Transcript Turning statistics into knowledge

Turning statistics into knowledge: use
and misuse of indicators and models
Data Day
Geneva May 18th
• Modeling: Partial vs General equilibrium
• The importance of estimation
• Indices
Turning statistics into knowledge
2
• Modeling: Partial vs General equilibrium
• The importance of estimation
• Indices
Turning statistics into knowledge
3
Modeling: Partial versus General equilibrium
Definitions
• Partial equilibrium implies that we only consider
a few markets at a time and we do not close the
models by including all economic interactions
across sectors (e.g., SMART, GSIM in WITS or
TRITS at the World Bank).
• In a general equilibrium setup all markets are
simultaneously modeled and interact with each
other (e.g., GTAP developed at Purdue
University).
Turning statistics into knowledge
4
Why partial equilibrium?
Advantages
• Minimal data requirement. We can take
advantage of rich WITS datasets. Crucial if
question is about:
– Bolivia or Uruguay and not the “Rest of South
America”
– Soya exports and not “Other cereals”
– Results of the trade model will feed poverty analysis.
Households produce corn or soya, not “cereals”.
Heterogeneity of impacts may be lost in a more
aggregate general equilibrium model.
Turning statistics into knowledge
5
Why partial equilibrium?
More Advantages
• Allows analysis of Doha negotiations more
accurately:
– In the WTO countries negotiate bound tariffs, not
applied (tariff “overhang” in many regions)
– Applied and bound tariffs are very different within
HS 10 Cereals. General equilibrium approach will
miss this.
Turning statistics into knowledge
6
Why partial equilibrium?
More Advantages
• Transparency
– Modeling is straightforward and results can be
easily explain. No “black box”.
• Easy to implement
– Excel sheet/SMART/GSIM
• Solves aggregation bias
Turning statistics into knowledge
7
Adding apples and oranges….
P
Pw+ta
Pw+Ta
Pw+Tf
Pw
Q
Apples
Oranges
Fruits
• No welfare cost associated with Ta: apples import
demand is perfectly inelastic. No tariff on oranges. So no
welfare cost associated with fruit protection.
• Aggregation bias suggests welfare loss =
Turning statistics into knowledge
8
Why partial equilibrium?
Disadvantages
• One has information only on a predetermined number of economic variables
(“partial” model of the economy)
• One may miss important feedbacks
– E.g., Labor market constraints. (But if you know
they are there you can model them)
• Can be very sensitive to a few (badly
estimated) elasticities.
Turning statistics into knowledge
9
• Modeling: Partial vs General equilibrium
• The importance of estimation
• Indices
Turning statistics into knowledge
10
The importance of estimation
Ex-post
• One can estimate the impact of a certain policy reform on
exports, trade creation, diversion, GDP growth, productivity
and with a bit of modeling utility (e.g., gravity equation)
Ex-ante
• One should estimate the critical parameters of the modeling
exercise (elasticities, economies of scale, etc..). Otherwise:
– Harris (1984) versus Head and Ries (1999)
– World Bank (2001) versus Hoekman et al (2004)
– GEP(2001) versus common sense
• Importance of comparing relative and not absolute results
Turning statistics into knowledge
11
But why do simulation results differ?
• Scenarios are not the same
– Full versus partial
– Different base years (benchmarks)
– Mixing with other reforms (fiscal policy, trade facilitation)
• Data are not the same
– GTAP data is standard, but PTAs, NTBs..
• Parameters (elasticities) are not the same
• Modeling assumptions differ
– Perfect versus imperfect competition
– Flexible versus rigid labor markets
– Endogeneity of TFP to trade openness
Turning statistics into knowledge
12
• Modeling: Partial vs General equilibrium
• The importance of estimation
• Indices
Turning statistics into knowledge
13
Indices: between analysis and narrative
• According to statisticians: “what cannot be counted
does not count”, but “do indicators try to count what
cannot be counted”?
• Composite indices are good for:
– Narrative
– And advocacy of particular reform/policy
– Decision making process if based on policies
rather than outcomes, and aggregated using a
proper technique.
Turning statistics into knowledge
14
Indices
Problems:
• Modeling versus estimation of weights of
different components (or subjective versus
objective criteria)
• Based on theory, not hand-waving (World
Bank’s OTRI versus IMF’s old TRI)
• Rankings and the importance of measurement
error (OTRI versus TRI or Doing Business)
Turning statistics into knowledge
15
Concluding remarks
• Keep it simple and transparent
• Don’t trust your guts: estimate everything you
can!
• Pay attention to measurement error
• Compare relative policy shocks not absolute
numbers
Turning statistics into knowledge
16