PDPE Market Analysis Tool: Market Integration

Markets are important determinants of food availability and food access. The extent to
which markets make food available and keep prices stable depends on whether
markets are integrated with each other. Integrated markets can be defined as markets
in which prices for comparable goods do not behave independently. If markets are well
integrated, it can be assumed that market forces are working properly, meaning that
price changes in one location are consistently related to price changes in other
locations and market agents are able to interact between different markets. If markets
are integrated, food will flow from surplus to deficit areas - and imports will flow from
port and border areas into the hinterland. High prices in deficit areas provide the
incentive to traders to bring food from surplus to deficit areas, making food available.
As a result of these flows, prices should decline in deficit areas, making food more
accessible to households.

What insights can this tool provide?

Prices usually give important indications on whether markets are integrated. Markets
are integrated if prices among different locations move in similar patterns, given that
the differences between prices is explained by the transfer and transaction costs as
food flows between the locations. Otherwise markets are segmented. This could, for
example, be a result of prohibitive transaction costs related to poor infrastructure in
remote areas or damaged roads or bridges because of a disaster.

When markets are integrated, food flows among regions and prices fluctuate less,
enhancing food security. Knowledge about market integration is, therefore, essential
for programming. The degree of market integration will inform the analysis of food
security, appropriate responses to a crisis, the extent of possible negative effects of
food aid and local procurement possibilities. Here are some examples:
Where markets are poorly integrated - and prices more volatile - vulnerable
households will experience more often high prices;
Regarding response options, cash transfers can be a response option if markets
are integrated, food is available and prices are relatively stable;
Local procurement is also highly dependent on market integration. WFP might
be able to procure locally with no detrimental effects on the market if food
flows from other regions1; and
In case of an emergency, the degree of market integration affects the
estimates for the required amount of food aid because traders might be able to
meet part of the food needs of the disaster-affected people.

How to analyse, interpret and use the data?

Analyzing market integration is done by comparing prices in different locations. The
chart below provides a framework to analyse prices. This framework can be discussed
step by step as follows:
• Step 1: Assess whether prices move in tandem or not. One could calculate
simple correlation coefficients or plot price series in a graph to check co-
movement. If prices co-move, markets may be integrated. However, high
correlation coefficients or price co-movement can be a result of other factors,

1 WFP could also procure locally in areas where a lack of market integration prevents available surpluses being moved out to deficit
areas.

1
 like a steady increase in all prices, rather than market integration. Checking for
outliers - caused by a specific phenomenon in one market e.g. - and stability of
price series overtime is also needed.
• Step 2: Analyse whether prices converge by calculating the average of price
differences between markets in a given period. Convergent markets are
integrated markets where prices are at the same level. A zero average suggests
absolute convergence, indicating that the markets may be well integrated. A
non-zero mean points to relative convergence, indicating that prices move in
tandem, but that there are price differences as a result of transaction costs.
• Step 3: Compare spatial price differences with transaction costs. If transaction
costs are higher than the price differences between two markets, it is likely
there is no incentive for traders to move food between these markets at a
period of time. Otherwise, the two markets are likely to be integrated.
• Step 4: Cross-check with traders if there is any reason why they might not move
food. Among other reasons, it is worth analysing risk factors such as seasonal
food availability and transport hindrances, changes in policy, security as
indicated in the big “cloud of the framework”.
• Step 5: Implications can be drawn for programming and response options. If the
above steps point towards market integration, food is available in markets and
prices are stable, cash transfers may be an option. If markets are integrated,
the effects of food aid on markets are also likely to be small and temporary.

Graph 1: Market Integration Framework

Evidence in favour of convergent, well-integrated markets

Yes 1 Comovement in Prices? No

Simple correlation
Potential Market coefficients to
Tentatively no
Integration assess the degree Market integration
of integration

Potential determinants of market integration:

Stability of price differences over the period - Price differences: seasonality, transport cost, transaction costs
(taxes, tariffs, subsidies, etc.), prices of substitutes or complements;
- Lack of connection: physical market access, trade and import flows
(formal and/or cross-border); and
Convergence? - Market functioning: demand aspects (purchasing power e.g.),
competition/collusion, insecurity …
Non-functioning markets
Zero mean difference Non zero mean difference Non-availability of food
2 Price volatility

Absolute Relative
convergence Convergence

Transaction costs
Food available through commercial channels Market
Low inflationary pressures or barriers 3
segmentation?

Policy implications/Response Food aid interventions

Cash options? Options assuming an in-depth +
Market Analysis Market support
Check for positive 5
4
trade flows
Food aid?
Food aid is still an Magnitude and
eligible option duration of price
instability at stake

2
Example: Grain wholesale prices in Ethiopia (2000-2006)
To illustrate the issue of market integration, let us consider the case of Ethiopia’s
wholesale prices for teff, wheat and maize in 3 main regions/cities: Addis Ababa,
Mekelle (Tigray) and Oromya. The price data used are extracted from the Ethiopian
Grain Trade Entreprise (EGTE) database and the estimate for transportation costs
between Mekelle and Addis Ababa from the Tigray Agricultural Marketing Promotion
Agency (TAMPA) bulletin. The steps and calculations are also in an attached Excel
spreadsheet: marktint.xls.

• Step 1: A rapid look at the price movements in the graph below for wheat
shows how Tigray price evolution is peculiar. When prices in Addis Ababa and
Oromiya went up swiftly (e.g. in July 2001 and September 2002), Tigray prices
increased much less. On the contrary, in 2004-2005, Tigray experienced two
price hikes (first half of 2004 and between March and July 2005) whereas Addis
Ababa prices were increasing more moderately. From the correlation
coefficients (around 80%) for the different commodities we could assume Addis
Ababa and Oromya grain markets are well integrated. Yet, the Addis Ababa and
Tigray regions have slightly different price variation patterns (correlation
coefficients between 55% (wheat) and 77% (maize)). Regarding price stability
over time, there seems, first, to be no significant outlier in any of the market
price series that could affect our calculations. Then, the 7-coefficients, which
give a measure of how quickly changes in one series (location) are transmitted
to another series (location), show that Tigray (7Tigray=0.45) is far inerter (much
less volatile) than Oromya (7Oromya=0.70) relatively to Addis price variations.

Graph 2: Wheat wholesale prices in Ethiopia

Wheat wholesale prices

400.0

350.0

300.0
Wholesale price (Birr/qt)

250.0

Addis Ababa
200.0 Oromya
Tigray

150.0

100.0

50.0

0.0
D 0

Ju 1

Se 1

D 1

Ju 2

Se 2

D 2

Ju 3

Se 3
03

Ju 4

Se 4
04

Ju 5

Se 5

D 5

Ju 6

Se 6
06
M 0

M 1

M 2

M 3

M 4

M 5

6
0

0
0
-0

-0

0
-0

-0

0
-0

-0

-0

-0

-0

-0

0
-0

-0

-0
p-

n-

p-

n-

p-

n-

p-

n-

p-

n-

p-

n-

p-
ec

ec

ec

ec

ec

ec

ec
ar

ar

ar

ar

ar

ar
Se

Source: EGTE

• Step 2: If we look at Addis Ababa average price differences with Oromya and
Tigray , we can develop the table below:

3
 Average Relative Average Relative
difference difference to difference difference to
Oromya - Addis Addis Ababa Tigray - Addis Addis Ababa
Ababa average price2 Ababa average price
(birr/quintal) (birr/quintal)
2001
Teff -63 -29% -9 -4%
Wheat -25 -16% 43 28%
Maize -21 -23% 36 39%
2002
Teff -55 -27% -15 -7%
Wheat -24 -16% 54 36%
Maize -13 -14% 28 30%
2003
Teff -48 -19% -15 -6%
Wheat -32 -15% 11 5%
Maize -34 -21% 7 4%
2004
Teff -37 -14% 1 0%
Wheat -26 -13% 40 20%
Maize -25 -17% 24 -17%
2005
Teff -39 -14% 11 4%
Wheat -20 -9% 44 20%
Maize -28 -16% 19 10%
2006
Teff -41 -11% -3 -1%
Wheat -49 -16% -10 -3%
Maize -30 -17% 25 14%

The pattern of differences seems constant over time. Exception can be made
for wheat prices in Addis Ababa rising above Tigray prices in 2006 and teff price
difference between Addis Ababa and Tigray, whose sign alternates but still
remains low (relative difference to Addis Ababa price smaller than 10%). We
can therefore assume there is no absolute convergence between the different
markets apart for Teff between Addis Ababa and Tigray where price difference
is close to 0. Oromya prices are on average lower than in Addis Ababa (49 birr
per quintal difference in 2006 for wheat) and generally higher for Tigray (25
birr per quintal difference in 2006 for maize).
• Step 3: Focusing on Tigray, TAMPA points at an average 40 birr per quintal
transportation costs in 2006 for grain from Addis Ababa to Mekelle (we assume
transportation cost to be constant over the period of analysis). This is
approximately the price difference for wheat for example (except in 2003 and
2006), reflecting existence of incentives for traders to move wheat from Addis
Ababa to Mekelle. This same transportation cost difference is not big enough
for maize trade.
• Step 4: Knowledge in the WFP Country Office concerning food flows among the
regions confirms that the flows are indeed better from Oromya to Addis Ababa
than from Addis Ababa to Tigray.
• Step 5: The Productive Safety Net Programme (PSNP) advisory board, on which
WFP sits, recently advised to reduce food-based safety net programmes in
Oromya region in favour of cash-based interventions. This could indeed be

2 Relative differences to Addis Ababa average price are the average difference divided by the average price in Addis.

4
 justified based on the degree of market integration. On the other hand,
Tigray’s situation needs to be closely monitored as most of the cash-based
woredas (districts) in the PSNP asked for switching to food.

Limitations
i) High volatility or persistent price differences need further analysis. A
temporary segmentation among markets might denote other issues and not
reflect behavioural changes that could involve programming adjustments. The
volatility of prices across locations, as well as in one given location, might
nevertheless give the wrong signals to households (frequent important
variations will blur households’ purchase intentions) and therefore increase
their vulnerability.
ii) Demand-side variables are often needed to understand the reasons of market
(dis)integration, as well as derive what the main implications for households’
food security are. A low purchasing power potential in an area may explain the
lack of incentives for traders to move food there. Conversely, in a segmented
market, prices will remain high due to the low food inflows, thus detetiorating
household food access. Unfortunately, data on purchasing power is not easy to
come by and often considered unreliable. One could nevertheless analyse the
terms of trade between food prices and livestock, cash crop prices or wages (or
other income-generating activities) to capture such situations (see MARKIT tool
on terms of trade).
iii) Often various retail markets are highly integrated with a well-identified
marketplace, such as, for example, a particular wholesale market (radiality). In
such cases, an analysis of price seasonality and transport cost changes through
the seasons are necessary. Unfortunately, the availability of those data is often
a problem. Proxies or specific tools can be used (see MARKIT tool on price
seasonality). For instance, if transport cost cannot be monitored, distance, fuel
prices and road conditions can be used to have a proxy indicator because they
have a direct impact on transport cost.

How to calculate the indicators

Following the steps given in the section above, here are some elements to analyse the
price data to capture market integration. The different steps are followed in the
attached Excel datasheet on Ethiopia data.
• Step 1: A simple graph of the different prices over time can often reveal
whether prices move in tandem or not. Normally wholesale prices are preferred
to retail prices because we assume that traders move large quantities between
markets while retailers sell locally3. It is important to use either only wholesale
prices between two markets or only retail prices between two markets, not a
combination. Using the same type of price will make it easier to compare with
transaction costs, as long as we are dealing with price differences. Prices in one
region might lag behind changes in prices in other regions because of the time
it takes to react to price differences and move food. But if a common pattern
exists, even with some delay, it is usually clear.
The correlation coefficient J could also be used to give an idea of the intensity
of the co-movement. It is usually computed as follows4:

3 There should not be a huge difference between retail price differences and wholesale price differences because the margins and
additional costs will likely be similar between wholesalers and retailers.
4 See CORREL() function in Excel

5
 T
( MarketAi MarketA) • ( MarketBi MarketB)
MarketA, MarketB = i =1
T T
( MarketAi MarketA) 2 • ( MarketBi MarketB) 2
i =1 i =1
where MarketXi represents the price of the commodity at time i in market X and
T
1
MarketX = MarketX i the average of the prices over the period T in
T i =1
market X. J varies between -1 and 1. The closer to 1, the more correlated the
prices are and allegedly the better the integration between the markets. The
coefficient of correlation is close to 0 when there is no (linear) link between
the two sets of prices and therefore presumably no flows between the markets
that could regulate the prices through incentives. As far as checking for
stability of prices is concerned, a rapid look at the graph helps identify outlying
values (abnormal prices in one market e.g.). Indeed, too many of them could
threaten the validity of the tools used in the next steps. An erratic data series
might therefore need some stand-alone analysis of the market situation.
Stability of a given market relative to a main market (Market ‘0’) can also be
measured through the M-coefficient:
T
1
( MarketX i MarketX ) • ( Market 0 i Market 0)
Cov ( MarketX , Market 0) T
MarketX = = i =1
T
Var (CPI ) 1
( Market 0 i Market 0) 2
T i =1
Market ‘0’ is referring to prices on a main market that can also be the
Consumer Price Index in the country, or its food subgroup. If the M-coefficient
is 0.8, the prices have varied approximately 0.8% in the market of interest
when the main market prices have varied 1%. The prices on this market are
therefore less volatile than on the main market. A M-coefficient of 0.5 (or 2)
means the price in Market X are moving twice as slow (twice as fast) and thus
indicating an inerter (more volatile) market;
• Step 2: The next step is the analysis of the difference between market A and
market B prices, if any. One can derive the average difference as follows:
T
1
Diff A B = ( MarketAi MarketBi )
T i =1
whose sign indicates which market has, on average, higher prices and whose
magnitude indicates how important the gap is. A difference close to zero
indicates an absolute convergence, i.e. potentially a very good integration
between the two markets. A non null difference - relative convergence - needs
more investigation. A chart of the price differences over time (rather than the
prices themselves) could be helpful. Along this line, one can compute the
relative difference to one market to understand what the estimated level of
difference represents. Relatively to market A price level, the difference will
then be:
T
1
( MarketAi MarketBi )
T
relativeDiff A = i =1
;
MarketA
• Step 3: From secondary sources (or from WFP logistics and procurement units),
estimates of transportation costs can be useful to explain price differences in
the case of relative convergence. These costs may explain market segmentation
if they are larger than the price differences calculated above because they

6
 make it unprofitable for traders to move commodities from one market to
another. A simple comparison over time between the transportation costs and
the price differences in a chart gives a good idea of the possible incentives for
traders;
• Step 4: A rapid checking of actual flows between the areas, through traders’
interviews/focus groups, is necessary to derive any conclusions on market
integration. Analysis through prices does not provide all the answers on market
integration. The actual existence of positive trade flows is the only sufficient
condition for market integration. Therefore, to the extent possible, traders
interviews enquiring about their willingness to move food and risks associated,
are extremely useful complement to price integration analysis; and
• Step 5: Market integration provides important elements for the response
analysis, including whether cash transfers are an option, and the design of
programmes (see MARKIT cash decision tool).

Data needs, data sources

Data needs Type and Data sources

transformation
Time-series for prices (per unit) Weekly or monthly WFP monitoring (FSMS);
of main food staple(s) in major data, plotted against NGOs (for some rural
urban/wholesale markets, time in graph, areas); government (for
including border markets and correlation coefficients major urban areas and
rural markets, if available. Data of price trends, spatial wholesale markets, import
needs depend on the depth of price differences, price prices); Ministry of
analysis required on market differences per major Agriculture
integration seasons

Transportation costs per unit Average cost per unit, WFP procurement/logistics,
between major markets, cost changes by season ad hoc traders’ interviews
including wholesale, border and
rural markets, by major seasons,
if available
Flows of food among these Basic set of questions Ad hoc trader interviews,
markets on the willingness and observations
risks for moving food

[WFP/PDPE, 7viii2007]