How reliable are national accounts estimates of agricultural output?

Measurement errors in macroeconomic aggregates such as GDP have been widely lamented, particularly in low-income contexts. This study investigates the reliability of one component of national accounts, agricultural sector output. Focusing first on the case of Mozambique, we use a series of 12 harmonized national agricultural micro-surveys to construct estimates of gross annual output in the sector. Compared to corresponding national accounts values for the period 2002-20, the micro-survey estimates are about 50% lower. A decomposition exercise indicates this gap is primarily driven by differences in base year levels, offset by higher rates of inflation observed in survey-based producer prices. Triangulating these estimates using household budget surveys, market price, and FAO production data, we find consistent support for the agricultural micro-survey estimates, suggesting real rates of total GDP growth may have been overestimated by 1 percentage point over the period. A cross-country comparison of national accounts and FAO production data indicates positive differences between estimates of agricultural output from the two sources are not unique to Mozambique, but with larger discrepancies among lower-income countries. Systematic investment in reliable and timely agricultural statistics is essential to track progress in the sector.