Lecture 5 Scenario Design for Regional Demand System

1
Lecture 5 Scenario Design for Regional Demand
System

• Laixiang Sun
• LUC, IIASA, Austria
• SOAS, University of London, UK

CHINAGRO 2nd Training Course 24 Sep. 2003,
CAS-CCAP, Beijing
2
Outline

• The basic of demand system in an AGE setup.
• Why must the design be systematic?
• What can we learn from households surveys?
• What can we learn from international comparison?
• Our approaches to have a systematic design.
• Concluding remarks.

3
1. Basic of demand system in an AGE setup
1.1. Linear expenditure system Most convenient
(discrete in time) setup for scenario design

• Choose Stone-Geary utility function for each
  individual consumer

• Maximising utility s.t. budget constraint yields
  the linear expenditure system

4
1. Basic of demand system in an AGE setup
1.2. Relationship between elasticities
expenditures
Partially differentiating the LES yields these
relationships

• In econometric analysis, we use households
  expenditure pattern to estimate elasticities.
• In scenario design, we involve in a reverse
  process Use acceptable future elasticities to
  establish future expenditure patterns (various
  shares).

5
2. Why must the design be systematic?

• Fine tuning income elasticities is not
  sufficient.
• It may violate consistent and constraint
  conditions given before (Section 1), including
  adding-up, symmetry, homogeneity, and
  non-negativity.
• It may lead to infeasible marginal shares of
  expenditures.
• Troublesome Engel properties.
• The typical problems of translating cross-section
  patterns into time-series patterns.
• The case of consumption vs saving in USA.
• Is it possible to have a systematic fine-tune?
• We may need more help from plural perspectives.

6
3. What can we learn from surveys?

• Estimate current patterns of consumption and
  expenditures across regions, rural and urban
  divisions, and income groups (an example from
  CCAPs tables).
• Various shares.
• Matrixes of elasticities (w.r.t. price,
  expenditure, and income).
• Understand the limitation of the estimation based
  on cross-section or pooling data.
• Same utility function
• Same probability distribution
• The estimates are suggestive or illustrative, but
  not deterministic!

7
Source CHINAGRO Working Package 1.7 Income
Growth and Life-style Change, by CCAP-CAS
8
3. What can we learn from international
comparison?

• Estimate consumption patterns across the
  development spectrum (different p.c. GDP levels).
• Difficulty Engel curves across development
  spectrum is non-linear.
• Marginal and average budget shares are also
  non-linear across development spectrum.
• These non-linearity is of fundamental importance
  for demand scenario design and analysis!

9
Example 1a Average (fitted) budget shares for
food products (at mean PPP prices, 1985)
Reference Changes in the Structure of Global
Food Demand, by J. Cranfield, T. Hertel, J.
Eales, P. Preckel, Purdue University, 1998.
10
Example 1b Marginal budget shares for food
products (at mean PPP prices, 1985)
Reference Changes in the Structure of Global
Food Demand, by J. Cranfield, T. Hertel, J.
Eales, P. Preckel, Purdue University, 1998.
11
Example 2 Non-parametric estimation of meat
demand and per-capita income (1975-97)
Reference Can We Feed the Animals? The Impact
on Cereal Markets of Rising World Meat Demand,
by M. Keyzer, M. Merbis, I. Pavel, C. van
Wesenbeeck, SOW-VU, 2003.
12
4. Our approaches to have a systematic design

• 4.1. Basic Strategy
• Run estimations and simulations based on AIDADS
  or extended LES with switches to establish
  relationship between consumption patterns (shares
  and expenditure elasticities) and income growth.
• Incorporate this externally calibrated
  relationship into the AGE with Stone-Geary form
  of utility function.
• The relationship can also be projected to the
  time dimension, with the help of an externally
  calibrated income growth patterns across regions,
  rural urban divisions, and income groups.

13
4. Our approaches to have a systematic design

• 4.2. Basic on AIDADS
• AIDADS stands for An Implicit, Directly Additive
  Demand System.
• It has been regarded as the best practice
  benchmark model to detect the relationship
  between consumer demand and income growth.
• It starts from an implicitly directly additive
  utility function as follows.

14
4. Our approaches to have a systematic design

• 4.2. Basic on AIDADS
• Solving the 1st order cost minimization
  conditions yields the budget share form

• If ag ßg for all g, AIDADS simplifies to the
  LES.

Reference Estimating consumer demands across
the development spectrum Maximum likelihood
estimates of an implicit direct additivity
model, by J. Cranfield, P. Preckel, J. Eales
T. Hertel. Journal of Development Economics, 68
(2002), 289-307. Projecting world food demand
using alternative demand systems, by W. Yu, T.
Hertel, P. Preckel, J. Eales, Purdue University,
2002.
15
4. Our approaches to have a systematic design

• 4.3. Basic on extended LES with switches
• Demand function is as follow

• The indirect utility function of this system has
  close-form expression and meets the requirements.
• Its marginal and average expenditure shares
  changes across the switching points.

16
5. Concluding remarks

• Fine tuning income elasticites alone may lead to
  inconsistency and a systematic scenario design of
  demand system is needed.
• Systematic design means to integrate plural
  perspectives and best-available information into
  a consistent framework. Consistency across income
  levels (or over time) is essential.
• Given the fact that improvement in data and
  estimation models/techniques is evolutionary,
  improvement in scenario design will follow the
  same track as well.