1. Doing (and Teaching) Economics for the Real World Dani Rodrik IEA, June 2017

2. What is the Nobel committee thinking?
What’s the big idea?
Can they both possibly be right?

3. In a nutshell…
“Economics is a science of thinking in terms of models joined to the art of choosing models which are relevant…”
Keynes (1938)
“There’s no easy line in summarizing my contribution,” he protested. “It is industry-specific. The way you regulate payment cards has nothing to do with the way
that you regulate intellectual property or railroads. There are lots of idiosyncratic factors. That’s what makes it all so interesting. It’s very rich It’s not a one-line thing.”

4. In the land of the Econ …where “modls” rule
Axel Leijonhufvud, “Life Among the Econ,” Western Economic Journal, September 1973.

5. … and determine attitudes towards other social sciences

6. … though even the natives cannot agree which “modl” to use

7. But why models: Borges on scientific method

8. Argument Models are key to the “scientific” nature of economics
understand complex social reality by laying bare a very large variety of causal relationships, one at a time
Economists have the wrong take on their discipline
succession of ever-better models, till we get the right model?
Economics advances not by settling on “the model,” but by generating useful collection of models
an inventory of partial explanations
non-universality and context-specificity
This view of economics counters typical critiques of economics
as well as economists’ own description of their practice
Economists are good at making models, but poor at navigating among them

9. Models at work: what does economics have to say on the effects of
minimum wages on employment?
competition on innovation?
expansionary fiscal policy on economic activity?
capital inflows on economic growth?
trade liberalization on economic performance? ….
Different models, different results
(and empirical analysis rarely ever conclusive, often pointing to different outcomes for different places and times)

10. Example: minimum wages
What are the employment consequences of a minimum wage imposed by the government?

11. Effects of minimum wage under two different kinds of market structureL
A competitive market

12. Effects of minimum wage under two different kinds of market structureL
A competitive market

13. Effects of minimum wage under two different kinds of market structureMC w w S S D D L L
A competitive market
A monopsonistic market

14. Effects of minimum wage under two different kinds of market structureMC w S S D D L L
A competitive market
A monopsonistic market

15. Can empirics do away with models?
“Let facts on the ground determine policy”
“We don’t need theory if we have good evidence”
RCTs, especially
Big Data
no need for theorizing any more; simply uncover the patterns in the data
Counterarguments:
every “fact” requires an interpretive frame
models provide the interpretive frame
example: toin cosses
regularities versus causation
without causal stories we don’t know when a regularity stops being one
RCTs require causal stories/extrapolation to be applicable, and we need models for that

16. Economic models as … fables
are simple dfgdfgdfgdfgdfgdfgd
dfgdfgdfgdfgdfgdfgd
are not real dfgdfgdfgdfgdfgdfgd
have clear storyline dfgdfgdfgdfgdfgdfgd
have characters that can be animals or objects dfgdfgdfgdfgdfgdfgd
typically have a moral
provide interpretive short-cuts
are multiple, one for every situation
Economic models
simplicity: ceteris paribus assumption
reality: stylized abstractions, untrue assumptions
storyline: clear cause-and-effect, if-then relationships
characters: random shocks, exogenous structural parameters, “nature”…
moral: policy implication
interpretation: analytic shortcut
multiplicity: context-specificity
(cf. Rubinstein 2006; Cartwright 2008)

17. Economic models as … experiments
isolate effects of specific cause/intervention
can be replicated by anyone
experiments produce dissimilar results in diverse settings
lab experiments do not pretend to represent “real world”
field experiments need to be extrapolated to other settings
in both cases, external validity not assured and has to be argued and supplied from outside
Economic models
clarify causal links by simplifying gdfgdfgdfg
can be reproduced by anyone
different models for different contexts
do not claim to be representations of real world
relevance of a model depends on “extrapolation”
additional techniques needed to sort out the usefulness/relevance of available models (see below)

18. What makes models “scientific”?
Clarifying nature of hypotheses: explicit causal chains
simplification both a necessity and a virtue: isolation => what precisely does an explanation depend on?
“a model is an experiment, and vice versa” (Mäki; Gilboa et al.)
Model selection
after the fact
in real time
A method for sorting out disagreements
“arguments that can be shown to be wrong” vs. those that are “not even wrong” (W. Pauli)
we can agree on what we disagree on, even when empirical evidence is too weak to discriminate among models
Accumulation of knowledge
how economics advances (slide)
Nature of “authority”
rests on “quality” of models (judged by principles widely shared by practitioners), not on reputation/status/network

19. What makes models “scientific”?
Clarifying nature of hypotheses: explicit causal chains
simplification both a necessity and a virtue: isolation => what precisely does an explanation depend on?
“a model is an experiment, and vice versa” (Mäki; Gilboa et al.)
Model selection
after the fact
statistical tools
in real time
applied policy analysis (e.g., growth diagnostics)
Nature of “authority”
rests on “quality” of models (judged by principles widely shared by practitioners), not on reputation/status/network

20. What makes models “scientific”?
Clarifying nature of hypotheses: explicit causal chains
simplification both a necessity and a virtue: isolation => what precisely does an explanation depend on?
“a model is an experiment, and vice versa” (Mäki; Gilboa et al.)
Model selection
after the fact
statistical tools
in real time
applied policy analysis (e.g., growth diagnostics)
A method for sorting out disagreements
“arguments that can be shown to be wrong” vs. those that are “not even wrong” (W. Pauli)
we can agree on what we disagree on, even when empirical evidence is too weak to discriminate among models
Nature of “authority”
rests on “quality” of models (judged by principles widely shared by practitioners), not on reputation/status/network

21. What makes models “scientific”?
Clarifying nature of hypotheses: explicit causal chains
simplification both a necessity and a virtue: isolation => what precisely does an explanation depend on?
“a model is an experiment, and vice versa” (Mäki; Gilboa et al.)
Model selection
after the fact
statistical tools
in real time
applied policy analysis (e.g., growth diagnostics)
A method for sorting out disagreements
“arguments that can be shown to be wrong” vs. those that are “not even wrong” (W. Pauli)
we can agree on what we disagree on, even when empirical evidence is too weak to discriminate among models
Accumulation of knowledge
not by one model replacing another, but by richer set of models
and better understanding of where they apply
Nature of “authority”
rests on “quality” of models (judged by principles widely shared by practitioners), not on reputation/status/network

22. Math and economic models
Models do not require math, in principle
any causal statement contains an implicit model
In practice, math often useful to
clarify (and make explicit) the nature of assumptions, relationships, conclusions
ensure conclusions follow logically from assumptions
“economists use math not because they are smart, but because they recognize they are not smart enough”

23. Maintained assumptions in economics
Are some assumptions (modeling conventions) so important that they always have to be preferred?
individual rationality
less and less a benchmark
“micro-foundations”
methodological individualism a strength of economics, but can come at cost of introducing other, misleading simplifications
e.g., representative agent, infinite horizon, rational expectations, …

24. How to figure out the relevant model – the craft of economics
Verify direct implications
Verify incidental implications (comparative statics)
Verify critical assumptions (cf. M. Friedman)
Verify mechanisms
Note parallels with external validity in randomized controlled trials
Theoretical models face an “external validity” problem just as lab experiments or RCTs do (which, by the way, are viewed as the gold standard in evidence).

25. Example: growth diagnostics as model selection (I)
Why is growth low:
Neoclassical model: physical and human capital
“Endogenous” growth model: R&D and competition
Trade & growth model: trade policy, exports
Dual economy model: structural change/industrialization
“Chicago” model: too much government
Structuralist model: too little government
Institutionalist model: property right & contract enforcement …

26. Example: growth diagnostics as model selection (II)
Explicit search for “diagnostic signals”:
“If story A is correct, signals x, y, z must be present…”
Direct evidence
“shadow” prices: returns to education, real interest rates, cost of transport,…
benchmarking potentially helpful
Indirect evidence
if a constraint binds, effects must show up in differential outcomes for activities that differ in their intensiveness in that constraint
informality, internalization of finance, self-enforcement of contracts
elimination of other plausible constraints
A constant cannot explain a change
high growth episodes of the past cause us to ask what has changed

27. What does all this imply for teaching economics
Cherish diversity of models
Emphasize plurality of critical assumptions, behavioral relations, rather than benchmark, competitive model
teach multiple models side by side
teach (or at least talk about) model selection in practice
Emphasize logics beyond competitive market eq’m
second-best thinking
coordination failures
small-numbers interactions: game theory models
Emphasize economics as a way of thinking, rather than a defense of/argument for free markets and efficiency

28. Re-evaluating critiques of economics
Simplistic/reductionist theories: a feature, not a bug
Inappropriate universalistic claims: a problem
Reification of markets and material incentives: perhaps
“Conservative bias”: not clear
Failure to predict: no; conditional predictions at best
Methodological biases, that crowd out new ideas: perhaps, but a feature of all scientific disciplines
Loss of ambition
from a program to transform society to merely understanding how a particular form of market society works
economists as dentists or plumbers

29. Real failings originate from behavioral and sociological aspects of profession
Mistaking a model for the model
expecting the same model works all the time
overlooking alternative models with different implications
over-confidence, hubris
Categorical preference for certain axioms
assumption of rational, forward-looking individuals operating in perfectly competitive markets
Preference for questions that are amenable to available tools of analysis
substantive implications of common tractability assumptions
neglect of issues involving scale economies until analytical tools were developed
Implicit political-economy theorizing in policy discussions
economists’ training endows them with no way to evaluate alternative social states other than through lens of allocative efficiency

30. Final word Recognition of economics as portfolio of models:
forces economists to be more humble about how much they really know
enables greater understanding of the variety of social phenomena
where such understanding is possible
closes some of the gap with other traditions in social sciences (critical, cultural, humanist, constructivist, interpretive)
an economist’s answer to “what about x which you left out of your model…?” is/should be “OK, let’s write down a model of it…”

31. Want to read more?