13
SAS–2026 A
© NISR
Table 3: Population size per district by stratum (Number of segments)
District
Stratum
Dominant hill
crop land
Dominant wetland
crops
Dominant
rangeland
Mixed stratum
Excluded stratum
Total
Nyarugenge
534
238
-
168
524
1,464
Gasabo
2,165
283
-
697
1,632
4,777
Kicukiro
461
179
-
233
1,000
1,873
Nyanza
5,688
520
-
500
744
7,452
Gisagara
5,197
397
-
824
1,077
7,495
Nyaruguru
3,568
343
-
1,300
6,027
11,238
Huye
3,160
346
-
1,466
1,496
6,468
Nyamagabe
5,344
263
-
1,154
5,352
12,113
Ruhango
5,663
336
-
489
487
6,975
Muhanga
4,983
237
-
760
1,200
7,180
Kamonyi
5,530
320
-
704
777
7,331
Karongi
5,757
117
-
726
2,159
8,759
Rutsiro
4,511
-
-
776
2,083
7,370
Rubavu
2,516
-
-
446
843
3,805
Nyabihu
3,481
-
-
671
1,896
6,048
Ngororero
5,580
134
-
461
1,276
7,451
Rusizi
3,731
155
-
886
5,500
10,272
Nyamasheke
4,584
134
-
953
4,839
10,510
Rulindo
4,144
304
-
625
1,219
6,292
Gakenke
5,934
249
-
671
966
7,820
Musanze
3,111
126
-
769
1,869
5,875
Burera
4,256
260
-
667
1,976
7,159
Gicumbi
5,883
208
-
950
2,176
9,217
Rwamagana
5,060
163
-
1,194
1,122
7,539
Nyagatare
6,591
516
9,112
1,112
4,050
21,381
Gatsibo
7,362
435
788
1,100
7,781
17,466
Kayonza
6,471
149
3,825
1,293
9,730
21,468
Kirehe
7,704
1,501
3,972
13,177
Ngoma
6,293
-
1,201
2,154
9,648
Bugesera
6,957
612
-
2,341
4,456
14,366
National
142,219
7,024
13,725
26,638
80,383
269,989
Source: NISR, SAS 2026
2.1.4. Sampling procedures
Out of the five defined strata, only the dominant hill crop land stratum, dominant wetland crops
stratum, dominant rangeland stratum and mixed stratum are considered as agricultural land poten-
tial. The remaining stratum is classified as non-agricultural land. Notably, clusters comprising tea
plantations and wetlands designated for paddy rice cultivation are not considered in the area sample
frame due to reasons stated above. Thus, SAS is conducted on four strata mentioned above. In the
first stage,1200 segments are selected and allocated at the district level using the power allocation
approach Bankier (1981). Within each district, the sampled segments are distributed across strata
according to a proportional-to-area criterion.
14 SAS–2026 A © NISR Table 4: Allocation of 1200 sampled segments per district by stratum District Agricultural land on hillside Agricultural land in marshland Rangeland Mixed Total Nyarugenge 12 6 2 20 Gasabo 22 4 3 29 Kicukiro 13 5 2 20 Nyanza 37 4 2 43 Gisagara 33 5 3 41 Nyaruguru 25 3 7 35 Huye 27 3 5 35 Nyamagabe 36 2 6 44 Ruhango 36 3 3 42 Muhanga 33 3 4 40 Kamonyi 36 3 4 43 Karongi 38 2 3 43 Rutsiro 34 4 38 Rubavu 21 4 25 Nyabihu 29 3 32 Ngororero 38 2 3 43 Rusizi 27 2 5 34 Nyamasheke 31 2 5 38 Rulindo 28 3 4 35 Gakenke 37 2 4 43 Musanze 24 2 4 30 Burera 30 2 3 35 Gicumbi 37 2 5 44 Rwamagana 34 2 6 42 Nyagatare 31 5 25 7 68 Gatsibo 38 3 5 5 51 Kayonza 32 2 13 5 52 Kirehe 45 9 54 Ngoma 39 6 45 Bugesera 45 3 8 56 Total 948 75 43 134 1,200 Source: NISR, SAS 2026 At the second stage, 25 sample points are systematically selected, following a special distance of 60 meters between points. The sample points serve as reporting units within each segment. Enumerators visit each point, identify and delineate the plot in which it falls, and collect records of land use and related information. The information recorded is used to characterize the whole segment, and the results are extrapolated to the stratum level. Hence, the combination of strata within each district provides district-level area-related statistics.
15 SAS–2026 A © NISR Map 4: Map showing square cluster(segment) with 25 sampled points Source: NISR, SAS 2026 2 .1.5. Weighting Procedures Based on the stratified two-stage sample design used with the new area frame, the first stage sampling probability for sample segments within each stratum is calculated as follows: Where: p1h= probability of selection of sample segments in stratum h (district by stratum) nh = number of sample segments selected in stratum h Nh = total number of segments in the area frame for stratum h in each stratum The second-stage probability was calculated at the plot level, based on the assumption that the plots within each sample segment were implicitly selected with PPS using the area of the plot as the measure of size. Therefore, the second-stage probability of selection can be expressed as follows:
16 SAS–2026 A © NISR Where: p2h= Probability of selection of the plot in segment h ghi= Number of grid squares selected in the i-th sample segment of stratum h; Ahij= Area of the j-th sample plot selected in the i-th sample segment of stratum h Ahi= Area of the i-th sample segment of stratum h; ghij= Number of selected grid squares in the j-th sample plot of the i-th sample segment of stratum h The weight of a sample plot is equal to the inverse of the first and second stage probabilities of selection: Where: WPhij= weight for the j-th sample plot in the i-th sample segment in stratum h 2 .1.6. Sampling errors computation Sample survey results may be subject to two types of errors: (i) sampling errors and (ii) non-sampling errors. Non-sampling errors encompass all sources of error not related to sampling and may occur throughout all aspects of the survey process, including data collection and processing. They are categorized into four types: coverage errors, measurement errors, non-response errors, and processing errors. Although researchers take steps to minimize such errors during the design and implementation phases of the survey, their elimination is practically impossible. Non-sampling errors, in particular, can be extremely challenging to identify and quantify accurately. Despite our best efforts, there’s always some degree of uncertainty associated in the survey results persists due to the presence of these errors. Sampling errors are associated with the sampling selection process, and arise from observing a sample rather than the entire population. They represent the disparity between an estimate derived from a sample survey and the true value that would result if a census of the entire population were conducted under the same conditions. In order to assess the precision of the most important estimates derived from the SAS 2026 Season A data, as well as the statistical efficiency of the agricultural area frame and sample design, it is important to calculate the sampling errors and corresponding coefficients of variation (CVs) for these estimates, such as the total area under each major crop. The sampling error of each estimate is measured by its standard error, which is the square root of the variance. The Complex Samples module in SPSS and Stata use a linearized Taylor series variance estimator that considers the stratification and clustering in the sample design. The SPSS Complex Samples software was used to calculate the sampling errors and CVs for estimates of the total area of major crops from the SAS data. The formula for the estimate of a total can be expressed as follows: Where:
17 SAS–2026 A © NISR Where: L = number of strata yhij = value of variable y for the j-th sample household in the i-th sample segment in stratum h The variance estimator for a total used by the Complex Samples module of SPSS and Stata can be expressed as follows: Variance Estimator for a Total: Where: yhij = value of variable for the -th sample plot in the -th sample segment of stratum h The survey estimate of a ratio is defined as follows: Where Yˆ and Xˆ are estimates of totals for variables y and x , respectively, calculated as specified previously. In the case of a stratified two-stage sample design, means and proportions are special types of ratios. In the case of the mean, the variable X, in the denominator of the ratio, is defined to equal 1 for each unit so that the denominator is the sum of the weights. For a proportion, the variable X in the denominator is also defined to equal 1 for all units; the variable Y in the numerator is binomial and is defined to equal either 0 or 1, depending on the absence or presence, respectively, of a specified characteristic for the unit. The variance estimator for a ratio used by SPSS Complex Samples and Stata can be expressed as follows: Variance Estimator for a Total: Where: , X Y
=
R ˆ ˆ ˆ
18 SAS–2026 A © NISR Fh : first stage probability for stratum h; is the finite population correction (fpc) factor yhij : value of variable for the -th sample plot in the -th sample segment of stratum Variance Estimator for a Ratio: Where: and are calculated using the formula for the variance of a total. In addition to calculating the standard error, the program also computes the Design Effect (DEFF) for the main indicator, which is the area under cultivation. The Design Effect is defined as the variance of an estimate based on the actual complex sample design divided by the corresponding variance from a simple random sample of the same size. It serves as a measure of the relative statistical efficiency of the sample design, taking into account both the stratification and clustering present in the sample design. The presence of clustering typically increases the design effect, owing to the intra-cluster correlation of plots within the segments. Simultaneously, the land-use stratification of the segments tends to decrease the design effects, as it proves to be more efficient than a simple random sample. This dual consideration of both factors provides a comprehensive assessment of the efficiency of the sample design in capturing the nuances of the area under cultivation. The estimates of the total area of major crops at the national level and the corresponding measures of precision (standard error (SE), the coefficient of variation (CV), the 95% confidence interval, the design effect (DEFF), and number of unweighted observation (n of sample plots) from the SAS 2026 Season A data are presented in Table 5
19 SAS–2026 A © NISR Table 5: Sampling Errors for major crops at the national level Season A 2026 data Crop name Estimate SE CV 95% Confidence Interval DEFF No. observations (plots) Lower Upper Maize 245,404 6,093 0.025 233,449 257,360 0.081 8,985 Sorghum 38,302 3,309 0.086 31,810 44,794 0.742 747 Beans 327,907 7,328 0.022 313,529 342,285 1.017 8,626 Paddy rice 17,209 85 0.005 17,043 17,375
2,488 Irish potato 55,310 3,102 0.056 49,224 61,397 1.421 1,811 Sweet potato 96,217 3,676 0.038 89,005 103,429 2.548 1,765 Soybean 28,546 1,435 0.050 25,731 31,360 0.124 1,231 Vegetables 19,244 1,228 0.070 15,082 19,900 0.541 736 Cooking banana 99,689 4,113 0.041 91,619 107,759 1.255 3,539 Dessert banana 38,127 1,594 0.042 34,999 41,256 1.621 3,496 Banana for beer 129,860 4,783 0.037 120,475 139,245 1.483 3,818 Cassava 236,357 6,111 0.026 224,365 248,348 1.314 5,277 Pea 9,563 1,133 0.118 7,340 11,787 1.792 441 Groundnut 9,262 928 0.100 7,441 11,083 0.935 201 Fruits 8,183 2,274 0.168 9,050 17,975 0.525 918 Source: NISR, SAS 2026 2.2. Data collection procedures The SAS data collection is carried out into two distinct phases. The first phase, referred to as screening, is done during the planting period. It consists of delineating all plots containing the sampled points within all sampled segments, as well as identifying all Large-Scale Farmers (LSF) who have grown crops during the current season. During this phase, information is recorded on agricultural land use, crops grown, crop area, and the expected harvest period. The second phase involves collecting data within the agricultural plots identified during screening activity. This phase captures information related to crop production, agricultural inputs, and agricultural practices. 2.2.1. Time frame and coverage During data collection for Season A 2026, the SAS was carried out across all 30 districts of the country, gathering data from 1,200 segments and 554 large-scale farmers. Data collection for the Season started on December 1st, 2025, and concluded on February 18th, 2026. The survey attained a 100% response rate, with complete coverage of all sampled segments and the full participation from all operators of the sampled plots, as well as all sampled large-scale farmers. 2.2.2. Field staff During this season, a total of 147 enumerators and 29 team leaders, all of whom were experienced, served in the field data collection after having completed a refresher training. To ensure data quality, high-level supervision was conducted throughout the data collection activities.