Analysis of effectiveness of modern information and communication technologies

on maize marketing efficiency

in Malawi markets

BY

SARAH E. TIONE CHOWA

Economist

Ministry of Agriculture, Irrigation and Water Development

Department of Planning

Malawi

1

Background • Malawi is an agro-based economy, with smallholder and

estate sub-sector.

• The main food crop in Malawi is maize which is held as synonymous with food security.

• Most rural households especially smallholder farmers are net purchasers of maize which makes them vulnerable to prices (Sahley et. al., 2005)

• Maize is traded throughout the country through ADMARC, Private traders and NFRA

2

Background • Although agriculture is important, markets do not work efficiently esp.

among smallholder farmers (Goletti and Babu, 1994 and Jayne et. al., 2008)

• Liberalization was intended to equip smallholder farmers with successful marketing instruments, it brought a new marketing challenge• That of poor access to reliable and timely market information or asymmetry of

market information.

• The lack of market information substantially increases transaction cost and reduces market efficiency, (Barrett 2008)

• Thus, inefficient markets and poor access to market information have mainly contributed to poor access to markets and small volumes of high valued products offered.

3

Transaction Cost drivers

State of Infrastructure (Z)

Information Services (IS)

Assets Endowment (K)

FARM HOUSEHOLD PRODUCTION

Subsistence Needs

Marketed Surplus (QN)

Market 2 or Village Market 2

Market 1 or Village Market 1

Regional Markets

International Markets

Conceptual Framework

4

Conceptual Framework• The theory behind the framework is based on

• Transaction cost theory from New Institutional Economics. The theory looks at;

• The transaction cost and market linkage, • Transaction cost and performance of spatially and temporally

separated markets and • Transaction cost and market participation (Kirsten and Karaan,

2005).

• Law of one price• Central to all forms of market integration or co-movement of

prices are transfer or transaction costs, comprising transportation, storage and processing charges plus a modest allowance for trader’s normal profit (Baulch, 1997).

5

Purpose of study • In 2004, Government of Malawi introduced Malawi Agriculture

Commodity Exchange (MACE) under Initiative for Development and Equity in African Agriculture (IDEA)

• The initiative promoted use modern ICTs to improve access to agricultural market information for all agricultural commodities including maize

• IDEAA-MACE aimed at• Linking buyers and seller, • Empower farmers with relevant and timely information • Provide a transparent and competitive price discovery mechanism

• The Market Information System used by MACE among smallholder farmers was Market Information Points (MIP), SMS and radio programme

6

MIP Source: MACE Base Line Survey by Phiri (2006)

Purpose of study

• Although MACE promoted use of modern ICTs, the question still remains “how much are the ICTs initiatives improving the efficiency of markets in Malawi”

• This study analysed spatial market integration of 9 markets before and after 2004 using threshold error correction models

8

Methods and Analysis• Market integration deals with linkages among markets

• Co-integration is a property of two or more variables which have shown to be integrated and have a long run equilibrium point

• The theory is based on the ‘Law of One Price’• That is, for spatial arbitrage to happen, price difference of a

commodity in spatially separated markets should be equal to marginal transfer cost of moving the product (Rapsomanikis et. al., 2006)

9

Integration and Co-integration

• Although prices might drift apart in the short-run, co-integrated markets have the tendency to co-move together in the long-run.

• Thus, error correction model was used to analyse integration and price transmission

• Both liner and threshold error correction models were used

• Before assessing price transmission in pre and post ICT initiative (MACE),

• The study determined the long – run co-integration and granger causality of all 9 markets from January 1992 to December 2009 using real price maize market data

10

Spatial price transmission• Price transmission was based on unidirectional co-integrating markets

• Based on assumption of symmetric price transmission

• Thus, price transmission was done and compared between Linear and threshold autoregressive error correction models

• LINEAR AUTOREGRESSIVE ERROR CORRECTION MODELS

……………… (1)

• The ηt is the error term used to defined the error correction model and it defines the deviation between prices in two different markets.

• When β = 1 the deviation becomes non stationary leading to no integration

tjtit PP

11

Linear Autoregressive Error Correction Models

• Using ηt, the linear autoregressive error correction equation is

……..………………..…2

• Where

• Using equation (2), the estimated ρ shows the adjustment parameter on lagged price difference

• It indicates the extent to which price differences in the previous period are ‘corrected’ back to equilibrium price

• The Linear Autoregressive error correction model fails to allow for a zone of trade inactivity or the ‘parity bound’ when price spreads fall below a threshold that reflects transfer cost between markets

1tt

jtitt PP

12

Threshold autoregressive (TAR) error correction model

• Assuming from equation (2) ηt follows a threshold autoregressive behavior

• Spatial price transmission in long-run equilibrium under competitive behavior is given as (Myers, 2008)

• If q = 0 (Regime 1)

• If q > 0 (Regime 2)

• If q < 0 (Regime 3)

Where ρit - ρjt = ηt ; q is the quantity of commodity traded between the markets in two way

direction; c the marginal transfer cost and it is assumed symmetric irrespective of the direction of trade flow

• Thus the TAR model can be estimated as;

cPP jtit

cPP jtit cPP jtit

13

• Regime I

• Regime 2

• Regime 3

Where ηt = pit -pjt and ct is long run transfer cost at t

There is a non-linearity at the threshold which allows the price spread to display different behavior inside versus outside a ‘parity bound’ defined by long transfer costs.

Since there is no separate transfer cost data, ct is determined by

………………………………………….. 3

where T is total number of observations,

Thus from the regimes,αt is used to determine the rate of price adjustment but does not show the

value of adjustment (The values of αt are btwn -1 and 0)

tttk

ktktt cif

......................1

1

tttk

ktktkttttt cifccC

......................)()()(1

111

tttk

ktktkttttt cifccC

......................)()()(1

111

itt PT

tC 2010 )1(

)(

14

TAR error correction model cont.

• To assess the value of adjustment, use half life.

……………….. 4

• Half life is the measure that helps interpret the spread of adjustment of price back to the parity bound in regimes 2 and 3.

• It is the time it takes for trade to increase and drive the price spread half way back to the parity bound, when there is a supply or demand shock that raises price spread above the parity bound

• It is given in weeks

• The TAR model was estimated without a constant

)1ln(/)5.0ln( h

15

Spatial Integration Of Maize Markets In Malawi

16

Causal relationship between co-integrating markets

• The results showed 8 unidirectional causal relationships between markets

• Karonga – Rumphi • Mitundu – Lilongwe • Mitundu – Lizulu • Bangula – Karonga • Mitundu – Karonga • Lunzu – Rumphi • Luncheza – Rumphi • Lunzu – Lizulu

• Price transmission was estimated based on the 8 co-integrating markets

17

18

Pre-ICT price adjustment results

• From the results, the fastest significant price adjustment factor was observed in Rumphi-Luncheza market link both in AR and TAR models

• The adjustment factor of 0.05 in AR model implies that it took 13.5 weeks for half of the price shock to return to the equilibrium price.

• In TAR model, the estimated adjustment factor of 0.07 implies that it took 9.7 weeks for half of the price shock to return to the equilibrium neutral price band

• This indicates that price adjustment speed is faster in TAR model because it considers the threshold where there is no price adjustment

• This shows TAR models are more appropriate in estimating price adjustment (Agrees with Van Campenhout, 2007; and Goodwin and Piggott, 2001),

19

Rumphi – Lunzu

545 -0.025***(0.005)

27.4 3.960 -0.034***(0.006)

20.0 0.057(0.041)

11.8 3.392 0.069(0.045)

9.7

Rumphi – Luncheza

821 -0.050***(0.106)

13.5 3.167 -0.069***(0.115)

9.7 -0.004(0.024)

173.0 4.421 -0.009(0.026)

76.7

Mitundu – Lilongwe

30 -0.030***(0.010)

22.8 4.129 -0.045***(0.012)

15.1 -0.120***(0.039)

5.4 3.556 -0.142***(0.041)

4.5

Mitundu – Lizulu

90 -0.014***(0.006)

49.2 1.740 -0.019***(0.007)

36.1 -0.186***(0.060)

3.4 1.944 -0.209***(0.062)

2.9

Lizulu – Lunzu

201 -0.024***(0.008)

28.5 2.1822 -0.035***(0.009)

19.5 -0.001(0.036)

692.8 1.6510 -0.003(0.037)

203.7

20

 Market Pair

Pre –ICT Post - ICT

Distance (km)

AR Model TAR Model AR ModelTAR Model

    Half-Life

Half-Life

Half-Life

Half-Life

Note: ρ denotes adjustment parameter on the lagged price difference (expressed as a percentage of mean prices in the two markets), δ is the estimated thresholds, expressed as percentage of mean price in the two marketsFigures in parenthesis are standard errors. *** and ** denote significance levels at 1 percent and 5 percent, respectively.

 Market Pair

Pre –ICT Post - ICT

Distance (km)

AR Model TAR Model AR Model TAR Model

    Half-Life

Half-Life

Half-Life

Half-Life

Karonga – Rumphi

176 -0.029***(0.010)

23.6 2.533 -0.043***(0.009)

15.8 -0.148***(0.025)

4.3 3.006 -0.189***(0.023)

3.3

Karonga – Mitundu

620 -0.030***(0.010)

22.8 3.107 -0.041***(0.009)

16.6 -0.065**(0.033)

10.3 2.878 -0.078**(0.034)

8.5

Karonga - Bangula

804 -0.078*** (0.018)

8.54 4.038 -0.124*** (0.015)

5.23 -0.069* (0.026)

9.69 3.5719 - 0.084* (0.024)

7.90

21Note: ρ denotes adjustment parameter on the lagged price difference (expressed as a percentage of mean prices in the two markets), δ is the estimated thresholds, expressed as percentage of mean price in the two marketsFigures in parenthesis are standard errors. *** and ** denote significance levels at 1 percent and 5 percent, respectively.

Post-ICT price adjustment results • The fastest significant price adjustment in post-ICT was observed in

Mitundu – Lizulu market link

• In standard AR model, the adjustment factor was 0.186 which implied a half-life of 3.4 weeks.

• In TAR model, the significant adjustment factor was 0.209 percent which implied a half-life of 2.9 weeks.

• The estimated threshold was 1.94 percent of the mean price.

• This entails that influencing factors that reduce transaction costs also influence the speed of price adjustment if there is a shock in the markets.

22

Comparison between pre – ICT and post – ICT price adjustment results

• The comparison was done on TAR co-integrating links that were significant in both periods

• That is Karonga-Rumphi; Mitundu-Lilongwe and Mitundu-Lizulu

• The integration between regional markets of Karonga and Mitundu show a faster adjustment in post-ICT period than in pre-ICT.

• It used to take 17 weeks for half of the price shock to return to parity bound in pre-ICT period with 3.1 percent estimated threshold.

• In post-ICT period, the estimated threshold was 2.9 percent and 9 weeks half-life.

• This implies that the transaction costs decreased and it was taking few weeks for half of the shock to return to the parity bound in post ICT period

23

Conclusion

• Estimated thresholds were lower in post ICT-TAR models and that it took fewer weeks for a shock to return half-way back to parity bound• This signify that modern ICT based market interventions influenced

reduction in search transaction cost thereby improving maize marketing efficiency in the post ICT period.

• Although price adjustment was faster in post-ICT period, the adjustment was not instantaneous.

• This can be attributed to, among other factors, transportation costs and market charges related to volume of trade (Myer, 2008).

24

Policy Recommendations• Enhance use of modern ICTs especially at farm gate

• Need to improve the market infrastructure to complement the efforts in reducing market information asymmetry

• Further areas of study• Use of parity bound models and threshold vector error correction

models that take into account asymmetric price transmission and estimate thresholds that vary over period of study.

• Compare use of different ICT tools in improving market efficiency

25

THANK YOU FOR YOU ATTENTION

26