Annual Agricultural Survey 2020-2021

The sample for the 2019/2020 marketing year has been renewed. As a reminder, the EAA was built on a two-stage survey, with census districts (CDs) as primary units (PUs) and agricultural households as secondary units (SUs), as defined in the RGPHAE 2013. In line with the broadening of the scope of the survey advocated by the AGRIS approach, the sampling plan has integrated from this campaign a stratification at the first degree, induced by that of the second degree, to better reflect the different agricultural activities and improve the efficiency of the estimates. The choice of a first-degree stratification induced by that of the second degree, although less effective than an independent first-degree stratification, was guided by the constraint of non-existence of relevant variables of interest in the RGPHAE sampling frame to discriminate against CDs.

The stratification took into account the relative importance of the main agricultural activities (in terms of household numbers) identified in the RGPHAE, namely rainy agriculture, livestock and horticulture. Four strata have thus been constituted as follows:

• the rain stratum only" which includes all households practicing only rain crops;
• the "Livestock only" stratum for households that practice livestock only;
• the "Horticulture and other crops" stratum, which includes households that mainly practice horticulture and incidentally other crops (forestry, fruit arboriculture, etc.);
• the "Rainfed-Livestock" stratum made up of households that practice both rainfed agriculture and livestock.

The sample size of the agricultural households to be surveyed was calculated by department (field of study) by setting a relative error of 10% on the variable of interest (see Technical Note in the Annex). At the national level, this resulted in a theoretical sample size of 7,300 distributed in 1,460 physical CDs, at a rate of 5 households per CD. At the end of the enumeration operation carried out in the rds physical samples, adjustments were made to take account of the actual updated size of the CDs, which led to a final size of 7 755, or 1 527 CD.

The average yields of the main crops are calculated per department on the basis of 60 plots drawn and distributed in 12 different CDs. As the collection of plot data often takes longer than expected, the drawing of yield plots is done a priori based on data from the 2019/2020 season. The basic assumption is that households have roughly the same farming practices from one campaign to the next. Thus, to constitute the sample of the plots that are the subject of yield square installation, part of the sample of the households surveyed is drawn. For each main crop and for each department, a sample of 80 households is randomly drawn from households that had practiced cultivation in 2019/2020, including 20 replacement households in case the drawn household did not renew the culture of the past campaign. If for the crop in question, the number of households that had practiced it is less than 60, the investigator is asked to complete the information from an out-of-sample household drawn in the same CD.