MARKETS AND DETERMINANTS OF FOOD PRICES - Food Security, Food Prices and Climate Variability - page 105

there is a lag in the price response between production and delivery of goods at a market
(Barrett, 1996).
The Granger-causality metric has been used to study grain market integration in many
analyses across the world. If prices, lagged by one or two months, from one market are useful
in forecasting prices in another market, even after controlling for own-lagged prices, then the
second market is said to Granger-cause price movements in the first. The procedure is usually
carried out within the framework of a bivariate regression, a vector autoregressive or error-
correction model and confirmed or rejected with an F-test on estimated coefficients. Some
analysts have taken the presence of Granger-causality to mean that shocks to prices in one
market may induce a significant response in another, with specific time lags. Others have
considered it as an indicator of the flow direction of information or goods between markets.
Baulch (1997) adds that if two-way Granger-causality exists, then prices are simultaneously
determined. Fackler and Goodwin (2001) point out the test only allows inferences about
lead/lag relationships and little can be said about the causal framework that underlies the
dynamic adjustments (Brown et al ., 2012).
To produce country-level indices, information was used from a continuously updated
price database comprising food prices from 232 markets in 35 countries, collected by the FAO
and FEWS NET. Most of the data is publically available from the FAO at the Global Informa-
tion and Early Warning System (GIEWS) food price website: www.fao.org/giews/price-
tool2/ . There are several data sources for each country. While a portion of the data can be
downloaded from FAO/GIEWS's PriceTool website, not all of the data are available at this
public database. While the FAO and FEWS NET use the same data sources (Ministries of
Agriculture), and most of the data is publicly available in some form (from newsletters, web-
sites and by request from local Ministries or from the World Food Programme), it is not
available from any one source in its complete, constantly updated form, and thus is a unique
dataset. There are 124 different commodities in the database from which we draw for the
study. FEWS NET experts deemed these commodities as appropriate to assess food security
in the selected area. The database has retail price data in local currencies from 1997 to 2011,
but the starting year of each series varies by market. The most complete information from
2004 through 2011 is used to form the price indices.
Combining the food price data into indices uses information on local production systems
and patterns of food consumption. FEWS NET Production and Market Flow Maps were
used to classify markets into major surplus, minor deficit and major deficit zones in each
country, creating marketsheds.
The trade flow maps can be found on the FEWS NET website ( www.fews.net ) under
Markets and Trade. For each country and region, the following indices were calculated:
• nationalcerealspriceindexfor35foodinsecurecountries,constructedusingpricesofcom-
modities that are important in the diet of food insecure communities (see Table 5.3 );
• regionalindicesforWestAfrica,SouthernAfrica,EastAfrica,CentralAmericaand
Central Asia for:
• cereals only, non-cereals only and all food prices available indices,
• capital cities only, non-capital cities and all markets indices.
While making the country and regional indices, the number of data observations available
from January 2004 to April 2011 was used as a weighting factor for each market and commodity