This paper challenges one of the main tenets of agricultural economics—that households behave as though they are single individuals, with production factors allocated efficiently between men and women. The efficient allocation of production factors is assumed to be true in virtually all models of the household in the economics literature, and it implies that yields will be the same on all plots planted to the same crop within a household in a given year, adjusting for plot characteristics.

The paper tests this assumption using an extremely detailed farm-level agronomic data set from six villages in Burkina Faso. An important characteristic of the farming systems in these villages, as in much of Sub-Saharan Africa, is that different members of the household simultaneously cultivate the same crop on different plots. This makes Burkina Faso an unusually opportune environment in which to test whether differences in who controls the plots (the man or woman of the household) is reflected in how resources are allocated across those plots.

The data are from a four-year (1981-1985) ICRISAT panel study of 150 poor households (less than US$100 per capita) in the six villages in three agroclimatic zone. Enumerators visited the sample households every ten days to collect information on inputs (labor, manure) and outputs on each plot since the previous visit, yielding usable data on a total of 4,655 cultivated plots. The data also included plot area, location, topo-sequence, and local name for soil (one of 89 possible soil types), making possible a much finer control for land quality, another production variable, than is generally possible with developing country data.

The analysis produced several surprising findings that contradict the assumption that household production factors are allocated efficiently between men and women.

Findings
The analysis produced several surprising findings that contradict the assumption that household production factors are allocated efficiently between men and women.

• Plots controlled by men have higher labor inputs by both men and children than do plots controlled by women.
• Plots controlled by women have labor inputs primarily from the women themselves.
• Nonhousehold labor (unpaid exchange labor) is used more intensively on plots controlled by men.
• Virtually all ma- nure (fertilizer) is concentrated on the plots controlled by men.

Previous studies have found differences in output per acre or per person but failed to isolate the source of these differences. This analysis found that the output of women's plots, and therefore total household output, could be increased by about 10 percent simply by moving some fertilizer from plots controlled by men to similar plots controlled by women.

Policy Implications
These findings suggest that a richer model of household behavior—one that takes into account internal household dynamics and recognizes that individuals compete as well as cooperate—is necessary for understanding the structure of agricultural production and for designing appropriate policy interventions.

These dynamics have affected a number of efforts to improve agriculture in Africa. For example, an attempt in the early 1980s to increase rice production among women farmers in Cameroon failed because rice was considered a male crop and any income generated from it would have been controlled by men, even if the crop were produced by women.

Consequently, few women entered into rice cultivation. Instead, they continued to grow sorghum, the product they controlled, despite its lower returns.

By contrast, a project in Togo to encourage soybean production succeeded precisely because. it was designed to ensure that the crop would remain in the hands of women. The soybeans were not introduced as a cash crop, which would have changed their status and made them a male crop; instead, they were promoted as legumes that could be used to make sauces. As a result, women began cultivating soybeans on their own small plots.

The effectiveness of interventions depends not just on how women respond to them, but how others respond to them as well. For example, a fertilizer subsidy that increases the output from women's plot may, in the longer term, cause husbands to reduce the amount of land allocated at marriage. Thus it is not sufficient for the design of better policy to recognize that households do not as single individuals. It is also important to recognize that changing the incentives or constraints faced by one household member may induce others to change their behavior in ways that frustrate the intention of the policy intervention. Only by using models that accurate reflect complex household dynamics will policy-makers be able to design policy interventions that are effective.