Docstoc

Spatial Pricing Efficiency in Fiji's Municipal Food Markets

Document Sample
Spatial Pricing Efficiency in Fiji's Municipal Food Markets Powered By Docstoc
					Spatial Pricing Efficiency in Fiji’s Municipal Food Markets 1
Chris Doucouliagos, Henry Haszler and Phillip Hone
3

2

The Fiji Islands economy is widely held to be constrained by a lack of market and transport infrastructure. Such infrastructure problems can reduce the efficiency of markets and result in wider than optimal disparities in food prices, both between markets and over time. In this paper we use monthly price data for 22 locally produced crops from Fiji’s major municipal markets, for the period 2001-2006, to identify the extent of the spatial pricing efficiency between these markets. Short-run price responses are estimated through SURECM. Relative price movements between markets indicate the potential for profitable arbitrage which will maximise the economic value of food to the Fijian economy. The arbitrageurs might be middlemen or individual farmers shifting their own produce between selling centres. We find that the municipal markets are generally spatially price efficient in the technical sense that prices return to their long-run relationships after a shock. But after a disturbance prices generally still take some two months to return to their equilibrium with the reference market, the capital Suva. This suggests there is scope for policy action to improve the spatial pricing efficiency of Fiji’s food markets. An enhanced market price reporting system is one option that deserves consideration. Key words: Law of One Price, pricing efficiency, arbitrage, Fiji

We wish to thank Nacanieli Takele, Sakiusa Tubuna, Richard Veit, Sakeasi Waikere and Paul Waqa, all staff of the Fiji Islands Ministry of Agriculture, Fisheries and Forests for the background information inter alia on the operations of Fiji’s municipal produce markets underpinning this paper. Paul Waqa (Fiji AgTrade) provided the municipal price data. As usual, all errors and omissions are the responsibility of the authors.
2

1

We are also pleased to acknowledge the financial support from the Australian Centre for International Agricultural Research (ACIAR). Again the authors accept responsibility for any errors or omissions.

School of Accounting Economics and Finance, Deakin University (Burwood), 221 Burwood Highway, Burwood, VIC, 3125. Email contacts are: Chris Doucouliagos – douc@deakin.edu.au ; Henry Haszler – henry@deakin.edu.au ; and Phillip Hone – hone@deakin.edu.au

3

Page 1 of 22

Spatial Pricing Efficiency in Fiji’s Municipal Food Markets
1. Introduction Fiji is a small, middle income country. The population of around 850,000 (775,000 at the time of the last census) is spread over 110 of the country’s 332 islands. A large part of the workforce is employed directly in agricultural production which is estimated to make up around 25 per cent of the Fiji Islands GDP. Given the importance of agriculture in the economy, the efficiency of agricultural product markets is a policy issue. Improvements in market efficiency are important as they tend to improve returns to rural producers and enhance food security to urban residents. In efficient markets regional shortages and surpluses of food are ameliorated through the movement of produce from regions of relative abundance to regions of relative scarcity. These movements are facilitated through the communication of price information to food consumers and producers. In Fiji, achievement of this efficiency may be impeded by difficulties in both moving produce within the country – particularly from island to island – and in obtaining timely and accurate information on market prices. Fiji is a relatively urbanised society with the five major regional centres accounting for 77 per cent of the urban population and 36 per cent of the total population (see Table 1). The delivery of food to these centres involves transporting produce from throughout the country. This transportation can be difficult. The road system between the main coastal market and population centres of Viti Levu and Vanua Levu is quite good. However, the road system away from the coast of Viti Levu is limited and poorly developed due to the difficult terrain, high construction costs and shortage of capital. Moreover, the multi island nature of the country means much inter-regional trade must take place via local shipping. This shipping tends to be expensive because port facilities are limited and freight volumes low. Also, the highly fragmented nature of the agricultural sector means many producers have limited access to transport equipment and economies of size in transport are rarely realised at the farm level. Currently information on municipal market prices is spread by both informal means and in weekly evening broadcasts on national radio in Fijian and Hindi. Even so, there are delays in distributing price information and, generally, formal distribution of printed market information appears to be limited. In some of the outer islands there is only limited access to telecommunications facilities and radio and TV– for example in some of the more remote areas electricity generators are run only at night. A further source of inefficiency could be the presence of less than competitive markets. Where the demand and supply elasticities differ between markets, and arbitrage between markets is costly and difficult, firms with market power are likely to engage in price discrimination. The outcome would be relatively higher prices in the least elastic market, increased profits to wholesalers and retailers and reduced community wellbeing. In this paper we report an investigation into the spatial pricing efficiency of Fiji’s major municipal produce markets. We also note some of the policy inferences suggested by the results of the analysis.

Page 2 of 22

Table 1 Population of Urban Centres: Fiji: Census 31 August 1996
Centre Viti Levu Central Division Deuba Korouvou Lami Nausori * Navua Suva * Western Division Ba * Lautoka * Nadi * Rakiraki Sigatoka * Tavua * Vatoukoula 214.6 1.6 0.3 18.9 21.6 4.2 168.0 111.1 14.7 43.3 30.9 4.8 7.9 2.4 7.1 7 2 3 11 8 14 9 Total Population 775.1 100.0% Total Urban Population Total Rural Population 359.5 415.6 46.3% 53.7% 15 18 6 5 12 1 Population ‘000 Size Rank Centre Vanua Levu Northern Division Labasa * Savusavu * Nabouwalu Seaqaqa Ovalau Eastern Division Levuka 3.7 3.7 13 30.1 24.1 5.0 0.6 0.4 4 10 16 17 Population ‘000 Size Rank

Note: The centres include “Unincorporated Towns and Other Urban Centres. The next census will take place in 2007. * Municipal market prices at these centres monitored by the Ministry of Agriculture. Source: Fiji Islands Bureau of Statistics, Key Statistics, December 2005.

2. Conceptual Model A market or set of interconnected markets can be described as informationally and spatially efficient for any set of related commodities if a) the prices at various markets generally move together, b) prices return reasonably quickly to their equilibrium relationships after shocks and c) any persistent differences between prices can be explained by the costs of arbitrage. The arbitrage costs will include the price risks to arbitrageurs associated from shifting produce between markets. Spatial price efficiency depends on the simultaneous existence of informational efficiency – in the sense that at least arbitrageurs are aware of the levels and movements of prices across the market system. Under the weak form of spatial efficiency, price differences between markets will occur even in spatially and informationally efficient markets. It takes time for price information to be recorded, communicated, absorbed and acted on. However, the existence and persistence of price differences exceeding the “efficient margin” will indicate unused arbitrage opportunities which if taken up will result in Pareto improvements in community wellbeing. The potential gains in economic welfare due to efficient arbitrage represent the possible returns to improving the spatial and informational efficiency of those markets, for example through enhanced information flows about market prices. For strong form spatial efficiency, a market system must meet the Law-of-One Price (LOP) such that prices are equal in the various markets so that the allocation of produce between markets equates the net marginal social value of the output across the markets. Where these marginal values diverge, the markets are inefficient. This strong form concept of the LOP and spatial efficiency are illustrated in Figure 1

Page 3 of 22

Figure 1: The Cost of Inefficiency – No Transport Costs
Price

Aggregate Supply ST D2 PT

D1

Aggregate Demand DT

0

Q2

Q1

QT

Quantity

NMV NMV2 P2

NMV1 0 P1

0

Q1

0

Q2

The stylised economy of Figure 1 has one good, x, and two geographically separated markets with separate demand curves D1 and D2. Production is assumed to take place only in Market 1 with the cost of this production reflected in the supply curve ST. In the absence of transport and other arbitrage costs, the efficient free market equilibrium would occur where supply equals the horizontal sum of the two demand curves (DT). The price would be P and the total quantity QT. Prices would be equated across the markets and would equal the cost of producing the last unit of output. Consumption in Markets 1 and 2 would be Q1 and Q2 respectively, which together sum to QT. In the lower panel in Figure 1 the equilibrium condition has been translated into a net marginal value (NMV) space. It is assumed for simplicity that once aggregate supply has been determined at QT, the marginal cost of supplying the two markets is the same and equals PT. The horizontal dimension
Page 4 of 22

reflects the total market supply while the two curves reflect the difference between demand in each market and total supply. The two curves NMV1 and NMV2 are therefore net marginal value curves reflecting the net contribution of an additional unit of x to community wellbeing in each market. The level of consumption in Market 1 is read left to right while the consumption scale in Market 2 is read right to left. The efficient outcome is indicated by the intersection of the two curves. Now imagine there is a divergence in prices between the two markets (P1 and P2 respectively). The shaded triangle in Figure 1 represents the cost to the community of this inefficiency. It shows the gain that could be made from the costless shift of relatively lowly valued units of x from Market 1 to Market 2.

Figure 2: Price Differences – With Transport Costs

NMV NMV 2 NMV’2 NMV1

P2T

0 P1T

0

Q1 Q2

Q’1 Q’2

0

In the absence of transport and information costs, all divergences between regional prices reflect market inefficiencies. However where transport and/or information costs exist, price divergences between regions may be consistent with optimal distribution of product between markets and an efficient market. Consider Figure 2 where we illustrate the case where supply of good x to Market 2 requires transport and other arbitrage costs equal to the vertical distance between NMV2 and NMV2’.

Now the socially efficient level of national output falls with a contraction in supply to Market 2 more than outweighing the expansion in sales to Market 1. Importantly, this socially efficient outcome is characterised by a divergence in prices between the two markets equal to the marginal value of the arbitrage costs 4 . In a distributional sense, the price in Market 1 falls and the price in Market 2 rises. The impact of arbitrage costs is to reduce community wellbeing with producers and consumers in Market 2 sharing the costs involved. Consumers in Market 1 gain from the divergence of supply to their market. They face lower prices and consume more in the presence of the transport costs to the other market. At the same time producers in Market 1 lose from this divergence between prices in the two markets. And of course in Market 2 the distributional impacts are the reverse of those in Market 1.
4

The outcome is socially efficient as long as the arbitrage services are provided competitively and in a technically efficient way. Page 5 of 22

This discussion of the distributional effects of spatial price differences draws attention to some commonalities between the literatures on spatial efficiency and price stabilisation (for the latter see Williams and Wright, 1991). The former literature deals with equalising prices over space while the latter deals with equalising – or stabilising – prices over time and transactions or arbitrage costs are considered in both literatures. Moreover, in both cases the distributional impacts are hidden in the ex post sense so that Gruen’s (1964) classic analysis of the hidden costs and benefits of price stabilisation should be readily reinterpretable in the context of spatial efficiency.

Figure 3: Price Differences due to Transport Costs and Inefficiencies

NMV NMV2

P2 NMV’2
NMV1

P2T

0 P1T

P1
0 0

Q1 Q2

Q’1 Q’2

The simultaneous presence of transport costs and market inefficiencies is also plausible (Figure 3). In this situation any price divergence between markets has two components – the welfare consistent divergence due to transport and other arbitrage costs and the welfare reducing component associated with market inefficiency given by the shaded area in Figure 3. There are a number of price comparisons between markets and products that can be made to analyse the extent of uncovered arbitrage opportunities and hence the informational and spatial efficiency of markets:
•

Multiple commodities in a single market – An example is provided in the first panel of Figure 4 below 5 . Note how closely bunched are the price index numbers of the 22 food items included in the Figure. While the actual prices diverge much more than the levels of their index numbers, there are clearly forces working to maintain some longer term parity between prices of these items. Single commodity group and single market – The example in the middle panel of Figure 4 is for root crops in Suva. Note the longer run stability of the relationship between the price of carrots

•

The data labels have been deliberately stripped from the charts making up Figure 4 and Figure 5 because the purpose of the charts is to illustrate broad relationships between the price series displayed. Unless otherwise indicated, the charts include data for the centres of Labasa, Lautoka, Nadi, Nausori, Savusavu, Sigatoka and Suva. The chart for “Suva: All Produce Items” includes all of the 22 products listed in Tables 3 to 6. Page 6 of 22

5

(the higher price series) and the starchy staples cassava, dalo (two varieties) and Kumala (sweet potato.
•

Single commodity group across multiple markets –The prices for two varieties of eggplant (long and round) in Suva (the generally higher prices) and Labasa provide the example here. In this case prices for the two varieties move closely together in the one market. This is, as would be expected from the result of an earlier analysis of the Rabaul market in Papua New Guinea (Epstein, 1961). However, the price series have diverged and remained apart occasionally, suggesting the possibility of unmet arbitrage opportunities – for example from late 2004 to the end of 2005. Single commodity across multiple markets – Examples of this comparison are given in Figure 5 – prices for Suva highlighted – which shows data for English cabbage, watermelon and lewena (a derivative of yaqona used to make kava). In this case the prices for English cabbage are obviously highly seasonal and quite closely integrated as are the prices for lewena. We would anticipate close integration of prices for lewena since it has a high value to weight ratio. By contrast, the relatively low value to weight ratio for watermelon may be the reason for the apparent maintenance of a price premium for watermelon in Suva. Certainly smallholder farmers are unlikely to be able to effectively arbitrage watermelon prices using local bus transport.

•

•

Within all these types of comparisons market prices should generally be more closely integrated: The more substitutable in consumption and production are the products – so prices of different dalo (taro) varieties can be expected to be more closely related than prices of dalo and kumala and prices of root crops should be more cointegrated than prices of all the traded produce; The higher the value/weight ratio of the products – improves the economics of transportation ; The less perishable the product – increases the possibility of holding over sale of product in period one at one market and moving the product to be sold in period two in the same or at another market.

• •

In this paper we concentrate on comparing prices for single commodities across Fiji’s major municipal produce markets. We test price movements for their consistency with the weak form of the LOP.

3. The Municipal Markets Fijian consumers have access to a wide spectrum of retail services. In the case of food, these range from modern supermarkets selling high priced imported and perishable (frozen) products to individual smallholders hawking their produce from house to house around their own homes and in the nearby towns. Between these extremes, there are municipal produce markets in the major urban centres serving the centres as a whole. These municipal markets are backed up by smaller markets catering to passing traffic or a local neighbourhood. At least in Suva, the central produce market includes stalls selling frozen goods such as meat and fish. The size and throughput of the municipal markets is probably roughly proportional to the urban populations they serve. However, strict proportionality is unlikely to hold because of the smaller neighbourhood markets that are most likely to be found in the largest urban centres. Fiji AgTrade – a group within the Ministry of Agriculture – currently monitors the wholesale and retail food prices at nine municipal markets.

Page 7 of 22

Figure 4: Patterns of Produce Prices: Various Comparisons
Suva: All Produce Item s (Index - Dalo Tausala highlighted)

3

2

1

0 2001-11

2002-06

2003-01

2003-08

2004-03

2004-10

2005-05

2005-12

2006-07

Suva: Root Crops ($F/kg - Dalo Tausala highlighted) 4

3

2

1

0 2001-11

2002-06

2003-01

2003-08

2004-03

2004-10

2005-05

2005-12

2006-07

Suva & Labasa: Eggplant Varietie s ($F/k g - Suva highlighte d)

3

2

1

0

2001-11

2002-06

2003-01

2003-08

2004-03

2004-10

2005-05

2005-12

2006-07

Page 8 of 22

Figure 5: Patterns of Produce Prices: All Markets Comparisons
Englis h Cabbage: All Marke ts ($F/k g - Suva highlighte d) 6 5 4 3 2 1 0 2001-11

2002-06

2003-01

2003-08

2004-03

2004-10

2005-05

2005-12

2006-07

Waterm elon: All Markets ($F/k g - Suva highk ighted) 4

3

2

1

0 2001-11

2002-06

2003-01

2003-08

2004-03

2004-10

2005-05

2005-12

2006-07

Le w ena: All Mark e ts ($F/k g - Suva highlighte d) 30

20

10

0 2001-11

2002-06

2003-01

2003-08

2004-03

2004-10

2005-05

2005-12

2006-07

Page 9 of 22

The nine monitored markets listed in Table 1 can be visualized as stretching in a line from Labasa and Savusavu on the north-western and south-eastern coasts respectively of Vanua Levu then to Viti Levu and to Nausori, which surrounds Suva’s airport somewhat to the north-east of Suva, then Suva and then with the other centres basically spread around the coast between Suva and Lautoka and then north to Ba and further to Tavua. (The map available at http://www.mapsouthpacific.com/fiji/index.html shows the location of most of these markets.) The seven monitored municipal markets on the main island Viti Levu are serve – in the 80 per cent of Fiji’s urban population (1996 census). Labasa and Savusavu on Vanua Levu serve 8 per cent of the urban population. These markets are connected by a transport network comprising road, sea and air transport modes. Two major roads – Kings Road and Queens Road – encircle Viti Levu. There is a ferry service from Savusavu to Viti Levu and Labasa is served by a commercial airport. Although the Fiji Islands cover a wide area of ocean, the area of Fiji’s inhabitable islands is not large – for example it is possible to drive around the main island Viti Levu in one (long) day. That means that the road distances between the major municipal markets on any one island are not great. Distance is relevant in the context of spatial pricing efficiency because of its impact on arbitrage costs. So the travel times and costs shown in Table 2 are arbitrage-relevant measures of distance.

Table 2: Measures of Inter-Market Distance
Lautoka Nadi Sigatoka Suva Nausori

Distance in Bus Travel Time (hours/minutes) Nausori Suva Sigatoka Nadi Lautoka 5.:50 4:50 2:00 0:30 0:00 Distance in Bus Fare ($F) Nausori Suva Sigatoka Nadi Lautoka 14.50 12.95 5.70 1.70 0.00 13.45 11.90 4.50 0.00 9.30 7.75 0.00 1.55 0.00 0.00 4:35 4:00 1:30 0:00 3:00 2:35 0:00 0:35 0:00 0:00

Note: Distances and fares to and from Nadi refer to Nadi Airport. Details for Nausori are by Sunbeam Transport via Kings Road. Other details for Pacific Transport via Queens Road. Sources: http://www.bulafiji.com/index.cfm?go=main.root, Accessed 26/01/2007.

The arbitrage between markets might be provided by professional “middlemen” or even by farmers just deciding to sell their produce at a different market. So the bus fares between the various markets plus some allowance for any additional costs of traveling with a bundle of produce should be an indicator of the transport margin. Professional arbitrageurs might transport produce at a lower unit cost. Of course someone moving produce between selling centres might need to give up a preferred site within a
Page 10 of 22

“home” market and instead sell from a peripheral site at the alternative market. Depending on the monetary and other benefits of the “home” site, the size of the economically efficient price margin might exceed the direct transport costs. Any arbitrage cost arising from differences in selling location within the home and distant market is less likely to be an issue for professional arbitrageurs who, in principle, could hold permanent stalls at preferred sites in more than one market. The time and fare “distances” between the various markets are not large in an absolute sense. As a result, given good information about prices any persistent price differences between markets should generally be small and any divergences should be quickly bid away. The municipal markets operate as both wholesale and retail markets. 6 Wholesaling occurs very early in the morning after which the markets shift into retailing mode. On Thursdays and Fridays Ministry of Agriculture officials visit the nine monitored markets to record both wholesale and retail prices. Nine markets are now regularly monitored. The range of produce covered has been extended and standardized to an extent. Currently prices are collected for around 55 items– including coconuts and other fruits, traditional Pacific Islands root crops and carrots, potatoes, a wide range of vegetables, spices and – last but in Fiji certainly not least – lewena and waka, forms of yaqona used to make kava, the national drink. Unfortunately fish, meat and other livestock products are not included in the range of produce covered. Based on the authors’ visits to the Suva and Labasa markets, the most important omissions are fish and eggs. 4. Analytical framework According to the weak-form of the LOP, international arbitrage results in an equilibrium such that domestic and international prices differ by no more than the cost of arbitrage, i.e. the marginal profit from arbitrage is zero. 7 And in a domestic setting regional prices would also differ by no more than the cost of arbitrage. Spatial efficiency requires that LOP holds. There is a large literature on the ways that LOP is analysed. Empirical investigations of LOP can be divided roughly into two groups. The first group uses time series techniques such as unit root tests, cointegration and VAR models to investigate spatial efficiency and price transmission processes. Examples of these approaches include: Gonzalez-Rivera and Helfand (2001); Thompson et al. (2002); Rashid (2004); Rapsomanikis et al. (2003-4); and Vollrath and Hallahan (2006). A second group uses maximum likelihood estimation to distinguish between integration and equilibrium. Examples of this literature include Tostao and Brorsen (2005) and Moser et al. (2006). While the former group uses only price data, the later typically also includes trade flow data. (For reviews and applications see Barrett 2001 and Barrett and Li 2002). There is, as yet, no theoretical consensus on the best way to investigate spatial efficiency. Our preference is for the first group mainly because of the lack of regional trade data for the agricultural commodities with which we are concerned.

6 7

The following information is based on personal communications from Paul Waqa, Fiji AgTrade, Ministry of Agriculture, Suva. This is derived from the Samuelson-Takayama-Judge spatial equilibrium models. Page 11 of 22

4.1 Law-of-One-Price Our analytical approach involves two steps. First, we commence with analysis of the LOP. Second, we estimate various systems of equations designed to capture the short-run dynamics of prices as they respond to shocks and change to reestablish long-run equilibrium. Generally the existence of LOP is tested through unit root tests on the difference between some given price of interest and a reference price, such as some benchmark world price. For details see Thompson et al. (2002). Note that finding stationarity in this difference is consistent with LOP but does not prove it. Conversely, non-stationarity of the price difference need not mean lack of spatial efficiency, as it can result from non-stationarity of transaction costs, especially transport costs. 8 Since we are dealing with a single country over a relative short period (2001 – 2006), we can safely assume that non-stationarity of transport costs will not be the reason for any failure to find evidence supporting LOP. We include in the system only those regions where unit root tests support LOP. Suva is our benchmark. Hence, we first perform unit root tests on the difference between the price of commodity j in region i and the price of commodity j in Suva. The unit root test results are available from the authors. 9 4.2 Short-run dynamics We provide two sets of estimates of the short-run dynamics. First, we follow Thompson et al. (2002) and use iterative SUR to estimate the following two-equation cointegration system (SURECM):

Δpit= λ0i(pit - pst) + ΣβΔpit-k + ΣγΔpst-k + u1it Δpst= λ1i(pit - pst) + ΣθΔpit-k + ΣφΔpst-k + u2st

(S1)

where Δ denotes the first difference filter, p denotes price, the subscripts i and s index respectively the ith region and Suva at period t, λ is the adjustment coefficient estimating the speed at which prices revert to their long-run equilibrium, k is the number of lags 10 and u1 and u2 are Gaussian error terms. We call this system the Short-Run Dynamics Single Commodity Arbitrage model. The model has been applied previously to the world wheat market (Thompson et al. 2002). The model focuses on the shortrun responses (the Δpit and Δpst) to a disturbance in the initial long-run equilibrium. It is important to note that this system considers only the own region price response and ignores cross-region responses. Note also that a separate equation is included for each pit for which LOP was found to hold, with a minimum of 1 regional price and a maximum of 5 (one each for Labasa, Savusavu, Nadi, Savusavu and Lautoka). 11

8 9

That is, LOP may actually hold but non-stationarity in transport costs gives the appearance of failure of LOP in the long-run.

We used both Augmented Dickey-Fuller and Phillips-Perron tests and used the 5% level of significance as our critical value. The ADF and PP tests mostly led to the same conclusion. Where the ADF and PP led to conflicting inferences, we relied on the PP for inference as it is a more robust test. We report the results using two lags. The results are robust to longer lags results.

10

Hence, at minimum S1 involves the estimation by SUR of two equations (one for region i and one for Suva) and a maximum of 10 (one for each of the five regions and five for Suva). Page 12 of 22

11

Our second approach is to follow Gonzalez-Rivera and Helfand (2001) who consider multiple cointegrating vectors – in addition to the own region price response to a disturbance in the long-run equilibrium, they also explore the own region response to a disturbance in an other region. We call this the With Cross-Region Effects model:

Δpit= λ0i(pit - pst) + λ0j(pjt - pst) + ΣβΔpit-k + ΣβΔpjt-k + ΣγΔpst-k + u1it Δpst= λ1i(pit - pst) + λ1j(pjt - pst) +ΣθΔpit-k + ΣθΔpjt-k + ΣφΔpst-k + u2st

(S2)

where the subscript j indexes the price of region j and where the number of cross-region effects varies from crop to crop according to the availability of data and the confirmation of LOP. While it is a more general model, there are several unresolved empirical issues associated with the Gonzalez-Rivera and Helfand model, especially their sequential process for ascertaining the regions that define a market 12 , and the interpretation of the cross-region responses. In their case study (Brazilian rice) and in the application of Rashid (2004) to maize, the sequential process worked well. However, we found that for many crops the starting point made a difference and that at times we ended up with conflicting configurations. Consequentially, we decided to include all regions in which LOP held as our starting definition of the market and to estimate ECMs using all regional data. This approach has the benefit of also including the same set of regions as does S1. 13 Our primary interest here is on the adjustment coefficient (λ). At one end, a zero coefficient means that there is no price adjustment between the two regions and that there is no tendency for a market to return to equilibrium. This indicates a spatially inefficient market. Alternatively, if the coefficient equals 1, then there is complete and rapid adjustment back to equilibrium. Hence, higher values of λ are consistent with higher levels of spatial efficiency. S1 and S2 are estimated separately for each of the 22 crops for which we have sufficient data. We measure spatial efficiency for only those crops whose prices were measured in terms of kilos or per dozen (coconuts). Prices for several crops, for example bele, Chinese cabbage, cowpea prices are quoted per bundle or per heap. The definitions of bundles and heaps change over time and across markets and there is anecdotal evidence that traders engage in “cost leveling”, varying bundle sizes to disguise underlying changes in prices 14 . Of the 22 crops, there are often missing price entries. We estimate S1 and S2 using the recorded data and interpolated data. We present the results using interpolated data, as there is no real difference between the two results.

In brief, the sequential process proceeds as follows. Commence with a core m number of municipal markets and test the number of cointegrating vectors. Then sequentially add a municipal market as long as there are m-1 cointegrating vectors. In their paper Gonzalez-Rivera and Helfand (2001, p. 579) state that: “Future research should study the econometric problems of sequential exclusion”. We are unaware of such work undertaken, but have found that the sequential exclusion process to be problematic for our dataset. Both the Thomspon et al. (2002) and the Gonzalez-Rivera and Helfand (2001) models have been developed and applied to single commodities for different regions. For example, in their empirical analysis, Rapsomanikis, Hallam and Conforti (2003-4) treated every agriculture commodity as a distinct item. An extension to this approach would be to estimate arbitrage in the multiple commodity case. Our Fiji data covers 22 different agriculture commodities. Arbitrage activities need not be limited to a single agricultural commodity, but may be spread across a range of commodities. Ignoring arbitrage opportunities may lead to misspecification. In order to accommodate this notion empirically, the systems S1 and S2 can be estimated to a particular class of agricultural commodities. This does however impose a large burden in terms of degrees of freedom. 14 On this score at least, Fiji’s municipal market traders have nothing to learn from the marketing practices followed in economically more advanced countries Page 13 of 22
13

12

5. Results The short-run dynamics captured by the SURECM for five different fruits are presented in Table 3. The estimates associated with the system S1 – the Short-Run Dynamics Single Commodity Arbitrage model – are shown in the top panel. Estimates of the Gonzalez-Rivera and Helfand model – the With Cross-Region Effects model – are presented in the bottom panel. For the sake of brevity, the crossregion price responses are not reported here but are available from the authors. 15 The first column for each set presents the own region price responses. For example, the coefficient for Labasa in column 1 (-0.39) is the adjustment coefficient in the ECM showing the speed of adjustment of Labasa prices to a disturbance from equilibrium. The coefficient is negative indicating that if Labasa prices increase compared to their long-run equilibrium with Suva, then Labasa prices will fall in the next period. Approximately 39 per cent of the adjustment is born in the one period, so t it will take about 2.5 months for the adjustment to complete. However, the estimate in the bottom Panel suggests a much faster rate of adjustment with approximately 81 per cent of the adjustment born in the one period, so full adjustment takes only a little over one month. The coefficients in the second column for each set show the response of Suva. In the case of bananas, all the coefficients are small in magnitude and are not statistically significantly different from zero. 16 The results for root crops, vegetables and lewena, waka and coconuts are shown in Tables 3 to 6.

The differences between the estimates in the top and lower panels of Tables 3 to 6 result from the inclusion of cross-region price responses which lead to five major differences in the results:
•

•

•

• •

There are more statistically significant adjustment coefficients in the With Cross-Region Effects model – 81 statistically significant (and correctly signed) own-region coefficients compared to only 62 from Panel A. λ is generally larger in S2. Comparing the median adjustment coefficients for the two models we find: for fruit the median is -0.24 compared to -0.47; for root crops it is -0.22 compared to -0.47; for vegetables it is -0.19 compared to -0.49; and for kava (ie lewena and waka) is it -0.09 compared to -0.48. The vegetables and kava markets appears to be more spatially efficient in system S2 and grossly inefficient in the S1 model. In four cases (pineapples, ginger, kumala and cucumbers), the coefficient for the own-region response for Nadi is no longer statistically significant implying that for these crops, Nadi and Suva are not integrated markets. In most cases, the goodness of fit of the S2 model is higher than for S1. There are no cases of instability in the S1 model but seven cases of apparent instability in the full model (two of these cases emerge in the market for ginger).

Possible instability is indicated where the adjustment coefficients have the wrong sign, implying that a disturbance from equilibrium triggers responses that drive further away from equilibrium rather than towards it. However, in this case it may be premature to conclude that these do represent instability as the sign differences could arise also from either trade flows (for which we currently have no data) or from cross-price effects.
15

Note that prior testing revealed that the best modelling approach was to include an intercept in the ECM but no trend. Details of the cointegrating equations are available from the authors.

We expect a positive coefficient for Suva, in order to ensure stability of the system. A positive coefficient means that, for example, as prices in Labasa rise and move away from their long-run equilibrium with Suva, prices in Suva rise to reestablish equilibrium. For the lower panel a negative coefficient for Suva is possible, depending on the cross-region responses. Page 14 of 22

16

Table 3 Adjustment coefficients, fruits, Fiji, 2001-2006
Bananas Own Region Price (1) Bananas Suva Lemons Own Region Price (3) Lemons Suva Pineapple Own Pineapple Suva Pawpaw Own Region Price (7) Pawpaw Suva Watermelon Own Region Price (9) Watermelon Suva

(2)

(4)

(5)

(6)

(8)

(10)

Short-Run Dynamics Single Commodity Arbitrage Labasa Savusavu Nadi Sigatoka Lautoka Adjusted R
2

-0.39*** -0.20** -0.42*** -0.32*** -0.41*** 0.06 to 0.41

0.02 0.02 0.01 0.02 0.01

na na 2.31 -0.03 -0.33*** 0.00 to 0.43

na na 0.02 0.12* 0.03

-0.26*** -0.24** -0.26*** -0.37*** -0.16*** 0.00 to 0.35

0.06 0.08 0.01 -0.01 0.05

-0.13* -0.30** -0.38*** -0.18** -0.18*** 0.05 to 0.17

0.02 -0.01 0.01 0.01 0.01

-0.02 -0.10** -0.79*** -0.05 -0.16* 0.00 to 0.27

-0.01 -0.01 0.02 0.01 0.00

With Cross-Region Effects Labasa Savusavu Nadi Sigatoka Lautoka Adjusted R
2

-0.81** -0.44** -0.43*** -0.35*** -0.77*** 0.14 to 0.43

-0.06 0.03 0.01 0.17 -0.11

na na -0.17** -1.16*** -0.46*** 0.06 to 0.54

na na -0.07 0.29 0.10

-0.43*** -1.09*** -0.09 -0.58*** -0.72*** 0.03 to 0.54

0.22 -0.03 0.01 -0.21 -0.34

-0.47*** -0.62*** -0.57*** -0.46*** -0.41*** 0.17 to 0.30

-0.09 0.04 -0.08 0.03 -0.19

-0.52*** -0.29** -0.80*** -0.48*** -0.31*** 0.09 to 0.39

-0.01 -0.01 0.02 0.01 0.00

Models include two lags in both the changes in the own region and Suva prices. na means insufficient observations. Cross-regional effects jointly statistically significant but not reported in the table. Iterative SUR estimation.

Page 15 of 22

Table 4 Adjustment coefficients, root crops, Fiji, 2001-2006
Dalo Tausala Own Region Price (1) Dalo Tausala Suva Dalo Others Own Region Price (3) Dalo Others Suva Cassava Own Region Price (5) (4) (6) Cassava Suva Ginger Own Region Price (7) (8) Ginger Suva Kumala Own Region Price (9) (10) Kumala Suva

(2)

Short-Run Dynamics Single Commodity Arbitrage Labasa Savusavu Nadi Sigatoka Lautoka Adjusted R
2

-0.03 -0.38*** na na -0.17* 0.08 to 0.31

0.02 0.02 na na 0.02

-0.09 -0.04 -0.21** -0.09 -0.36*** 0.00 to 0.33

-0.01 0.05 0.01 0.02 -0.01

-0.26*** -0.32*** -0.22** -0.23** -0.21** 0.02 to 0.43

0.05 0.05 0.04 0.01 0.01

-0.31*** -0.07* -0.22*** -0.17** -0.33*** 0.12 to 0.27

0.01 0.01 0.00 0.00 -0.01

-0.18** na na -0.22* na 0.26 to 0.40

0.06 na na 0.03 na

With Cross-Region Effects Labasa Savusavu Nadi Sigatoka Lautoka Adjusted R
2

-0.45*** -0.73*** na na -0.30*** 0.10 to 0.33

0.04 -0.06 na na 0.29***

-0.38*** -0.43*** -0.20** -0.68*** -0.84*** 0.09 to 0.36

-0.29** 0.08 -0.02 0.35*** 0.09*

-0.52*** -0.29** -0.80** -0.48*** -0.31*** 0.11 to 0.40

0.03 -0.02 0.05 0.07 0.12***

-0.78*** -0.64*** -0.09 -0.71*** -0.78*** 0.00 to 0.47

-0.17 -0.44** -0.10** 0.07 -0.05

-0.25* na na 0.06 na

-0.45 na na 0.02 na

Models include two lags in both the changes in the own region and Suva prices. “na” means insufficient observations. Cross-regional effects jointly statistically significant but not reported in the table. Iterative SUR estimation.

Page 16 of 22

Table 5 Adjustment Coefficients, Vegetables, Fiji, 2001 – 2006
Carrots Own Region Price (1) Carrots Suva Bongo Chilies Own Region Price (3) Bongo Chilies Suva Cucumber Own Region Price (5) Cucumber Suva Eggplant (Long) Own Region Price (7) Eggplant (Long) Suva Eggplant (Round) Own Region Price (9) Eggplant (Round) Suva Pumpkin Own Region Price (11) Pumpkin Suva

(2)

(4)

(6)

(8)

(10)

(12)

Short-Run Dynamics Single Commodity Arbitrage Labasa Savusavu Nadi Sigatoka Lautoka Adjusted R
2

Na -0.02 -0.30*** -0.14 -0.17*** 0.04 to 0.29

Na 0.00 0.02 0.06* 0.01

-0.06 -0.14* -0.42*** -0.05 na 0.00 to 0.17

0.04 0.02 0.03 0.02 na

-0.23** -0.60*** -0.24** -0.06 -0.18* 0.00 to 0.37

0.02 0.02 0.01 0.02 0.01

-0.09 Na -0.09 -0.13 LPF 0.07 to 0.21

0.07 Na 0.06 0.27* LPF

-0.13 -0.11 na -0.18* -0.11* 0.10 to 0.32

0.01 0.02 na 0.03 0.01

-0.34*** -0.64*** -0.25** LPF -0.30 0.00 to 0.35

0.01 0.02 0.01 LPF 0.01

With Cross-Region Effects Labasa Savusavu Nadi Sigatoka Lautoka Adjusted R
2

Na -0.89*** -0.34*** -0.45*** -0.41*** 0.18 to 0.43

Na 0.04 0.09* -0.06 -0.08

-0.19* -0.55*** -0.51*** 0.22*** na 0.16 to 0.46

0.29 -0.36** 0.11 0.24 na

-0.63*** -0.88*** -0.19 -0.95*** -0.26*** 0.06 to 0.54

0.08 -0.03 -0.03 -0.12 -0.10*

-0.55*** Na -0.10** -0.50*** LPF 0.08 to 0.24

0.11** Na 0.04 0.05 LPF

-0.44*** -0.81*** na -0.64*** -0.55*** 0.20 to 0.44

0.10* -0.17 na 0.15 -0.18

-0.59*** -0.52*** -0.65*** LPF -0.53*** 0.25 to 0.30

0.14 0.16** 0.01 LPF -0.06

Models include two lags in both the changes in the own region and Suva prices. na means insufficient observations. LPF means LOP does not hold. Cross-regional effects jointly statistically significant but not reported in the table. Iterative SUR estimation.

Page 17 of 22

Table 6 Adjustment coefficients, vegetables and other agriculture products, Fiji, 2001 – 2006.
English Cabbage Own Region Price (1) English Cabbage Suva French Beans Own Region Price (3) French Beans Suva Okra Own Region Price (5) Okra Suva Lewena Own Region Price (7) Lewena Suva Waka Own Region Price (9) Waka Suva Coconuts Own Region Price (11) Coconuts - Suva

(2)

(4)

(6)

(8)

(10)

(12)

Short-Run Dynamics Single Commodity Arbitrage Labasa Savusavu Nadi Sigatoka Lautoka Adjusted R
2

-0.44*** -0.30** -0.32*** -0.79*** -0.67*** 0.00 to 0.14

0.15** 0.10 0.03 0.13 0.07

-0.09 na na -0.02 na 0.00 to 0.02

0.08** na na 0.08* na

-0.34** -0.20*** -0.07 -0.09* -0.07 0.04 to 0.21

0.10* 0.01 0.03 0.04 0.04

0.02 -0.11 LPF LPF LPF 0.11 to 0.17

0.04 0.01 LPF LPF LPF

-0.39*** -0.15** -0.05 -0.12** -0.03 0 .00 to 0.15

0.05 0.00 0.03 -0.01 0.00

-0.07* -0.17 -0.11* -0.24** na 0.00 to 0.24

0.00 -0.01 0.00 0.02 na

With Cross-Region Effects Labasa Savusavu Nadi Sigatoka Lautoka Adjusted R
2

-0.54*** -0.26** -0.41*** -0.68*** -0.79*** 0.11 to 0.26

0.17 0.17** -0.05 0.56*** -0.19

-1.15*** na na -0.11** na 0.03 to 0.44

0.15 na na -0.30*** na

-0.67*** -0.35*** -0.29*** -0.62*** -0.07 0.16 to 0.45

0.21 -0.27** -0.08 0.10 0.13

-0.29*** -0.36*** LPF LPF LPF 0.10 to 0.23

0.34*** 0.11* LPF LPF LPF

-0.52*** -0.68*** -0.41*** -0.64*** -0.45*** 0.20 to 0.33

0.22* 0.07 -0.02 -0.15 0.15

-0.23** -0.95*** -0.41*** -0.33*** na 0.06 to 0.45

0.03 -0.05 0.11 0.25** na

Models include two lags in both the changes in the own region and Suva prices. na means insufficient observations. LPF means LOP does not hold. Cross-regional effects jointly statistically significant but not reported in the table. Iterative SUR estimation.

Page 18 of 22

6. Determinants of spatial efficiency Most studies on spatial efficiency ignore the equally important issue of the determinants of spatial efficiency and, hence, the important policy issue of how to improve it. In this section we offer a preliminary exploration of some of the determinants of spatial efficiency. The dependent variable is made up of the absolute value of the estimates of λ. The explanatory variables are: (a) three dummy variables for crop differences with root crops as the benchmark, with a dummy variable for fruit a dummy variable for vegetables and a dummy variable for yaqona, (b) the distance in miles between a region and Suva and (c) a set of regional dummies. The results are presented in Table 7. Columns 1 and 3 present the results without regional dummies, using the estimated λ from the models without and with cross-regional effects, respectively. Columns 2 and 4 present the equivalent results when regional dummies are added (with Nadi as the base). In all cases, distance has a positive coefficient and is statistically significant. This is surprising and suggests that spatial efficiency is greatest amongst the more distant regions. It is consistent with agents in more remote regions taking greater interest in prices in Suva than in their generally nearer markets. In this case the dominance of the Suva market may simply reflect the size of the various markets. Recall that Suva’s population is roughly four times the population of Lautoka, the next largest urban centre. Perhaps there are economies of scale in arbitrage so it is just more profitable for professional arbitrageurs to concentrate on Suva? Table 7 Determinants of spatial efficiency (dependent variable = absolute value of λ)
Variables Without Cross-Region Effects (1) Distance Fruit Vegetables Yagona Regional dummies Adjusted R-squared N 0.004 (4.88)*** 0.127 (3.26)*** 0.062 (1.12) -0.057 (-1.11) NO 0.16 90 (2) 0.008 (4.99)*** 0.060 (1.53) 0.000 (0.00) -0.107 (-2.92)*** YES 0.21 90 With Cross-Region Effects (3) 0.012 (13.40)*** 0.119 (3.69)*** 0.060 (1.61) 0.043 (0.50) NO 0.21 90 (4) 0.013 (5.70)*** 0.060 (2.77)*** 0.001 (0.04) 0.010 (0.17) YES 0.36 90

Figures in brackets report t-statistics using robust standard errors.

Fruit has a positive sign and is mostly statistically significant indicating that spatial efficiency is greater in the fruit market than in the other markets. This suggests that the returns to the more effective dissemination of price information may be lower for fruits than other crops. The generally superior spatial efficiency of the fruit market may reflect the fact that many Fijians with access to some land will grow at least some of their own bananas and pawpaws. Once planted, these fruits require little attention compared to other crops. It may be therefore that arbitrage of market prices through subsistence household production is more effective for the easily grown fruits than other crops.

Page 19 of 22

7. Summary of Findings Spatial efficiency is one important component of an efficient market system in which arbitrage plays a critical role in matching demand and supply between regions. We used monthly retail level price data for 22 locally produced crops to assess the extent of spatial efficiency across Fiji’s major municipal produce markets and to explore the determinants of spatial efficiency. An extension to this paper could explore different definitions of the market and could also consider additional dimensions of pricing efficiency such as across a commodity group – for example all root crops – within the one market and also across substitute products from different food groups Meanwhile, we see five results emerging from our analysis. • First, the weak form of the Law of One Price seems to hold at the retail level for most crops across Fiji’s major municipal produce markets. That means most markets appear to be generally integrated and so our results appear to be consistent with spatial efficiency – defined narrowly without allowance for arbitrage costs – among Fiji’s regional food markets. This is an important – even if preliminary – result. It means that Fiji may be near to maximizing the economic value – to consumers and farmers – of its produce markets. • Second, it appears, nevertheless, that some markets are not spatially efficient and further research is needed to identify the reasons for this. Particularly problematic is the market for lewena where LOP does not hold. 17 Supply factors could be important. For example, some part of the yaqona crop acts a “savings bank in the ground” because of the considerable flexibility in harvesting the crop. That is, farmers harvest some of their yaqona to meet “social obligations” such as school fees and church donations. So market prices may play little role in the harvesting and marketing decisions for yaqona for some producers and at some times of the year. • Third, while we found most markets to be spatially efficient, the speed of adjustment remains a concern. The estimates derived from the With Cross-Region Effects model yield more rapid adjustment. However, for most regions and for most crops it still takes prices around two months to reestablish long-run equilibrium. For markets that operate daily this is relatively slow response to price shocks indicating incomplete spatial efficiency. • Fourth, in most cases, the short-run adjustment dynamics do not involve Suva. That is, regional price shocks are self-adjusting. This is probably due to the size of the Suva market. A disturbance in the long-run equilibrium price between region i and Suva, sees prices in region i bear the adjustment. This is particularly so in the case of fruit, where there is no instance of Suva prices adjusting. For some crops, the price response of Suva suggests instability when the with cross-region effects model is used. Analysis of trade flow data is needed to confirm this. • Finally, the often slow adjustment back to equilibrium suggests there may be grounds for policy action to improve the dissemination of information about market prices amongst both consumers and farmers. Currently, price reports from the individual centres are faxed to Fiji AgTrade at the Ministry’s head office in Suva and combined to generate weekly average prices for all centres. The information is published in the Ministry’s Market Watch report. This is distributed to key personnel within the Ministry and to a number of agricultural exporters and commercial farmers. In all, 33 copies of the Market Watch report are distributed
17

Moreover, lack of data prevented analysis for some crops, particularly Kumala, Dalo and French beans. Page 20 of 22

– of which 23 copies are circulated within the Ministry. The Ministry has recently significantly upgraded its web site (www.agriculture.gov.fj) from a static to a live information site that permits downloading of reports and other information. It should eventually be possible to easily download the regular price reports from the new site. Even with the change to the web site, the price information is likely to remain severely underutilised since most Fijians do not have access to the internet, certainly not at home and not in the rural areas amongst most of the farming community. Given all the effort involved in gathering and compiling the information there appears to be a compelling case for maximizing its economic value by ensuring it is made much more widely available. There are many ways that price information could be more widely disseminated. For example, newspapers could be encouraged to carry market reports as a regular feature in their Friday and Saturday editions. And – as many farmers take a break around noon to escape the midday sun – local radio stations might broadcast market summaries as part of their midday news services. The decision to introduce these sorts of information services would obviously deserve a proper benefitcost analysis. This evaluation should include an analysis of the hidden impacts on the economic surplus of consumers and producers under conditions where prices are equalized up to the level of arbitrage costs. Nevertheless, it seems a priori that the benefits could be quite substantial while the costs should be relatively low. After all, the market prices are collected and collated already and the newspapers and radio stations may both see the dissemination of this price information as an opportunity to better serve their clients. References Barrett, C.B. (2001). Measuring Integration and Efficiency in International Agricultural Markets, Review of Agricultural Economics, 23(1): 19-32. Barrett, C.B. and Li, J. R. (2002). Distinguishing Between Equilibrium and Integration in Spatial Price Analysis, American Journal of Agricultural Economics, 84(2): 292-307. Epstein, T.S. (1961). A Study of Rabaul Market, Australian Journal of Agricultural Economics, 5: 49-66. Gonzalez-Rivera, G. and Helfand, S.M. (2001). The Extent, Pattern, and Degree of Market Integration: A multivariate approach for the Brazilian rice market, American Journal of Agricultural Economics, 83(3): 576-92. Gruen, F. (1964). Some Hidden Gains and Losses in a Wool Reserve Scheme, Australian Journal of Agricultural Economics, 8(2), 181-8. Moser, C., Barrett, C.B. and Minten, B. (2006). Spatial Integration at Multiple Scales: Rice markets in Madagascar. Manuscript. Rapsomanikis, G., Hallam, D. and Conforti, P. (2003-4). Market Integration and Price Transmission in Selected Food and Cash Crop Markets of Developing Countries: Review and Applications, Commodity Market Review, 2003-2004. Rashid, S. (2004). Spatial Integration of Maize Markets in Post-Liberated Uganda, MTID Discussion Paper No. 71, IFPRI. Thompson, S.R., Sul, D. and Bohl, M.T. (2002). Spatial Market Efficiency and Policy Regime Change: Seemingly Unrelated Error Correction Model Estimation, American Journal of Agricultural Economics, 84(4): 1042-53.
Page 21 of 22

Tostao, E. and Brorsen, B.W. (2005). Spatial Efficiency in Mozambique’s Post-Reform Maize Markets, Agricultural Economics, 33(2):205-14. Vollrath, T. and Hallahan, C. (2006). Testing the Integration of U.S.-Canadian Meat and Livestock Markets, Canadian Journal of Agricultural Economics, 54: 55-79. Williams, J. C. and Wright, B.D. (1991). Storage and Commodity Markets, Cambridge University Press, Cambridge.

Page 22 of 22


				
DOCUMENT INFO
Shared By:
Stats:
views:149
posted:5/23/2009
language:English
pages:22
Description: The Fiji Islands economy is widely held to be constrained by a lack of market and transport infrastructure. Such infrastructure problems can reduce the efficiency of markets and result in wider than optimal disparities in food prices, both between markets and over time. In this paper we use monthly price data for 22 locally produced crops from Fiji’s major municipal markets, for the period 2001-2006, to identify the extent of the spatial pricing efficiency between these markets. Short-run price responses are estimated through SURECM. Relative price movements between markets indicate the potential for profitable arbitrage which will maximise the economic value of food to the Fijian economy. The arbitrageurs might be middlemen or individual farmers shifting their own produce between selling centres. We find that the municipal markets are generally spatially price efficient in the technical sense that prices return to their long-run relationships after a shock. But after a disturbance prices generally still take some two months to return to their equilibrium with the reference market, the capital Suva. This suggests there is scope for policy action to improve the spatial pricing efficiency of Fiji’s food markets. An enhanced market price reporting system is one option that deserves consideration.
JFEI Nicol JFEI Nicol Technology http://www.techfoxin.com
About