1. Agricultural Market Information Systems in Africa: renewal
and impact
Montpellier (CIRAD), March 29-31, 2010
Measuring the Impact of Public Market Information System
g p y
on Spatial Market Efficiency in maize markets in Benin:
Application of Parity Bounds Model.
Sylvain KPENAVOUN CHOGOU
University of Abomey-Calavi, Benin
kpenavoun@yahoo.fr
2. Outline
Introduction
Estimation approach and data
Results and discussion
Results and discussion
Conclusion
3. Introduction
In Benin, 70% of the total labour force is
employed in agriculture and the share of the
employed in agriculture and the share of the
sector in export earnings is more than 60%
(Cotton);
Agricultural liberalization reforms undertaken by
Benin Government in the 1990s to build up
Benin Government in the 1990s to build up
efficient markets that benefit poor or smallholder
farmers;
MIS promoted as an accompanying measure of
reforms, supported by FAO and GTZ, etc.
reforms supported by FAO and GTZ etc
4. Introduction
MIS was expected to:
correct the information asymmetries;
correct the information asymmetries;
give more bargaining power to farmers;
make a more transparent market,
strengthen competition get the institutions
competition, get the institutions
right, reduce transaction costs and improve
market integration and market efficiency;
g y
and then contribute to improve the well
being of the producers who live in rural areas.
being of the producers who live in rural areas
5. Introduction
So, large positive impacts are expected from MIS,
but empirical suffisant works to show them are
missing in SSA (Tollens, 2006).
Several key questions still not answered carefully:
e.g.
Have small farmers obtained better arrangements
(market or contract) when selling their surpluses?
( k t t t) h lli th i l ?
Have small farmers obtained better access to
market?
What is the extent to which the reforms have
improved the spatial market efficiency?
improved the spatial market efficiency?
6. Introduction
• Research objective
Measure how the agricultural reforms, in particular
M h th i lt l f i ti l
the PMIS, have affected the market performance of
maize, the major staple food crop in Benin.
, j p p
7. Empirical model: Parity Bounds Model
p y
Distinction between market efficiency and
market integration
Spatial market efficiency is an equilibrium condition
whereby all potential profitable spatial arbitrage
opportunities are exploited;
t iti l it d
Spatial market integration is defined as the extent to
which demand and supply shocks arising in one
which demand and supply shocks arising in one
location are transmitted to other locations (Barrett
and Li, 2002).
Market analysis depends on available data
7
8. Table 1: A Hierarchy of Market Analysis Methods
Method group Examples Characteristics
Level I methods Price correlation test Price correlation is relatively simple way to measure
utilize only price (Lele, 1967; Jones, 1968) market integration but suffers from various weaknesses
data, assume Test market integration for the marketing system as a
constant inter‐ Delgado’s variance whole instead of pair-wise test of market integration;
market transfert cost decomposition approach The method purge out the common trends and seasonality
(Delgado, 1986) present in the price series before testing for market
integration
This method allow testing market segmentation, short‐run
The Ravallion method market integration, long‐run market integration between local
(1986) and central markets after controlling for seasonality, the
common trends and autocorrelation
Engle and Granger
Take into account the presence of stochastic trends in the
(1987) cointegration
price series but pair-wise test of market integration.
analysis
Take into account the presence of stochastic trends in the
Johansen (1988)
price series
Multivariate cointegration
Test market integration for the marketing system as a
analysis
whole .
9. Table 2: A Hierarchy of Market Analysis Methods
Method
Examples Characteristics
group
Level II Based on the idea that the presence of
methods Threshold Autoregression or ransaction cost creates a « neutral band »
combine Threshold Cointegration
g whithin which the prices in different markets
transaction (Blake, and Fomby, 1997; are not linked;
cost and Goodwin and Piggott. 2001; Does not require the observation of
price etc.) transaction cost;
data, and
thus more Allows to measure the probability of being
closely in different market effciency regimes that
resemble are consistent with the equilibrium notion
spatial Parity Bounds Model of that all spatial arbitrage opportunities are
equilibrium Spiller and Wood (1986); being exploited (Enke 1951; Samuelson
theory
eo y Se o ,
Sexton, Kling a d Carman
g and Ca a 1964; Takayama and Judge 1971);
96 ; a aya a a d 9 );
(1991); Park, Rozelle and Can indicate not only wether the markets
Huang (2002); Baulch (1997); are efficient but also the extent to which the
Penzhorn et Arndt (2002). markets are efficient;
Possible estimate with incomplete price
series
10. Table 1: A Hierarchy of Market Analysis Methods
Method
group Examples Characteristics
Level III
methods
combine
trade flow, Parity Bounds Model of
Allow a clear distinction between spatial market
price data Negessan and Myers (2007)
effciency and spatial market integration
integration.
and time- or Barrett and Li (2002).
series
transaction
cost data
11. Table 2: Trade regimes between two markets
Trade regimes
Trade
Pit − Pjt = TC jit (1) Pit − Pjt p TC jit (2) Pit − Pjt f TC jit (3)
λ11 λ21 λ31
With Trade
Perfect market efficiency or Imperfect integration Imperfect integration
perfect market integration Market inefficiency Market inefficiency
λ12 λ22 λ32
No trade
Market efficiency Market efficiency Segmented disequilibrium
λ1 = λ11 + λ12 λ2 = λ21 + λ22 λ3 = λ31 + λ32
With or witout trade
Market efficiency condition Autarky market condition Market inefficiency
condition
λi is the probability of being in regime i et λij is the probability of being in sub-regime j of regime i.
TC jit is the transfer cost for trading from market j to market i at time t.
g
Pit et Pjt are prices in markets i and j, respectiv ely.
11
12. Market performance measure with Parity Bounds Model
Market performance has several characteristics that together help
describe the development of the market:
Efficiency rate;
Inefficiency rate;
Arbitrage rate;
Arbitrage opportunity rate;
Autarky rate;
12
13. Arbitrage opportunity rate ( λ1 + λ3 )
It is the probability
It is the probability that the arbitrage opportunities exist between two
the arbitrage opportunities exist
markets.
Arbitrage rate, a more appropriate measure of integration
The probability that arbitrage is observed when arbitrage opportunities
Th b bili h bi i b d h bi ii
exist or the extent to which arbitrage opportunities are realized by
traders.
λ1
Arbitrage rate =
( λ1 + λ3 )
Autarky rate
Autarky rate
The percent of trading periods in which two regions do not trade because
price differences are less than transaction costs.
λ2
Autarky rate =
( λ1 + λ2 + λ3 )
13
14. Subperiods for Parity Bounds Model estimation
We estimate a parity bounds model of interregional trade for four
parity-bounds
subperiods to characterize how multiple aspects of market
performance change during the process of liberalization and
measure the impact of PMIS:
1988 à 1992: Period before reforms;
1993 à 1996: Period with PMIS: publication of monthly bulletins
993 à 996 e od t S pub cat o o o t y bu et s
and the broadcasting of prices and market information on national
public radio; Market infrastructure investment; Major investment
in transport investment (road, easy to buy occasional cars, etc.);
1997 à 2000: Improvement of PMIS with the posting of maize
prices at different locations on each market place.
2001 à 2007: Broadcasting of prices and market information on
several regional and rural radios, development of GSM, sms and
web services.
14
15. Likelihood function
In order to estimate the probability of being in one regime or another we need
another,
to define the likelihood function:
T
L = ∏⎡λ1 f1t + λ2 f2t + (1− λ1 − λ2 ) f3t ⎤
⎣ ⎦
with the density functions for each
regime:
t =1
⎡ it ⎤ ⎡ ⎤ ⎡ ⎤
⎡ ⎡
( σ ⎤⎤
P − Pjt −TCjit u ⎥⎥
1 ⎢ P − Pjt −TCjit f⎥ = ⎢ 2 ⎥ϕ ⎢ Pit − Pjt −TCjit ⎥ ⎢1−Φ⎢ it σe ⎥
)
f1t = ϕ 2t
⎢ 2 2⎥ ⎢ 2 2 ⎥⎢
⎢ ⎢ ⎥
⎥⎥
σe ⎢ σe ⎥ ⎣(σe +σμ ) ⎦ ⎣ (σe +σμ ) ⎦ ⎢ ⎢ (σe +σμ )
1 1 1
2 2 2 2 2
⎣ ⎦ ⎢
⎣ ⎥⎥
⎦⎦
⎣
⎡ 2
⎡
⎤ ⎡ P − Pjt −TCjit ⎤ ⎢
⎡
( )σ ⎤⎤
− P − Pjt −TCjit ν ⎥⎥
⎢ it σe ⎥
f3t = ⎢ ⎥ϕ ⎢ it
⎥ ⎢1−Φ⎢ ⎥
⎢ (σe2 +σν2 ) 2 ⎥ ⎢ (σe2 +σν2 ) 2 ⎥ ⎢ ⎢ ⎥⎥
1 1 1
⎣ ⎦ ⎣ ⎦⎢ (σe2 +σv2 ) 2
⎢
⎣ ⎥⎥
⎦⎦
⎣
16. Data
The data used in the study are monthly maize consumption
y y p
price series over the period January 1988 to December 2007.
This study analyses the price time series observed in seven
market places, well distributed over the major mai e
market places, well distributed over the major maize
consumption and/or production regions:
Cotonou, Azove and Ketou (located in the South),
Bohicon and Glazoue (located the central region) Parakou
and Glazoue (located the central region), Parakou
and Nikki are northern markets;
Parakou, Bohicon and Cotonou are urban markets;
Kétou, Glazoué, Azové and Nikki are rural markets.
Kétou Glazoué Azové and Nikki are rural markets
These data were collected by ONASA (Office National d’Appui
à la Sécurité Alimentaire).
Following Baulch (1997), we have constructed the time series
transfert costs using one‐time transfert costs estimate from
interviews with traders and for adjusting for inflation.
16
17. Results and discussion
Table 3: Maximum Likelihood Estimates of Parity-Bounds Model for
maize market in Benin
Arbitrage Autarky rate
Efficiency Inefficiency Arbitrage
Time
rate ⎛ λ2 ⎞
⎞ ⎜ ( λ + λ + λ ) = λ2 ⎟
rate rate opportunities
⎛ λ1 ⎜ ⎟
Periods
( λ1 ) ⎜
⎜ (λ + λ ) ⎟
⎟
⎝ 1 2 3 ⎠ ( λ3 ) rate ( λ1 + λ3 )
⎝ 1 3 ⎠
0.22 0.81 0.61 0.17 0.39
1988-1992
(0.11)
(0 11) (0.33)
(0 33) (0.34)
(0 34) (0.33)
(0 33) (0.33)
(0 33)
0.06 0.37 0.51 0.43 0.49
1993-1996
(0.07) (0.47) (0.37) (0.40) (0.37)
0.21 0.50 0.39 0.40 0.61
1997-2000
(0.22)
(0 22) (0.46)
(0 46) (0.37)
(0 37) (0.44)
(0 44) (0.37)
(0 37)
0.23 0.35 0.16 0.61 0.84
2001-2007
(0.21) (0.33) (0.30) (0.35) (0.30)
0.16 0.54 0.41 0.43 0.59
1988 2007
1988-2007
(0.13)
(0 13) (0.44)
(0 44) (0.36)
(0 36) (0.41)
(0 41) (0.36)
(0 36)
The estimated standard errors for each parameter estimate are reported in
parentheses.
The results presented are averages of each parameters estimate with the level of the
15 pairs of studied markets. Th are the estimates of 75 equations.
i f t di d k t They th ti t f ti
17
18. Conclusion
We find that:
The marketing reforms did not significantly improve the
Th k ti f did t i ifi tl i th
degree of efficiency or of spatial integration of markets;
But they did induce new marketing opportunities, which still
B t th did i d k ti t iti hi h till
remain under‐exploited;
The rate of autarky, which measures the spatial range over
Th f k hi h h i l
which transactions did not occur between two markets due to
high transaction costs, shows a decreasing trend over time.
g , g
Improvements are observed on a few markets.
18
19. Conclusion
However, the high levels of inefficiency prevent the system
from providing farmers and consumers the services they need.
This study therefore recommends the implementation of
more efficiency‐raising policies in order to encourage
competition and allow the system to fulfill the expectation of
farmers and consumers.
19
21. Agricultural Market Information Systems in Africa: renewal
and impact
Montpellier (CIRAD), March 29-31, 2010
Measuring the Impact of Public Market Information System
g p y
on Spatial Market Efficiency in maize markets in Benin:
Application of Parity Bounds Model.
Dr Ir Sylvain KPENAVOUN CHOGOU
University of Abomey-Calavi, Benin
kpenavoun@yahoo.fr