The guts of a GUT:
Elements of a Grand Unified
Theory of Growth
Lant Pritchett
LACEA
November 12th, 2010
Outline of the presentation
• What is growth theory a theory of? The
four facts a growth theory should explain
• Growth phases and phase transitions
versus a single linear equation of motion
• “Institutions”: general or specific?
• Equations of motion for “institutions” a la
Hirschman: “unbalanced growth” through
“backward linkages” in institutions
Four Facts about Growth
• Small group of countries with sustained, nonaccelerating, stable growth of 1.8-2.0 ppa
producing very high levels of output.
• Small group of countries very near subsistence
(hence long-run growth near zero)
• Small group of countries with very rapid growth
over extended periods
• Growth rates lack persistence over time—very
low correlation of growth from one period to the
next—growth is (mostly) an episodic condition
(not a characteristic)
First Fact: Long-run stability in
growth among now leaders
• The growth rate 1870 to 2003 of the 16 leading countries is
1.89 ppa with std dev across countries of only .33
• Predicting US GDP per capita in 2003 using only data from
1870 through 1907 and the simplest possible linear trend in
natural logs produces a forecast off by 2 percent ( 29,037
actual versus 28,242 predicted GK 1990 dollars)
• Median forecast error for 70 year ahead prediction of all
leading countries from pre-depression data is 3.9 percent!
• The median acceleration of these 16 countries from 18901915 versus 1980-2003 is only .14 ppa
• High levels of per capita output produced by moderate,
sustained, stable, non-accelerating growth.
Long-run stability: example of US
(Cover of Jones’s book on growth)
Same figure, Denmark
Predict 2003 levels—≈100 years
ahead—almost perfectly
Data 1890-1901 to
estimate a trend
Fact II: Small number of countries very near
subsistence (hence zero long-run growth)
• Can infer growth from level if you are willing to
assume a minimum level of output—the “Adam
and Eve” level
• The maximum growth could have been over any
period is the growth that takes you from “Adam
and Eve” at the beginning to the current
observed level.
• Countries still near “Adam and Eve” levels
implies slow growth—often very slow growth.
• Conversely one can ask when the leaders were
at the currently observed levels
Poorest countries in Maddison data (GK
1990 $) have cumulatively very slow growth
Country
GDP per capita in 2003,
GK 1990 units
(Maddison 2007)
Maximum growth rate
since 1870
(minimum=450)
Zaire (DRC)
212
-0.57%
Burundi
477
0.04%
Central African
Republic
511
0.10%
Niger
518
0.11%
Sierra Leone
579
0.19%
Eritrea and Ethiopia
595
0.21%
Guinea
601
0.22%
Tanzania
610
0.23%
Afghanistan
668
0.30%
Zambia
689
0.32%
Haiti
740
0.37%
Mapping current income into modern
economic history (post 1820)
Mapping the poorest countries into
pre-modern history
Peace of Westphalia
1648
Fact III: A small number of countries have grown
27 (or more) years at very rapid rates—2 to 3
times higher than the historical pace of the leaders
Fastest 27 year episode in the recent (post 1950) data
JPN
East Asia
1950
7.54%
BWA
Africa
1963
7.29%
CHI
East Asia
1980
7.16%
TWN
East Asia
1961
7.05%
KOR
East Asia
1970
6.74%
HKG
East Asia
1960
6.29%
SGP
East Asia
1963
6.19%
THA
East Asia
1970
5.42%
MYS
East Asia
1958
5.02%
BRA
Latin America
1953
4.94%
IDN
East Asia
1965
4.71%
COG
Africa
1960
4.51%
VNM
East Asia
1980
4.48%
Extended rapid growth episodes
are concentrated in two regions
(East Asia and Europe)
Of the 21 fastest 27 year growth episodes in the PWT6.3
data:
10 are East Asia: Japan, China, Taiwan, Korea, Hong
Kong, Singapore, Thailand, Malaysia, Indonesia,
Vietnam
8 are European (ish): Romania, Greece, Spain, Portugal,
Ireland, Austria, Italy, Israel (European ARS)
Only exceptions: Botswana, Brazil, Congo (?! go figure)
Fact IV: Economic growth is (mostly) a
condition of countries, not a
characteristic
• Characteristics are relatively stable empirical
features—being left-handed, being tall (for people),
having coast-line, speaking Spanish (for countries).
• Conditions are relatively impermanent empirical
features—having a cold, being hungry (for people),
having just won the World Cup, having recently had an
earthquake (for countries).
• Characteristics have high persistence and high intertemporal correlations (the left handed are left handed),
conditions have low persistence and low inter-temporal
correlations (people with colds are not “the colds”).
R-Squared of growth on past
growth
Per capita growth is condition-like—R-squared of
current growth on past growth is .05 at 5 year
horizons and less than .13 even at 25 year
horizons–in contrast population growth is a
characteristic-like
0.800
0.700
0.600
0.500
growth, gdp per capita
0.400
growth, population
0.300
0.200
0.100
0.000
5
10
15
20
Horizon (n) of t+n on t-n
25
World distribution of N year growth rates
across countries
Proportion
Of periods in
Growth range
Low growth
Country
What a “growth as
Characteristic”
world would look like
Medium growth
Country
High growth
Country
World Distribution
0
2
4
Growth Rate
Growth as a condition—countries
change growth rates and span the
possible range of growth experiences
Medium growth country which
Spends more time in rapid growth
and in slow growth than the world
Distribution of episodes
World Distribution
Ghana—has more episodes of superrapid growth and of negative growth
than the world distribution
More time in
negative growth
More time in
rapid growth
How leaders grow—centered in the
middle (e.g. Great Britain)
No negative
No rapid
What stars look like…Singapore with
lots of very rapid and no negative
(pretty characteristic-like growth)
What a consistently slow grower
looks like (Niger)
But countries with the exactly same growth rate
have very different distributions—Colombia, overall
1.9 ppa—concentrated
Chile—growth of 2.1 ppa—but with
more bust and more boom—
roughly the world’s experience
Brazil—overall 2.6 ppa—more boom,
not so much bust, more stagnation
Congo—grew “faster” (2.4 ppa) than
Colombia or Chile—but most time was either
boom or bust
What “the same” growth looks like
Encompassing theory to explain all
four facts
• Set of leading countries with steady growth
around 2 ppa as an “absorbing” state
• Set of lagging countries with growth near zero
(for a very long time)—(some consistently,
others booms followed by busts)
• Set of countries with rapid growth (at least twice
historical pace of leaders) for extended periods
• Lots of countries doing all of the above (negative
and zero and moderate and rapid growth) back
and forth in episodic fashion (e.g. discrete
looking starts and stops)—averaged out to near
non-converging growth on the leaders
Big problem with a “unified” theory
of growth
• “Institutions” are an
important, causal, driver
of levels of income
(AJRobinson, Hall and
Jones)—which we can
identify because
institutions have long
persistence
• “Institutions Rule” in that
they are claimed to drive
out “policy” (e.g. Rodrik,
Subramanian, and Trebbi,
Easterly and Levine,
AJRobinsonT)
• “Growth has low
persistence” (Easterly et
al.)
• “Growth is episodic” with
discrete starts and stops
(Ben David and Papell,
Pritchett)
• Accelerations and
decelerations of growth
are common (Hausmann,
et. al.) are common, even
among poor countries
(Jones and Olken)
Measured rankings of aggregated
components of quality of
“institutions” tend to be very stable
Bureaucratic
Quality
Corruption
Law and Order
0.80
0.71
0.77
198596
0.82
0.70
0.64
199709
0.78
0.70
0.81
198509
0.62
0.58
0.58
0.60
R-2 of current on past
Indicator from ICRG
Five year
periods
0.50
0.40
0.30
0.20
0.10
0.00
Democratic
Accountability
0.72
0.65
0.70
0.51
Horizon (years)
Socioeconomic Risk
0.67
0.47
0.73
0.63
"Institutions" Grow th
5
12
Damn. Its both.
What a single growth model (esp.
single linear equation of motion)
cannot do
• Most “determinants” of growth are characteristics (e.g. having a
coast, good “institutions”) have very high persistence, but growth
has low persistence
• Related, most growth equations cannot predict the onset of
episodes of either growth accelerations or growth decelerations
(Hausmann, Pritchett, Rodrik)
• The magnitudes of growth dynamics are all out of whack with typical
“micro” estimates—much larger total level “impacts” than would be
predicted.
• Parameter instability is a specification test—and regressions fail
• Standard growth models are getting worse as more and more is
“TFP”—the residual
• Cannot distinguish between covariates that have impact “within
state” versus variables associated with higher/lower growth state
transitions—e.g. do “institutions” play a role within states or with
transitions across states?
The problem
We now have a growth empirics (and some
accompanying theories) that explains
everything except precisely what we
wanted a theory and empirics for—to tell
us how to accelerate growth rates
In waterfalls, water…well, it falls
…except when it doesn’t: water has
had a phase transition to ice
Simple “states and transitions” simulations with
chosen transition probabilities and constant within
state growth dynamics can mimic all growth facts
Collapse
Stagnation Moderate
Rapid
gc(..)
gS(..)
gM(..)
gR(..)
πCC(..)
πCS(..)
πCM(..)
πCR(..)
πSC(..)
πSS(..)
πSM(..)
πSR(..)
πMC(..)
πMS(..)
πMM(..)
πMR(..)
πRC(..)
πRS(..)
πRM(..)
πRR(..)
Notation: πRC(..)—probability of transition from Rapid to Collapse,
gi(..)—within state growth dynamics
Two needs for a GUT of Growth
• A “states and transitions” model that can
explain phase transitions across growth
states (e.g. from stagnation to boom, from
boom to crisis)—and why some countries
but not others stay in growth booms.
• An equation of motion of “institutions”—
how do “institutions” evolve to explain the
four big facts of growth?
“Institutions”: General or Specific?
• While there are demonstrable general
differences between countries in the overall
quality of institutions, there are also huge
variances of institutions within countries
• The “quality” of the institutional environment as it
affects specific industries (and/or firms) varies—
the “institutions” for tea versus textiles versus
pharmaceuticals
• Moreover, with weak institutions there is huge
variances across firms—the “policy action” that
is the results from the application of the policy
depends on how it is chosen
Do the rules matter more or less than deals?
More variance across firms than across
countries
“Favored” versus “Disfavored” firms have
massively different experience—even with
the same rules
• Comparing Doing Business indicators of three
different indicators from the Enterprise Surveys
(e.g. days to get a construction permit, days to
get an operating license, days to clear customs)
• Massive differences between the DB estimates
and ES estimates in general
• Massive differences across firms—up to a year
between 10th and 90th percentile in time to get
construction permits
Same country: Peru
90th percentile a year
DB 210 days
10th percentile 9 days
Three levers to explain the world
• A “product space” with products arrayed
conceptually according to the extent to which
their “capability” or “functionality” inputs are
similar
• The “receptivity” with which sector/firm
performance translates into increased public
action to augment capability
• The specificity of the “acceptable ask” in
receptivity—person/firm specific to economywide (an element of politics which is
“institutionally” constrained (or not)).
An empirical product space (goods
arrayed by how likely they are to be
co-exported)
A simplified, conceptual product space arrayed by
the similarity of “public action” inputs (laws,
regulations, infrastructure, skills, etc.)
Cluster of activities with similar
Public action inputs
An industry produces when its “intrinsic”
profitability (determined by technology,
endowments, world prices) plus contribution of
public inputs exceeds a threshold—once in
production it climbs up towards the potential
Cluster of activities with similar
Public action inputs
So far, this is “Monkeys and Trees”—The second
lever is the receptivity is how the public inputs
respond to production--first in height
Firms/industries who are
producing use some of
their revenue/profits to
ask for additional public
actions---better
regulation, protection,
specific inputs, tax breaks
Cluster of activities with similar
Public action inputs
The third element is the “acceptable ask”—what is
it that a firm/industry can lobby for? Possibilities?
The third element is the “acceptable ask”—what is
it that a firm/industry can lobby for? Only level
playing field—government is responsive, in
principle, but only for “all actors” actions
Moderate spillovers (to neighbors in this product
space)
Firm/industry specific ask—which could include
harming (lowering profitability of) close competitors
So, here is the dynamics: industry becomes
“capable”, begins to expand output, attempts to
pull up a tent after them, the shape of the tent is
constrained
Positive dynamic feedback
loop with (a) receptivity and
(b) spillovers
So, here is the dynamics: industry becomes
“capable”, begins to expand output, attempts to
pull up a tent after them, the shape of the tent is
constrained
Positive dynamic feedback
loop with (a) receptivity and
(b) spillovers
This is Hirschman(esque)
• Hirschman’s “backward linkages”—but instead
of in a purely input/output space of market
mediated demands leading to “integration” the
“backward linkages” are in the public action
space, mediated by politics (and hence the
institutions of politics which construct both
receptivity and acceptable ask
• “Uneven” development as it is the receptivity as
output expands that finds/identifies the needs for
public inputs—the “balanced growth”
perspective cannot anticipate what is needed.
Five Scenarios
• Centered on remote section of product space—
so no feedback loop (a model of boom and bust
due to politically/institutionally mediated
resource curse).
• Acceptable ask is too narrow (oligarchic
chimneys)
• No feedback because no activities to generate
pressure (poverty trap)
• Less dense part, booms that peter out.
• Dense part of product space, responsive,
limitations on “specificity” of lobbying
Scenario I: Success in a part of the product space
that is “remote” (institutionally determined spillover
to distance in product space)
Implications:
Growth spurt as economy moves up the industry
Stagnation afterwards
Economy hinged to ups and downs in potential
(e.g. terms of trade)
Average “public input” quality never improves
“Institutional quality” has high variance—great
Institutions for tea/coffee/oil/gold
Scenario II: Acceptable ask unconstrained—
oligarchic chimneys—the “private sector” owns the
state—and naturally hates competitive capitalism
(e.g. “Seize the state, seize the day”)
•Bursts of rapid growth
•Self-limiting
•Variance increases
(deals, not rules)
•Average institutions do not
improve
Scenario III: Threshold (far) exceeds existing
contribution of public action inputs at intrinsic
profitability—a “Zombie” (Ghost country)
Scenario IV: Production starts in moderately
dense part—but doesn’t spread
•Larger/longer growth boom—might
only peter out at medium income levels
•Might have high average level of
public inputs (just not in dense part)
•“Institutions” might be middling (on averag
Scenario V: “Sweet spot”—profitability in dense
part of product space, government receptive,
acceptable ask is constrained—everything is
possible
Positive dynamic feedback
loop with (a) receptivity and
(b) spillovers
Staying in the sweet spot for
growth
• Dense parts of the
(institutionally ordered) product
space
• Responsiveness of
government to needs of
emerging industries (“high
bandwidth” responsiveness)
• Right level of spillovers of
industry demand--tents steep
enough to generate reform, flat
enough to not build oligarchic
chimneys
Outcomes
• (Potentially) rapid growth with
S-curve dynamics into newly
profitable industries (within
state dynamics faster or
slower)
• Small exogenous shocks (to
profitability or capabilities) sets
off rapid growth with no
change in overall “institutional
quality”
• Public inputs gradually expand
so that measured “quality”
across all industries eventually
re-converges
Output/growth dynamics as move through
product space: size and duration of boom
determined by potential (and within state
dynamics)
Level of
output
Potential NR plus
little cluster plus bug cluster
Potential NR plus
little cluster
Just NR
Steady growth of
leaders
Perpetually poor
countries
Once at maximum capability
growth constrained by expansion
of potential (both new industries
and moderate growth of existing)
Interaction of low “endowment”
and poor public inputs—can be a
long time
Rapid catch up
Sweet spot of feedback (dense,
receptive, limited ask)
Episodic growth with
booms and busts
•Sparse part of product space
•Oligarchic chimneys (weak
institutions)
•Level playing field too low (selfdiscovery)
Long and Short of Growth Theory
Unification
• In long-run (or cross section) prosperous countries have
“high quality” institutions and poor countries have bad
institutions
• In long-run prosperous countries produce in nearly all
parts of the product space (since they have all
capabilities) (not specialization)
• Small exogenous changes can make segments of the
product space profitable—even with “weak” overall
institutions (but policy reform might not)
• Some booms die and some survive—depending on the
location in product space, receptivity, and acceptable
ask—leading the observed variability with growth booms
and busts (over and above pure resource boom/bust
Four ways of staying in the sweet
spot (experiences of rapid growers)
• Pre-commitment to integration with higher
institutional quality regions (so that foreign
players pressure institutions)
• Industrial policy through large industry
groups
• Ordered deals as industrial policy
• Keeping up with the zones’s
Rapid Growers through precommitment? Periphery of Europe
(Greece, Italy, Portugal, Ireland, Spain)
• Commitment to join “Europe” forced limitations
on the ability of domestic firms to lobby for
“specific” benefits, even during periods of “weak”
institutions
• Put pressure for better “level playing field”
directly into policy decision making
• Other examples? (Caribbean states that
remained pre-committed (e.g. Puerto Rico),
more recently EU accession for early entrants
(e.g. Poland), Hong Kong)
Rapid Growers with “managed
capitalism” (e.g. Japan, Korea)
• Industrial organization of large groups with
interests in multiple sectors and organized
inter-penetration of business and
government at highest levels
• Inter-penetration allowed for “high
bandwidth” responsiveness
• Multiple groups with opposing interests
represented prevented oligarchic capture
• Performance driven by government
Rapid growers with authoritarian
corruption: “Ordered Deals” corruption
as industrial policy
• Industrial policy is the differential favoring of
public inputs (of various kinds, including policy
and/or its enforcement) across industries
• Corruption is the use of public authority for
private gain
• Indonesia is a case where the combination of
the two provided for rapid growth for 30 years as
“ordered deals” were available for a modest
number of selected industrial groups (e.g.
military, family, some conglomerates)
Replication of “good institutions” in
a limited place: Keeping up with
the Zones’s
• Create a favorable investment climate only
inside a greenhouse--firms inside a designated
“space” (which may be conceptual, not
geographic) get favored institutional
environment
• Makes new activities possible
• Is there a dynamic of replication and
responsiveness?
• Other examples? (Mauritius (the 22nd fastest
grower), zone boomlets in other economies (e.g.
Dominican Republic?)
Descargar

The guts of a GUT: Elements of a Grand Unified Theory of