1. What Has Happened
in Macroeconomics
(and what still needs to be done)
Noah Smith
Bloomberg View
2. Where is macroeconomics used?
1. Academic macro
• Uncover the nature of the macroeconomy
• Advise policy
2. Central banks
• Inform policy
• Forecast the economy
3. Finance and consulting industries
• Forecast the economy
• Analyze policymakers’ behavior
3. Academic Macro
• Dominated by DSGE methodology
• “Microfounded” models with maximizing agents
• “Structural” parameters
• General equilibrium
• Typical but not necessary features
• Rational Expectations
• Infinite horizons
• Representative agents
• Utility of consumption and leisure
4. Central Bank Macro
• Some DSGEs
• SEMs
• FRB/US
• Linear equations
• Large # of parameters – joint estimation is essentially hopeless
• Structural assumptions but not “microfounded”
• Individual judgment, informal models
Why have central banks not jettisoned old approaches?
5. Industry Macro
• Non-structural models for forecasting
• Lots of judgment, informal modeling, and heuristics
Why don’t industry macroeconomists use DSGE models?
• Lack of talent?
• Lack of exposure?
• No need to forecast policy effects?
Probably none of these. Industry people just don’t believe existing
DSGE models accurately describe the dynamics of the economy.
6. Criticisms of Macro
Since the 2008 financial crisis, macroeconomics has come under
increasing public criticism.
• Why didn’t macroeconomists predict the crisis? Or at least see the
risks? Or at least warn about the possibility?
• Why didn’t macro models include finance?
• Why was the policy response to the Great Recession so confused and
contentious?
7. Criticisms of Academic Macro
Critics named “Paul”:
• Paul Romer, “The Trouble With Macroeconomics”
• Macro models follow fads set down by Lucas, Sargent and other thought
leaders
• Paul Pfleiderer
• Models that are ostensibly just thought experiments are actually used to
inform policy
• Paul Krugman
• Intentional suppression of Keynesian ideas
• Mistaking beauty for truth
8. Criticisms of Academic Macro
Critics not named “Paul”:
• Simon Wren-Lewis
• DSGE models too complex and opaque to be used quickly
• Ricardo Caballero
• Implausible microfoundations persist due to herding/tradition
• Brad DeLong
• Not enough attention to history
• Narayana Kocherlakota
• Implausible microfoundations persist due to herding/tradition
9. Caballero Critique
Caballero (2010):
“My point is that by some strange herding process the core of macroeconomics
seems to transform things that may have been useful modeling short-cuts into a
part of a new and artificial ‘reality,’ and now suddenly everyone uses the same
language, which in the next iteration gets confused with, and eventually replaces,
reality. Along the way, this process of make-believe substitution raises our
presumption of knowledge about the workings of a complex economy…
Incidentally, this process of selective measures of success also weakens the initial
motivation for building the microfoundations of macroeconomics, which is to make
the theory testable. A theory is no longer testable when rejection is used not to
discard the theory, but to select the data moments under which the core model is
to be judged. This practice means that well-known major failures just become
‘puzzles,’ which are soon presumed to be orthogonal to the output from the
quantitative model that is to be taken ‘seriously.’”
10. Kocherlakota Critique
Kocherlakota (2016), “On the Puzzling Prevalence of Puzzles”:
“To an outsider or newcomer, macroeconomics would seem like a field that is haunted by
its lack of data, especially good clean experimental data. In the absence of that data, it
would seem like we would be hard put to distinguish among a host of theories with distinct
policy recommendations. So, to the novice, it would seem like macroeconomists should be
plagued by underidentification or partial identification.
But, in fact, expert macroeconomists know that the field is actually plagued by failures to fit
the data – that is, by overidentification.
Why is the novice so wrong?...
The mistake that the novice made is to think that the macroeconomist would rely on data
alone to build up his/her theory or model. The expert knows how to build up theory from a
priori restrictions that are accepted by a large number of scholars. (Indeed, in the academe,
that’s exactly what it means to be an expert macroeconomist.) Those restrictions are what
give the models their empirical content. As it turns out, the resulting models actually end
up with too much content – hence, the seemingly never-ending parade of puzzles.”
11. A Couple of Macro Puzzles
• Forward guidance puzzle: Standard model implies far-future forward
guidance is infinitely powerful
• McKay, Nakamura and Steinsson (2015)
• Neo-Fisherian puzzle: With simple model tweaks, low interest rates
are disinflationary
• Schmidt-Grohe and Uribe (2016)
• Cochrane (2016)
• Williamson (2013)
12. Questions for Academic Macro
1. What new or different things should macroeconomists try with the
DSGE approach?
2. Should other approaches be tried?
3. How can empirical evidence help improve theory?
13. Basic Problem: Macro is Really Hard
• High causal density
• Requires highly stylized assumptions
• Measurement difficulty
• Aggregation issues
• Are macro variables even the right things to measure?
• Uninformative data
• Aggregate shocks make cross-sectional analysis impossible
• Small samples: time series approaches are slow to converge
• Natural experiments usually dubious (Romer dates)
14. DSGE: The Standard Model
Smets-Wouters (2007)
• Used by many central banks
• Basic RBC core: Euler equation, classic shocks, budget constraints
• Imperfect competition in intermediate goods
• Habit formation
• Capital adjustment costs, capacity utilization costs
• Sticky prices and wages
• Taylor rule
• Log-linearization
• TFP shock, labor supply shock, preference shock, government spending
shock, inflation target shock, investment shock, interest rate shock, wage
markup shock, price markup shock, equity premium shock
15. Efforts to Improve Standard Model
Common ideas for improving DSGE models:
1. Financial frictions (obvious!)
2. Heterogeneity of households and firms
3. Search and matching
4. Departures from Rational Expectations
• Ragnar Söderberg Foundation project:
http://ragnarsoderbergsstiftelse.se/macroeconomic-policy-analysis-21st-century
• Heterogeneity
• Search and matching
• Heterogeneous beliefs
16. Financial Frictions
• Brunnermeier, Eisenbach and Sannikov (2012): Surveys models with financial
frictions
• Del Negro, Giannoni and Schorfheide (2014): Adds Bernanke-Gertler-Gilchrist
“financial accelerator” mechanism to Smets-Wouters model
• Similar to Christiano, Motto and Rostagno (2012)
• Matches large output response in Great Recession
• Inflation not too off
• Finds financial shocks are most important shocks
17. Adding Labor Search
• Diamond-Mortensen-Pissarides model
• Has difficulty explaining unemployment, as shown by Shimer (2005)
• Christiano, Eichenbaum and Trabandt (2016)
• Replace sticky wages with labor search
• Find required amount of stickiness is much lower than other models
• Mitman and Rabinovich (2014)
• RBC-type DMP model with unemployment insurance policy
18. Heterogeneous Agent Models
• Pioneered by Krusell and Smith (1998)
Heterogeneous Agent New Keyensian (HANK) models:
• Kaplan, Moll and Violante (2017)
• Large effects of monetary policy
• Role for fiscal policy
• Rognlie and Auclert (2016)
• Show persistent demand shortfalls due to rise in inequality
• Relies on assumption of constant interest rates
19. Abandoning Rational Expectations
The most closely guarded assumption in macro, next to general
equilibrium itself, is probably Rational Expectations. So why not
abandon it?
Woodford (2013): “Macroeconomic Analysis Without the Rational
Expectations Hypothesis”
• Learning
• Restricted perception
• “Near-rational” expectations
20. Learning Models
• Sargent
• Marcet and Sargent (1989): least-squares learning
• Cogley and Sargent (2005): model uncertainty and monetary policy
• Large list of papers at http://www.tomsargent.com/learning.html
• Evans and Honkapohja (2001)
• Survey book about learning models
• Matthes and Rondina (2012)
• Learning by agents and central bank
21. Restricted Perceptions
• Coibion and Gorodnichenko (2012)
• “Information rigidities”: Sticky information and noisy information
• Evidence of persistence in inflation forecast errors
• Persistence is reduced after big visible shocks (e.g. 9/11 attacks)
• Survey measures of expectations
• Coibion and Gorodnichenko (2015): inflation expectations explain “missing
disinflation” in Great Recession
• Coibion, Gorodnichenko and Kumar (2015)
22. Near-Rational Expectations
• Farhi and Werning (2017)
• Uses level-k thinking, heterogeneity, and incomplete markets
• Weakens effect of monetary policy, resolves forward-guidance puzzle
• Kozlowski, Veldkamp, and Venkateswaran (2015)
• Agents use recent macro data to estimate their model of the economy
• Generates persistent business cycles
• Garcia Schmidt and Woodford (2015)
• Uses imperfect foresight to rule out Neo-Fisherian effects
• Kueng (2015)
• Finds excess sensitivity to large predetermined payments
23. Bigger Departures from Rational
Expectations
• Gabaix (2017)
• Inserts short-horizon optimization directly into the consumption function
• Predicts long-run Neo-Fisherian effects
• Large role for fiscal policy
• Hirshleifer, Li and Yu (2015)
• Introduces extrapolative bias into expectations
• Asset pricing model, but includes production economy
24. Agent-Based Models
• Agent-based modeling = large-N simulation of agents with behavioral
rules
• Under certain rules and with certain regularity assumptions, can be
equivalent to DSGE
• Many DSGE models already use simple “agent-based” approach
• No well-known results so far
• Similar to sunspot models – potentially true, but useless?
• Same data problems as DSGE
25. What Makes a Good Macro Theory Good?
But let’s step back a moment and ask: What is the goal of these and other
improvements? How is success defined?
1. Forecasting power?
1. In-sample fit?
1. Out of sample fit?
1. Plausibility?
1. Publishing papers?
26. Forecasting Power
Despite improvements, even the best DSGE models have limited forecasting
ability.
• Gurkaynak, Kisacikoglu, and Rossi (2013): Show that DSGE model forecasts
often underperform univariate AR models
• Carriero, Galvao and Kapetanios (2016): Compare large number of macro
forecasting methods, show that DSGE models have average performance
If a model can’t forecast, can it be a well-specified structural model? Yes, if:
• Little persistence in the economy
• Policy “thermostat”
27. In-Sample Fit
• Danger of overfitting! Graph from Chari, Kehoe and McGrattan (2008)
showing inflation predictions of Smets-Wouters (2007) model:
28. Out-of-Sample Fit
• In most sciences, models gain acceptance only after performing well
out-of-sample.
• This requires that parameters not be re-estimated or re-calibrated!
• In macro, there are few different “domains” to test models out-of-
sample.
• Other countries
• Wait a couple of decades
• There is a validation/estimation tradeoff – information used to
estimate model parameters can’t be used to test the models
29. Out-of-Sample Fit and Falsification
• In most sciences, models with poor out-of-sample fit are labeled
“wrong”.
• In macro, “all models are wrong.”
What is the penalty for a macro model with poor out-of-sample fit?
• Modification
• Gradual loss of interest
• Nobel Prize
30. Pessimistic Sociology
A pessimistic sociologist might view the macro field as a self-perpetuating
system for extracting rents (mostly, status/respect).
• Undergrad econ majors (future businesspeople) pay to learn macro.
• Central banks, IMF, etc. want academic pedigree
• Media is interested in macro and seeks “expert” input
• Grad students who leave macro have good computational skills
• Macro research = signaling for entry into a cushy exclusive club?
(No, I don’t actually believe this. Most macroeconomists I know just want to
figure out how the world works! But it’s good to avoid this perception.)
31. Idea: Back to Microfoundations
It’s possible that attempts at “microfoundations” didn’t go far enough, or
were never serious in the first place.
• No demands that models fit micro data
• One “structural” parameter, 100 estimates
• Compare Christiano papers on Great Recession!
• A representative agent is not “micro”
• Basic assumptions are preserved a priori, at the cost of hideous complexity
• Standard “microfoundations” bear little resemblance to what modern
microeconomists do
Idea: Get the pieces right, then build up!
32. Example: Euler Equation
Most DSGE models are built around some version of the following
simple equation:
This equation has two observables: consumption, and the real interest
rates.
With assumptions about the value of 𝛿 and the functional form of u(c),
we can test this equation directly.
33. Testing the Euler Equation
Canzoneri, Cumby and Diba (2006) test several Euler equations against
observed interest rates.
The Euler equation doesn’t just fail to match observed interest rates,
but often the correlation is negative!
34. A Few Assumptions in a Typical Macro Model
A partial selection of the assumptions from Christiano, Eichenbaum and
Evans (2005):
• Production consists of many intermediate goods, produced by
monopolists, and one single consumption good" that is a CES combination
of all the intermediate goods.
• Firms who produce the consumption good make no profits.
• Firms rent their capital in a perfectly competitive market.
• Firms hire labor in a perfectly competitive market.
• New firms cannot enter into, or exit from, markets.
• All capital is owned by households, and firms act to maximize profits (no
agency problems).
35. Assumptions in a Typical Macro Model
(Cont.)
• Firms can only change their prices at random times. These times are all
independent of each other, and independent of anything about the firm, and
independent of anything in the wider economy. (This is "Calvo pricing". The magic
entity that allows some firms to change their prices is called the "Calvo Fairy").
• The wage demanded by households is also subject to Calvo pricing (i.e. it can only
be changed at random times).
• Households purchase financial securities whose payoffs depend on whether the
household is able to reoptimize its wage decision or not. Because they purchase
these odd financial assets, all households have the same amount of of
consumption and asset holdings.
• Households derive utility from the change in their consumption, not from its level
("habit formation"). Households also don't like to work.
• Households are rational, forward-looking, and utility-maximizing.
Why should anyone think a model with these assumptions is “plausible”?
36. Why Accept Bad Assumptions?
Milton Friedman (1953):
“Consider the problem of predicting the shots made by an expert billiard player. It seems not at all
unreasonable that excellent predictions would be yielded by the hypothesis that the billiard player
made his shots as if he knew the complicated mathematical formulas that would give the optimum
directions of travel…Our confidence in this hypothesis is not based on the belief that billiard players,
even expert ones, can or do go through the process described; it derives rather from the belief that,
unless in some way or other they were capable of reaching essentially the same result, they would not
in fact be expert billiard players.
It is only a short step from these examples to the economic hypothesis that under a wide range of
circumstances individual firm behave as if they were seeking rationally to maximize their expected
returns…and had full knowledge of the data needed…Now, of course, businessmen do not actually
and literally solve the system of simultaneous equations in terms of which the mathematical
economist finds it convenient to express this hypothesis, any more than leaves or billiard players
explicitly go through complicated mathematical calculations or falling bodies decide to create a
vacuum.”
This is often interpreted to mean that it doesn’t matter if the pieces of macro models match micro
data, as long as the models match macro data.
37. Problems With the Pool Player Analogy
• Problem 1: Lucas Critique – If model elements don’t fit micro data, how can
we assume parameters are structural?
• Problem 2: Throwing away micro data throws away most of the data!
• Problem 3: Seems implausible that a machine built from broken parts could
work.
• Problem 4: Without micro data as an anchor, fads and tradition determine
model elements. That requires increasing model complexity.
• Hard to estimate
• “Degenerating research program”, loss of interest
38. Micro-Focused Macro
More macroeconomists are doing research into the “pieces” of macro
models. A few examples include:
• Bils, Klenow and Malin (2016) disaggregate wage data to study stickiness of
prices and wages
• Wong (2016) measures differences in consumption patterns by age
• Forsythe (2016) measures differences in cyclical employment patterns by
age
• Stroebel and Vavra (2014) measure response of consumption to house
prices
This work is mostly being done by young people. It’s humble in scope and
sticks close to the data. Old people should encourage this work!
39. Slightly-less-macro Macroeconomics
Interesting Question: What is “macroeconomics”?
Does it have to include fully specified models of the business cycle?
How about focusing on more sophisticated, richer, more empirically
grounded theories of consumption, investment, and other “pieces” of the
macro puzzle?
• Labor search theory as an example – how about consumption, investment, etc.?
• May require simulation to solve → agent-based modeling!
Why not leave the dream of a credible fully specified business cycle model
for the future?
40. The Micro Approach Answers Most Critics
• Caballero
• Kocherlakota
• Romer
• Pfleiderer
• Krugman (somewhat)
As for simple models, history, and judgment, this is more about central
bank and IMF culture than academia.
41. Ultimate Goal: Macro Models That Work
1. Use clever empirics and simple theory to understand micro
behavior
2. Fewer pointless theory sections in papers
3. Build models from reliable microfoundations
4. Test models out-of-sample to validate
5. Insist on consistency of parameter values across papers
6. Go slow, allow central bankers to use judgment and simple models
in the meantime