CLIMATE VARIABILITY, REAL EXCHANGE RATE AND AGRICULTURAL TRADE IN MOROCCO: AN ARDL BOUNDS TESTING APPROACH (1963-2023)

Résumé

Abstract

Morocco's agriculture faces two structural vulnerabilities at once: extreme climate sensitivity and deep trade dependence. Cereal cultivation is overwhelmingly rainfed, and the country's agricultural trade coverage ratio, the value of agricultural exports divided by that of agricultural imports, has fallen from a surplus of 1.47 in 1970 to a deficit ratio of 0.54 in 2023. Whether this deterioration is associated with climate variability, and over what horizon, has not been established on long historical data. This paper estimates the long-run relationship between precipitation, temperature, the real exchange rate and this coverage ratio over 1963-2023, using annual data drawn entirely from primary sources: FAOSTAT, the ERA5 reanalysis and the World Bank. Domestic agricultural output is excluded from the baseline specification because the estimand of interest is the total effect of climate: since climate is expected to act on the ratio through production, conditioning on production would identify only the residual direct effect. The autoregressive distributed lag bounds testing procedure of Pesaran, Shin and Smith (2001) is applied, with lag order selected by the Schwarz criterion on a pre-declared grid and inference conducted on Newey-West standard errors. Unit root tests place precipitation at order zero and the remaining series at order one, none at order two. Two sets of estimates are reported, of different evidential status. On the full sample the bounds test gives an F statistic of 4.22, which exceeds the upper critical value at the 10% level alone: this is evidence of a long-run association, retained here as a historical benchmark. Parameter constancy is rejected on that sample, with a break dated 1973 on which three independent procedures agree. On the post-break regime, taken here as the preferred specification, the bounds test gives F = 8.52, above the upper critical value at the 1% level, so the evidence of cointegration is strong. The error correction term is -0.343 and significant at the 1% level, implying a half-life of 1.6 years, and the long-run precipitation elasticity is 1.12, significant at the 1% level: a 10% shortfall in rainfall is associated with a durable deterioration of the coverage ratio of about 11%. That elasticity remains positive and significant at the 5% level for every regime starting between 1971 and 1976. Temperature shows no statistically identifiable independent effect within the linear specification. Climate adaptation and water-efficiency policies could generate a measurable trade benefit by reducing the sensitivity of agricultural trade to rainfall.

Keywords: Climate variability; Agricultural trade coverage ratio; ARDL bounds test; Error correction; Real exchange rate; Morocco; Food sovereignty.