How important is misallocation in explaining the income differences across
countries? A recent literature in development and growth economics has focused
on misallocation across sectors, firms, and plants.
1
This literature has found
evidence that the dispersion of productivity across production units seems to be
consistently higher in poor countries than in rich ones. Such productivity
differences have the potential to account for a large fraction of the cross-country
income differences. In an aggregate sense, misallocation across sectors or firms
reduces aggregate total factor productivity (TFP).
A challenge in this literature is to distinguish misallocation from other sources of
dispersion in productivity, such as technology shocks, measurement error, and
adjustment costs of various kinds. This paper seeks to disentangle these different
sources of dispersion in an environment where cross-firm differences in
productivity are very large, aggregate productivity is low, and market failures
undoubtedly contribute to cross-firm frictions in the allocation of resources.
Specifically, we take advantage of extraordinarily rich data from farms in three
countries in Africa, for which we have detailed panel observations on the same
firms producing identical outputs on different plots in the same time period. Since
farmers face no market imperfections in allocating resources across their own
plots within a growing season, we can use these plot-level data to identify
misallocation more precisely. Our strategy allows us to disentangle the
productivity dispersion that arises from misallocation from that stemming from
measurement error or heterogeneity in technology and inputs (including
production shocks).
The agricultural sector provides a valuable window through which to study firmlevel
misallocation. Most firm surveys have relatively few observations on
different plants or establishments operated by the same firm, and this makes it
difficult to disentangle firm management from any unobservable characteristics of
the plant or factory. Another advantage we have relative to firm surveys is that
our producers are producing highly homogeneous products, with little market
power. Consequently, we can compare the output of different firms (farms)
without worrying about markups and pricing strategies.
Understanding the extent of misallocation in agriculture is also of interest because
of evidence that low agricultural productivity can explain – at least in a
mechanical sense – a large fraction of the cross-country dispersion of output per
worker (Caselli 2005, Restuccia et al. 2008). A cluster of recent papers has
suggested that there may be very large dispersion in productivity at the level of
farms and farmers, potentially indicative of misallocation at this micro level.
2
These papers point out that in poor economies, very large fractions of the
workforce are employed in agriculture, in contrast to rich countries, where very
few people earn a living from farming. In economies where two-thirds of the
people are farmers, it is reasonable to ask whether they are all good at farming –
and whether market failures of various kinds may induce too many low-skill
farmers to remain in agriculture.
Restuccia and Santaeulalia-Llopis (2017), in particular, have raised the intriguing
possibility that much of Africa’s productivity deficit might be attributable to
misallocation within the agricultural sector. In particular, they find suggestive
evidence, based on data from Malawi, that too much farmland is managed by lowskill
farmers. If true, this finding might offer an explanation for sub-Saharan
Africa’s low productivity in agriculture. Indeed, it might by extension help
explain the region’s low levels of income per capita. The finding also suggests a
relatively straightforward solution – albeit one with great political complexity –
namely, the liberalization of land and input markets, so that the best farmers can
eventually buy out those farmers who lack the skill to farm productively.
The misallocation hypothesis for African agriculture is particularly plausible
because of abundant evidence that the continent’s agricultural markets work
poorly – for land, rural labor, intermediate goods, and output. Much land lacks
formal title, and rural labor markets are often poorly integrated. Empirical tests
consistently reject the hypothesis that markets are complete.3