This paper attempts to address the following question: Why did a climatic shock in 2004 result in
a severe food crisis in Niger in 2005, whereas the shock of 2000 did not? Using a unique dataset that
combines information on prices, transaction costs, trade flows and market structure and conduct, this paper
empirically examines market performance in Niger in 2005 as a potential explanation for the crisis. Using
traditional time-series tests, we find that markets in Niger are partially integrated, with higher degrees of
integration during drought years. Food crisis regions are persistently less integrated than non-crisis regions,
although this pattern was reversed in 2001. In addition, cereal markets consistently violate the no spatial
arbitrage conditions, with positive profits occurring in approximately 60 percent of market pairs.
Nevertheless, spatial arbitrage opportunities were significantly reduced during the 2004/2005 marketing
season. Using a threshold autoregressive model (TAR) to measure market efficiency, we show that markets
are inefficient, with slower rates of adjustment during positive-profit regimes and in food crisis regions.
These findings are robust to the inclusion of market pair effects and other explanatory variables, as well as
instrumental variables (IV) estimation using the Arellano-Bond estimator. We provide some evidence on the
mechanisms behind poor market performance in Niger. By analyzing trader-level data, we find that markets
appear to be competitive, and hence do not explain persistent positive profits. However, a gas price shock in
2004 appears to have increased spatial transaction costs, reducing incentives to trade in 2004/2005. In
addition, lower levels of cereal production in key Granger-causing markets appear to be associated with a
reversal of spatial arbitrage opportunities with Nigeria, a key trading partner. Finally, while the phase-in of
cell phones between 2004-2006 appears to be associated with a reduction in price dispersions, this effect is
ambiguous in cell phone and non-cell phone regions.