Econometric analysis of historical price data from the region over three decades suggests
that positive NDVI is associated with markets that appear to be less integrated, as evidenced
by the decreasing effect of neighboring price lags in a good year on a markets observed price.
Essam conducted an analysis, using model fit criteria that shows that by including specific
regime variables that interact with lagged price bands from neighboring markets, we can
improve the overall fit of our base cereal market performance model. The regime variables
are based on NDVI anomalies that provide the overall abundance of millet in the region,
providing information that was previously not available to the model.
Essam also uses probability modeling to assess the ability of NDVI anomalies to predict
future price regimes for inclusion in his millet price prediction model. The results suggest that
NDVI anomalies from May, June and July have a rather limited ability to predict accurately
future price regimes. However, as he included additional NDVI anomalies from previous
months (March and April) that capture weather conditions and the amount of senesced
vegetation on the ground before the start of the growing season, his ability to predict price
regimes increased dramatically. This may be due to the importance of the previous year's
moisture conditions on the current year's food stocks. NDVI can show variation in biomass
in crop residue and other senesced vegetation, as well as differences in soil moisture over large
areas (McNally et al ., 2013).
The model also showed that in bad years, on average, markets appear to be better integ-
rated and in good years, on average, markets are less integrated, where integration is measured
as the degree of price transmission from neighboring markets. Market integration is an
important metric for estimating the impact of local weather shocks as well as international
price signals on local food security (Essam, 2013). Market characteristics can be inferred from
their response to different shocks.
Impact of weather and global food price shocks
The behavior of local markets assessed in the model described above has different character-
istics that can be summarized in a general typology, presented in Table 7.2 . The typology
provides an explanatory matrix to describe why we see differences in the impact of weather
and international price shocks on different markets. The typology is a useful way to think
about why each market is affected by these shocks and how these impacts are likely to change
over time. It is not meant to capture all the different sources of variability but to simplify
complex relationships so that they can be examined more easily. It is unlikely that these rela-
tionships will stay the same during periods of extreme stress, such as during very high inter-
national prices or very low food production due to weather shocks, but categorizing how
markets are affected will help information about these shocks be more easily used. Econo-
metric models that use NDVI as one of the input parameters, such as the study conducted by
Essam described above, will allow further exploration of these issues as the length of the price
time series as well as the locations that have price data expands.
Market-commodity pairs not affected by either international price or weather
In countries that actively control food prices, we expect to see shocks have less of an effect on
local prices. Kenya, Malawi, Zambia and Zimbabwe can be classified high-intervention