Download 5 Supply responsiveness - Christopher B. Barrett

MIFIRA Framework
Lecture 5
Supply responsiveness
Chris Barrett and Erin Lentz
February 2012
Market integration
• For those with less familiarity with price analysis
techniques, review monitoring and analyzing data
documentation from LRP Learning Alliance.
– Lentz, E.C. (2011) “LRP: Monitoring and Analyzing Data 25 March
2011.doc”. Draft.
– Lentz, E.C. (2011) “Lentz 11 LRP Price Analysis - How Prices Change.ppt”
Draft.
– Lentz, E.C. (2011) “Lentz 12 LRP Price Analysis – Approaches to
Analysis.ppt” Draft.
• The spreadsheet “Maize Kenya price series detrend and
deseasonalize.xls” works through an example of how to deflate,
deseaonsalize, and then correlate historical maize price series
across several markets in Kenya
2
Market integration
• Core question: how price elastic is supply and thus how
much (unintended) price change might result from food
security interventions?
• Need to establish degree of spatial market integration.
Main method: study patterns of price co-movement.
– Prices that vary independently signal segmented markets (one
market’s equilibrium is within the price band created by the
central market price and transactions costs).
– Prices should co-move (roughly one-for-one) in markets that
are integrated (price bands are binding).
3
Supply Responsiveness
MIFIRA questions we address:
1c. How much additional food will traders supply at or near
current costs?
2a. Where are viable prospective source markets?
2b. Will agency purchases drive up food prices excessively
in source markets?
2c. Will local or regional purchases affect producer prices
differently than transoceanic shipments?
4
Why market integration?
Need to understand how well integrated local market is
with external markets in order to predict supply response
and price changes post-intervention.
Price
Slocal
Paut
Pint
Sexternal
Dlocal
Quantity
5
Theory of market integration
Two distinct concepts of market integration:
1) Tradability: physical flows of commodity (or
indifference to engaging in trade). This concept focuses
on trade flows and not on prices or transactions costs.
2) Competitive spatial equilibrium: Inter-market price
dispersion should be bounded by the commercial costs
of arbitrage if competition drives the marginal returns to
trade to zero. This approach focuses on prices (and
sometimes transactions costs) and not on trade flows. If
prices move independently, the local market is within
the price band. If they comove, it’s integrated.
6
Price bands
A “price band” is established by a central market price and
the fixed and variable costs of transacting with that market.
Local prices can fluctuate in equilibrium within that band.
Demand
Fixed cost
Pc
Variable cost
Supply
Pi
Nontradables
price band
Px
Quantity
7
Spatial Price Analysis
Step 1: Acquire and understand data
Find out how, where and when the data are collected.
Common errors:
- Not all wholesale price series.
- Raw vs. processed commodity prices.
- Not in common currency and physical unit (e.g., US$/MT)
- Time periods not matched correctly (weeks vs. months)
- Are data day-specific observations or period averages?
8
Spatial Price Analysis
Statistical methods for estimating market integration
First, plot the data and look for patterns and outliers.
Second, look for common trends. Beware that bivariate
correlation coefficients are a widely used by flawed measure
due to the distortion caused by common trends.
9
Spatial Price Analysis
Statistical methods for estimating market integration
Instead, detrend the data to clean out common inflationary
and seasonality patterns and then plot and correlate the
series.
Compute an index for each month, lumping all locations’
prices and years together, and then deflate each observation
by the appropriate month’s index.
At a minimum, clean out a common trend by firstdifferencing the series, then correlate Δp1 and Δp2 instead of
levels.
10
Spatial Price Analysis
Statistical methods for estimating market integration
Now plot and visually examine the detrended and
deseasonalized data series.
Compute the bivariate correlation coefficient between the
series. Where multiple markets’ series available, generate the
full matrix of bivariate (detrended) correlation coefficients.
Lilongwe
Dowa
Ntchisi
Dowa
0.88
Ntchisi
0.69
0.87
Kasungu
0.68
0.74
0.84
Mchinji
0.80
0.83
0.79
Source: Barrett (2009), MIFIRA
Kasungu
0.77
11
Spatial Price Analysis
A multivariate regression approach
λ = partial correlation b/n p1 and p2 controlling for
common exogenous factors
If 0<λ≈1, then there is market integration.
12
Spatial Price Analysis
Limitations to using price comovement analytic
Simple spatial price analysis must be interpreted with
caution. This is especially true when:
- trade flows are discontinuous (as w/ “flow reversals”).
- the costs of spatial arbitrage change significantly (and nonrandomly) over time.
13
Parity Bounds Model (supplementary)
Parity Bounds Model
Statistically estimates how frequently prices consistent with:
(i) market integration under spatial equilibrium
(ii) market segmentation in equilibrium
(iii) market disequilibrium
Typically beyond the capacity of MIFIRA analysts due to
data, computational and time constraints.
14
Parity Bounds Model (supplementary)
Parity Bounds Model
Example from Moser et al. (Agricultural Economics 2009):
Commune-level rice market integration in Madagascar
Market condition
Subregion
Region
National
Integrated equilibrium
68.9%
5.5%
12.8%
Segmented equilibrium
21.5%
31.1%
83.0%
Disequilibrium
9.6%
63.4%
4.3%
Some points worth noting:
1) Market integration can vary dramatically across scales,
depending on the distances and the nature of local
competition among traders.
2) Look for geographic heterogeneity in mkt integration.
15