Alternate History? ('History 2.0')

From 1998 We Had 2030 Tools:

2030 Data: Panel data at the household level:

• Integrated HH balance sheets
  • Registry, or Mint.com-type

• With
  • Expectations ('What will [asset] prices do?')
  • Explanations ('Why'?)
  • Hypotheticals ('What would you do if?')
  • Uncertainty, Perceived Constraints ...
    • Whatever models say matters

2030 Models: Expectations Endogenized

Many competing theories right now

• Fading Memory
• Diagnostic Expectations
• k-Level Thinking

Winner Will Have Two Features

1. Expectations have an important extrapolative component
   • Perception of $\Delta p_{t} \Rightarrow \mathbb{E}_{t}[\Delta p_{t+1}]$

Older Similar Theories:

Winner Will Have Two Features

1. Heterogeneity in the interval used for extrapolation
   • Shiller looks back 150 years
   • Some don't remember events from 10 years ago
     • Low-memory types: Young people, e.g.

... Toy 2030 Theory: Anatomy Of Bubbles

1. Some genuine good news arrives
   • Causing initial $\Delta p_{t+1} > \mathbb{E}_{t}[\Delta p_{t+1}]$

1. Extrapolative $\mathbb{E}[\Delta p]$ spreads to "susceptible" like disease
   • Dies out if not enough are "susceptible"
     • People with access to funds ...
     • ... and short 'memory'
     • ... who are optimists
   • Infection $\approx$ amount marginal extra demand

1. Bubble collapse: like "recovery" from infection

Prehistory

Mid-90's Productivity Acceleration

• Greenspan's "New Economy" (~1996)
• Becomes Consensus:
  • Economic Report of the President (2001)

$\Rightarrow$ Boom in asset prices

• Stock Market (1994-2000)
• Also big $\uparrow$ in house prices (~1997-2002)

Citations:

The Boom: History 1.0 - Facts

2000-01: Contra , housing was not the cycle

• Barely a blip in house prices or construction

2001-06: "Global Savings Glut ('GSG')"

• $\Rightarrow$ Interest rates lowest since Great Depression

• Continuing $\uparrow$ in availability of credit across the board

  • Subprime and prime

  • Primary and nonprimary

Other Citations

The Boom: History 1.0 - Contemporaneous Views

• First prominent claims housing bubble is inflating?
• "That Hissing Sound"
• 'Financial innovation'
  • 'Something funny going on around here'
• Popular culture
  • "Flip That House" first episode: July 2005

The Boom: History 1.0 - Contemporaneous Views

"It is Not A Bubble"

• 2005 JEP Paper

Instead, It's Improved Fundamentals

• "New Economy", Low $R$, 'financial innovation'

Citations

The Boom: History 2.0 (2002-2006)

• 2030 Theory: Improved fundamentals are prereq for a bubble
  • Not an argument against a bubble
  • Bubble? Depends on "infectiousness" of $\Delta p$
    • How new marginal demand
    • Buy because of $\Delta p_{t} \Rightarrow \mathbb{E}_{i,t+1}[\Delta p_{t+1}]$

The Boom: History 2.0 (2002-2006)

• 2030 Data: Many in "susceptible" (=marginal) pool in 2002-03
  1. Beneficiaries of 'financial innovation'
     • New people who can participate
       • (Mian and Sufi, Gramlich)
     • Others who can borrow more
       • (Adelino et al)
  2. "New Economy" beliefs + Bush tax cuts ...
     • People with money to invest ...
     • ... pessimistic about stock market (after dot-com bust)
     • ... but optimistic about house prices (extrapolating)

History 2.0 (2003/4-2006): Infection Spreading

• 2030 Data: We see expectations and options changing heterogenously
  • Among marginal buyers, we see $\uparrow$ in $\mathbb{E}_{i,t}[\Delta p_{t+1}]$
    • We ask them why $\uparrow$ in $\mathbb{E}_{i,t}[\Delta p_{t+1}]$
      • They say, basically, "momentum"
    • and the marginal buyers say they are buying:
      • because $\mathbb{E}_{i,t}[\Delta p^{h}_{t+\bullet}]$
      • Many are "real estate investors"
  • Among nonbuyers we at the same time see decline in expectations
    • Dispersion of beliefs may be predictive

History 2.0 (2003/4-2006): Infection Spreading

Before 2030 Data: "You know it's a bubble" when

• 1920s: Shoeshine boys give stock tips (Joe Kennedy)
• 2000s: 'Flip That House': Hairdressers, bartenders become flippers

2018 Data Soo: News index gets us partway

2030 Data are the plural of anecdotes.

Boom to Bust (2007-2008): History 1.0

Competing Interpretations of Great Recession

1. Huge negative shock to credit supply
   • Eggertson and Krugman, Guerrieri and Lorenzoni, Mian and Sufi
2. Huge increase in uncertainty
   • Bloom, many others

Boom to Bust: History 2.0 (2007-08)

• In 2006-07, expansion of credit stops
  • Low memory marginal types extrapolate quickly:
    • Low memory former optimists become pessimists
    • $\Delta p^{h}_{2007} < \mathbb{E}_{2006}[\Delta p^{h}_{2007}]$
      • $\Rightarrow \mathbb{E}_{2007}[\Delta p^{h}_{2008}] < \mathbb{E}_{2006}[\Delta p^{h}_{2008}]$
  • Hissing sound gets loud
  • Explains slowdown 2006-2008q2
    • Guess: Does not explain collapse between 2008q2-2008q4

Boom to Bust: History 2.0 (2008q2-2008q4)

2030 Data

• Consumption collapsed even:
  • For people who are never going to want to borrow
  • More for people whose expectations deteriorated more
  • In regions where there had not been a boom

2030 Theory

• Degree of uncertainty is a "fundamental"
  • We see huge increase in uncertainty
  • Those whose uncertainty increased more, cut $C$ more

The Boom: History 2.0 (2002-2006, Macropru edition)

• Macropru regulators know how to do micro 'stress tests' <!-- * Prudence: $\mathbb{E}[u^{\prime}]$
  • At date $t$, see dist'n of balance sheets and $\mathbb{E}_{t}[\Delta p^{h}_{t+\bullet}]$
    • Can see unusual participation by marginal types -->
  • 2030 Theory: We know what circs cause defaults
    • Different for investors vs primary owners
  • Can simulate defaults under alternate future histories:
    • Productivity growth, interest rates, uncertainty
      • ,
• By 2006, micro stress tests reveal major fragility to minor shocks
  • Vulnerability to massive departure of marginal demand
    • "Investors" as well as primary

The Boom: History 2.0 (2002-2006, Macropru edition)

Which Macropru policies do what?

• 2030 Theory+Data: First-and-only mortgage: default when
  • Underwater + Negative income shock ('double trigger')

• 2030 Theory
  • : Theory that matches 'double trigger' facts implies
    • Debt-To-Value rules insulate against income shocks
    • Payment-To-Income rules insulate against $p^{h}$ shocks
    • With ARMs, Debt-To-Income insures $r$ shocks (?)
  • Calibrated stress tests tell you which to adjust

The Boom: History 2.0 (2002-2006, Macropru edition)

New kinds of macropru rules

• Countercylical rules that target "speculative" demand. Examples:
  • Risk-weighted capital rules where "riskiness" rises with
    • Proportion of aggregate lending for non-primary-residence
    • Proportion of buyers who say they are buying because $\mathbb{E}[\Delta p^{h}]$ high

The Boom: History 2.0 (2002-2006, Macropru edition)

Consequence? History 2.0 differs from 1.0

• Size of bubble is smaller
• For a given bubble size, consequences are milder

Part 2: Agenda

Modeling (near term priorities)

Incorporate real estate investing in HH problem

• On top of "primary housing" choices
  • Difference: Less "utility cost" of defaulting
• Governed by same expectations, explanations
• In model eqbm, consumers face choice between:
  • Stock market
  • Real estate

New focus of models (and analysis thereof):

• Tell us what to do on surveys
  • Expectations and Explanations of what?
  • Whose expectations and explanations?

Examples

• Kaplan, Mitman, Violante (2018):
  • Expectations are central explanation of Great Recession
  • Their model: Changed beliefs about "fundamentals"
    • $\Rightarrow$ ask people beliefs about fundamentals
      • 'I think $R$ is permanently lower'? or
      • 'Prices have been going up (resp. down)'
• Garriga and Hedlund (2018):
  • Liquidity dries up when housing market tanks
  • $\Rightarrow$ first-order precautionary effects
  • Who to target?
    • Marginal sellers in good and bad markets
    • Ask them how their behavior is affected by liquidity

Dynamics (not just steady states)

• Especially for the marginal players
• Sluggishness in $p^{h}$ and behavior comes from:
  • Search frictions
  • Information frictions
    • Everybody knows everything instantly: won't work
• Very hard

Modeling Expectations and Explanations (Desiderata)

• Ideally, same deep model for Everybody
  • Difference in deep parameter like "memory"
  • Most diffs in behavior explained by circs
• Behavioral foundations strongly disciplined by evidence
  • "Other people are like me"
  • "Representativeness Heuristic"
  • Fading Memory
  • ...

Deep Improvements in Modeling Practice

Need a DYNARE for HA modeling

• ECB can run Riksbank model calibrated to Italy
• Riksbank can run Fed model calibrated to Sweden
• Pontus can run Pavel's model; etc

Feasible with modern collaborative software development tools:

• Modular
• Open-source
• Platform-Independent
• Automatic testing/debugging tools
• Robust reproducibility

Getting There?

• Institutional support of infrastructure development
  • Like DYNARE has had
  • As is done in other scientific/technical fields
    • Astronomy, Artificial Intelligence, Bayesian Statistics, Biology, ...
• Changes in professional equilibria
  • "Publication"
    • Referees need to be able to run your code
    • Readers need to be able to reproduce your results
    • Otherwise, it's not really "public"
• Beginning: Econ-ARK project

Abolish Consumer Expenditure Surveys

Replace them with Consumer Expectations Surveys

• Get expenditures from admin data (Mint.com, registries)
• Use precious survey time asking:
  • expectations
  • explanations: 'Did you buy that second house because $\uparrow$ in $\mathbb{E}_{i}[\Delta p^{h}]$?'
  • whatever else models say is important
• Oversample potential marginal decisionmakers
  • e.g. intensive focus on new homebuyers

"2030 Data" and "2030 Theory" By 2030?

• Depends, In Large Part, On You!
  • People in this room and people you know
• Great Recession cost trillions of dollars
  • "2030 Data" and "2030 Theory" might have prevented it
• Getting to "2030 Theory" and "2030 Data" - Cost
  • A few millions
• Central Banks, IMF, Statistical agencies:
  • Seems like a no-brainer
• Please make me a retrospective prophet!

Citations

References