesis in real world housing data. 


Now let’s turn to Arzu Uluc, who discusses her co-authored work with Rafa Baptista, Doyne Farmer, Marc Hinterschweiger, Katie Low and Daniel Tang. You can find a link to Arzu’s description of her research by clicking on the image here.


The 2008 financial crash was not just an economic crisis; it was also a crisis for the dominant research paradigm: neo-classical economic theory. The paper that Arzu presented is an example of one of the alternative approaches to economics that have flowered as a response to a perceived failure of the neo-classical approach. Standard neo-classical economic models assume that the world is populated by rational people with infinite powers of future perception. Arzu and her co-authors use instead an agent based modelling (ABM) approach.

Agent Based Modelling draws its inspiration from the success of computer modelling techniques in the natural sciences.  For the uninitiated, I recommend the Online Guide by Robert Axelrod and Leigh Tesfatsion at the University of Iowa. In an agent based model, human beings are modelled as automata that respond in predetermined ways to their environment. Agent based models are highly non-linear complex systems that can generate a wide range of endogenous propagation mechanisms. The hope is, that by designing and experimenting with new policies in these models, we may learn how to design policies that stabilize real world economies. 

Arzu et al build an agent based model with four types of agents, each of which behaves in ways that mimic the behaviours of their real-world counterparts and they find that their computer model displays a rich pattern of complex dynamics. Importantly, the simulated dynamics of the model respond to policies that limit the loan to income ratio. When the market restriction is in place, the amplitude of credit cycles is diminished.  This is an approach that I expect to see applied more widely in the coming years.


Before you go here is a preview of the issues I’ll tackle in my next few posts – have a listen to the clips from other speakers at the conference last July, and watch this space.


Do Low Interest Rates Punish Savers?

01 James Bullard.png
11 Kevin Lansing.png

This is the second of my posts on the conference: Applications of Behavioural Economics, and Multiple Equilibrium Models to Macroeconomic Policy, held at the Bank of England on July 3rd and 4th. I feature two papers written by officials from the Federal Reserve System. James Bullard, President of the Federal Reserve Bank of St. Louis, discusses the implications of his recent research for low interest rates. And Kevin Lansing, a Research Advisor at the Federal Reserve Bank of San Francisco, discusses his work on multiple equilibria. 


James Bullard: President of the Federal Reserve Bank of St. Louis

James Bullard: President of the Federal Reserve Bank of St. Louis

Jim's fascinating and timely presentation was based on a research paper that he co-authored with Costas Azariadis, Aarti Singh and Jacek Suda. In that paper, Jim and his co-authors construct a stylized model of an economy with long-lived people and they study the determinants of interest rates. In a video, linked here, Jim discusses the implications of his academic research for a hot political topic: the persistence of very low interest rates almost ten years after the end of the 2008 financial crisis. In his view, there is not much the Fed can do about low interest rates; monetary policy is simply allowing the price of borrowing, aka the interest rate, to equate supply and demand.  In his words, “you don’t want to encourage the suppliers too much or you’ll get oversupply: you don’t want to encourage the demanders too much or you’ll get a shortage, so the price brings these two into balance”.

96
 

 
Normal
0




false
false
false

EN-US
X-NONE
X-NONE

 
 
 
 
 
 
 
 
 


 
 
 
 
 
 
 
 
 
 
 


 <w:LatentStyles DefLockedState="false" DefUnhideWhenUsed="false"
DefSemiHidden="false" DefQFormat="false" DefPriority="99"
LatentStyleCount…

Figure 1: Source: Bauer and Rudebusch

Although he doesn’t say it explicitly in his talk, Jim is referring here to what economists call the ‘real interest rate’; that is, the money interest rate adjusted for expected inflation. The real interest rate is the price that borrowers must pay to trade future for present consumption goods and much recent research suggests that this price has fallen steadily over the past couple of decades as the Chart in Figure 1 shows. This chart is reproduced from a Federal Reserve Bank of San Francisco Letter, by Michael Bauer and Glenn Rudebusch which discusses possible reasons for the decline in what Jim Bullard calls the “neutral price”. This is the market real interest rate that equates the supply of savings by lenders with the demand for savings by borrowers. Critics of the Fed blame loose monetary policy and past Fed mistakes for the current period of low rates and although I have some sympathy for this argument, it is not easy to come up with convincing explanations for permanent effects of Fed policy on the real as opposed to the monetary interest rate. In my view: the jury is still out.


Jim takes the position in his talk that the neutral real interest rate is independent of central bank policy. But is it? A second theme that arose at the conference is that any given central bank policy can lead to more than one equilibrium outcome. That theme was taken up by Kevin Lansing, a Research Advisor at the Federal Reserve Bank of San Francisco. I discuss Kevin's work below. You can hear Kevin describe his paper, in his own words, here.


Kevin Lansing, Research Advisor at the Federal Reserve Bank of San Francisco

Kevin Lansing: Research Advisor at the Federal Reserve Bank of San Francisco

Kevin's pioneering research research explores the idea that there may be more than one inflation rate associated with any given monetary policy. In normal times, the interest rate is positive and the central bank raises or lowers it to cushion the effects on the economy of shocks to demand and supply. But if the central bank lowers the interest rate too far and too quickly, as it did in 2008, the economy may become stuck in a low inflation rate trap. One possible solution that Kevin discusses is to raise the central bank’s inflation target. This is an idea that has been floated by a number of authors; (see for example, the Vox piece by Laurence Ball of Johns Hopkins University). Kevin finds that although a higher inflation target may mitigate the problem, it doesn’t remove it entirely.

How should we reconcile these two conference contributions? 

Jim is sympathetic to the idea that there may be multiple equilibria and he has discussed the idea elsewhere that the economy may switch occasionally between two regimes. In one regime, the real interest rate is high: In the other regime it is low. He does not take a stand on what causes the switch from one regime to the other.

Kevin’s research addresses a related, but different question. Like Jim, Kevin does not take a stand on what determines the real interest rate. He asks, how does central bank policy influence inflation and he points out that any given central bank policy may be consistent with more than one inflation rate.

What is Keynesian Search Theory?

search image.jpg

I've been asked recently how my work on search theory differs from the approach pioneered by Peter Diamond, Dale Mortensen and Chris Pissarides (DMP). I call the DMP approach classical search theory, and in my book Prosperity for All, I distinguish classical search theory from a new approach; Keynesian Search Theory, that I introduced in my paper Aggregate Demand and Supply, published in 2008. I coined the term Keynesian search theory, in Prosperity for All, to clarify the distinction between my work and previous literature. If a macro model is closed with classical search theory, beliefs are determined by economic fundamentals. If a macro model is closed with Keynesian search theory, beliefs represent a new fundamental that is determined by the psychology of market participants. I call this new fundamental; the belief function.

Here's how I described this issue in a post on this blog in 2014

"Until recently, new-Keynesian economists didn't bother to model unemployment. Instead, they followed the new-classical approach in which all that matters is labor hours spent in paid employment. More recently, a number of authors including  Bob Hall, and Mark Gertler and Antonella Trigari have incorporated explicit models of search unemployment into otherwise standard macroeconomic DSGE models. ...  What is different about more recent work ...  [in classical search theory, that builds] ...on Hall's 2005 paper, is the way the model is closed. 

... When a firm meets a worker in a search model, the worker and the firm enter a bilateral bargaining situation. The worker would be willing to accept a job, and the firm would be willing to employ the worker, for any wage that is greater than the worker's reservation wage and less than the worker's marginal product. In his 2005 paper, Bob showed that one way to close the model, is to assume that the wage is fixed.

...That way of attacking the problem is, ... [in my view] ... a mistake. The new-Keynesians are squeezing the square peg of labor market search theory into the round hole of Samuelson's  neoclassical, synthesis. 

... An alternative approach, that has a better shot at understanding the data, ... [is to drop] ... the bargaining assumption completely and ...  [ to assume instead ] ... that firms and workers are price takers in the labor market..."

In classical search theory, firms and workers bargain over the wage, and the value of a firm depends on the outcome of the bargaining process. Expectations about the value of assets, aka beliefs, must ultimately be consistent with labor market fundamentals.

In Keynesian search theory, this process is turned on its head. Market psychology is a new independent fundamental that determines asset prices and the division of the product, between workers and firms, must ultimately be consistent with asset market beliefs. In this new way of seeing the world, market psychology is a new causal factor.

Here is how I presented the contrast between these approaches in Prosperity for All [Page 121]:

"[Keynesian search theory] explains permanent shocks to the unemployment rate as inefficient shifts from one unemployment equilibrium to another. If I am right, we can and should try to counteract permanent shocks to the unemployment rate by active intervention in which the central bank, acting as an agent for the treasury, buys or sells shares in the stock market to smooth out financial fluctuations."

As I argue in this video, Keynesian search theory offers the promise of a marriage between psychology and economics. Classical search theory does not. If you are young, ambitious and uncorrupted by tired establishment views, consider joining me in developing a truly interdisciplinary approach to the future of macroeconomics.