FOOD PRICES AND SEASONALITY - Food Security, Food Prices and Climate Variability - page 120

of its caloric demand: if they sell 30 percent of their own-produced cereal post-harvest for
cash income, they then need to buy back 20 percent of the total production in the summer
(110% - 30% + 20% = 100%); due to price fluctuations, the household is a net buyer by value
given a loss of 10 percent of total production (+30% - 2[20%]) = -10%). By having additional
sources of cash income or understanding when prices are likely to rebound and delaying sale
until that point, the household could avoid paying this penalty (Brown et al ., 2009). Informa-
tion on the interaction of current growing conditions and likely impact of global food prices
and typical seasonality could provide this information.

Millet in Mali, Burkina and Niger

Early analysis of the link between satellite remote sensing observations of vegetation in Mali,
Niger and Burkina Faso and the seasonality of millet prices from 1982 through 2007 used a
decomposition technique that allows the quantitative relationship of the seasonality of the
environment to that seen in the food prices (Brown et al ., 2008). Brown et al . (2006, 2008)
used monthly millet prices from 445 markets in Niger, Mali and Burkina Faso. The data were
obtained from the early efforts of FEWS NET to gather and analyze food price (Chopak,
1999). The data have been kept in the local currency (CFA) and the series vary in length and
begin in different years (Niger in 1982, Mali in 1987, Burkina Faso in 1989). Because these
time series span the 1980s and 1990s as well as most of the 2000s, they provide historical
perspective that data beginning after 2000 cannot (Brown et al ., 2006, 2008).
As was discussed in Chapter 3 , remotely sensed vegetation data can be used to represent
growing conditions for millet in West Africa in the two studies. Normalized difference veg- Normalized difference veg-

etation index (NDVI) data have been used extensively in the Sahel to detect variations in
vegetation production, and have been shown by a number of authors to be correlated to local
crop yields (Fuller, 1998; Funk and Budde, 2009; Jarlan et al ., 2005; Vrieling et al ., 2011), and
precipitation (Nicholson, 2005; Nicholson et al ., 1998). The AVHRR NDVI data has 8 km
spatial and monthly temporal resolutions (Tucker et al ., 2005). The AVHRR sensor has
appropriate spatial, spectral and temporal resolutions for West Africa (Becker-Reshef et al .,
2010; Justice et al ., 1991; Prince et al ., 1990). The mean of a ive-by-ive-pixel box (40 × 40 km)
around on each market was calculated from monthly maximum value NDVI composites.
This focus on only the immediate area around each market was a weakness of the analysis,
since it is well known that many larger markets draw on farms for grain across a broad region,
far larger than simply the area immediately next to a market (Aker et al ., 2010; FEWS NET
2009). This weakness is addressed in new analysis presented in Chapter 7 .

West African production patterns

Like many semi-arid agricultural regions, the growing season in West Africa is monsoonal,
with strong vegetative response to the rapid increase in soil moisture after the first few rains
in June or July and lasting through September, with nearly all food grown during this period.
A rapid increase in food availability occurs after the harvest in October to November. As the
next growing season nears, food prices begin to rise as local food supplies fall and more house-
holds resort to purchasing food on the market. At the peak of the growing season, demand is
high and supply is low, resulting in a growth in food prices, particularly in regions that are
isolated. Plate 15 shows the time series for millet prices and vegetation from the three West