Transcript Doing growth diagnostics in practice

Growth Diagnostics in
Practice
Applied Inclusive Growth Analytics Course
June 29, 2009
Susanna Lundstrom, PRMED
Outline


What is so special about diagnostics?
Three common tools, with pros and cons:


Cross-country regressions
(Growth Accounting)
International Benchmarking and Indices

All are used in Inclusive Growth Analytics, but more is added


Basic Principles in Growth Diagnostics
Applications throughout the course!
See Hausmann, Klinger and Rodrigo (2008)
Growth research vs. Growth diagnostics

In growth research the subject is growth with
countries as observations


In growth diagnostic, the subject is the country



What do policy X do on average, to a randomly selected
country?
What particular problem does the country in question
have? How would this specific country react to policy X?
Informed by growth research…
Analogy: Medical research and practice
Cross-country growth regressions

“What factors affect growth in a typical country?”
^
^
^
^
g t   log  yt 1   1 x1t   2 x2t  ..   N xNt   t




Countries are just observations
But running out of degrees of freedom
Very important in understanding potential
constraints
If used with interaction terms – more context
specific g   log  y    x   x x  ..   x  
^
t
^
t 1
1
^
1t
2
^
2 t 1t
N
Nt
t
^
^
g t
 1   2 x2t
x1t


Given the presence of X2t, X1t will have this effect on gt
But running out of degrees of freedom…
Cross-country growth regressions
- Some characteristics

Normally assumes separability (if no interaction terms)



Assumes linearity



All x’s are substitutes of each other
I.e., you can compensate failures in one area by overperformance in other areas. But if there are binging
constraints….
Assumes monotonicity and linearity in the x’s - in the
absence of squared forms



The impact of variable x1 on growth is independent of the level
of the other x’s
I.e., no context-specific interactions taken into account
Increases in x from any level increases g
Increases in x from any level has the same effect on g
All β’s are the same for all countries
Cross-country growth regressions
- Some characteristics, cont

Only data that you have for all countries can be
included – often outcome rather than policy based


Example: Private credit/GDP ratio instead of a policybased index of financial liberalization
No price information in the equation

Supply or a demand problem?



Low supply, high price – potential constraint
Low supply, low price – low demands and not necessarily a
constraint
Results sensitive to the elimination of outliers

Sensitivity to groups of outlier – indicates contextsensitive effects
International Benchmarking
- Comparative countries
Similar countries (landlocked, conflict…) but with different GDP


Ideally, one should aspire to a “natural experiment” where the
selected benchmark is a replica in all but one respect to the country
under study
“Role models”


Where would we want to be in 2, 10 and 20 years?
A particular country or set of countries which performance or welfare
indicator wants to be attained by the studied country
Partial correlation
with a group of
countries
Compare with the
expected value
(fitted line) given
the GDP level in
the country
Gross Savings Ratio to GDP vs Real Per Capita GDP.
Average 1999-2007.
70
Gross Savings / GDP (Current
US$)

60
50
40
30
20
Tajikistan
10
0
-10
4
5
6
7
8
LN (Real GDP per capita US$)
9
10
11
International Benchmarking
- Indices and rankings


See examples on the IG website (“Data and
References”)
Complex systemic outcomes


How to collapse to a single dimension?



Does not map easily into policy
Take the average
Assumes linearity and separability
Assumes monotonicity and linearity in x:s


Is the optimal number of licenses zero?
Is an increase from 1 to 2 the same as from 9 to 10?
International Benchmarking
- Surveys and perception data

Sample selection bias



The binding constraint causes firms not to exist
and biases the survey
You talk to “camels” rather than “hippos”
International comparison of opinions


What does it take for a Swede to complain about
corruption?
Do all nationalities have a tendency to complain
more about taxes than human capital?
International Benchmarking
- General problem

Not obvious the focus should be on the areas
where you perform poorly compared to other
countries


Poor supply (and hence a constraint) ..or low
demand (and hence not obvious)?
Depends on the economic structure (human
capital more important in the US than Zambia)
Basic principles of country diagnosis
- Examples will follow throughout the course…
If a constraint is binding, then…
1.
The (shadow) price of the constraint should be high
2.
Movements in the constraint should produce
significant movements in the objective function
(e.g. GDP, or income of a specific group of ind.)
3.
Agents in the economy should be attempting to
overcome or bypass the constraint
4.
Camels and Hippos: Agents less intensive in that
constraint should be more likely to survive and
thrive, and vice versa
1. The (shadow) price of the constraint is high

Relative scarcity of a factor.


Estimate prices using regressions


Look at prices if they exist (interest rates, wages, etc.)
Example: Mincerian regressions
In other cases with no market price:


Infrastructure: congestion as the “price to pay”
Non-market valuation techniques, either based on
revealed or in stated preferences.

Ex: Hedonic prices, ICAs stated preferences
2. Differences in the constraint should produce
differences in growth


A constant cannot explain a change
Differences

Between time periods (look at trend breaks)


Between regions


Do growth periods following a relaxation of the
constraint? Do growth decrease when it is present?
Why are some prosperous and some lagging? The
constraint in question present in one but not the other?
Between groups of firms differently affected by
the potential constraint
3. Agents in the economy must be engaging in
efforts to overcome or by-pass the constraint

Examples:




Border controls  smuggling
Poor financial intermediation  growth
occurs within business groups (conglomerates)
Industry specific public goods binding 
growth in sectors less sensitive to specific
inputs or unusual level of cooperation among
successful producers.
Property rights and contracting binding 
Mafia
4. Camels vs. hippos



The surviving sector (“camels”) are those least
intensive in or least dependent on the binding
constraint (“water in a desert”)
In a good investment climate (“environment with
water”) the economy, and the thought comparative
advantages, may look differently (“hippos”)
What do analysis of camels reveal about potential
constraints?



Are successful groups particularly connected to the political
system?
Any “missing factors” within the successful industry?
The importance of analyzing hippos?

Example: Informal ICAs