6
After completion of the survey, a final series of regional debriefings was organised for field staff in order to obtain their views on possible improvements in field procedures and questionnaire design and contents.
This feedback has been incorporated in the development of the next survey round in 2016.
2.6 Sampling design
The sampling design of the ALCS 2013-14 ensured results that were representative at national and provincial level, as well as for Shamsi calendar seasons.
5
In total 35 strata were identified, 34 for the provinces of Afghanistan and one for the nomadic Kuchi population. Stratification by season was achieved
by equally distributing data collection over 12 months within the provinces. For the Kuchi population, the design only provided sampling in winter and summer when communities tend to temporarily settle. The
distribution of sampling areas per province was based on an optimal trade-off between precision at the national and provincial levels.
6
The sampling frame used for the resident population in the ALCS 2013-14 was the pre-census household listing conducted by CSO in 2003-05, updated in 2009.
7
Households were selected on the basis of a two- stage cluster design within each stratum. In the first stage Enumeration Areas EAs were selected as
Primary Sampling Units PSUs with probability proportional to EA size PPS. Subsequently, in the second stage ten households were selected as the Ultimate Sampling Unit USU.
8
The design thus provided data collection in on average 170 clusters 1,700 households per month and 2,040 clusters 20,400 households
in the full year of data collection. The Kuchi sample was designed on basis of the 2003-04 National Multi-sectoral Assessment of Kuchi
NMAK-2004. For this stratum a community selection was implemented with PPS and a second stage selection with again a constant cluster size of ten households. The 60 clusters 600 households for this
stratum were equally divided between the summer and winter periods within the survey period. The reality of survey taking in Afghanistan imposed a number of deviations from the sampling design. In
the first six fieldwork months areas that were inaccessible due to insecurity were replaced by sampled areas that were scheduled for a later month, in the hope that over time security conditions would improve and the
original cluster interviews could still be conducted. In view of sustained levels of insecurity, from the sixth month of data collection onward, clusters in inaccessible areas were replaced by clusters drawn from a
reserve sampling frame that excluded insecure districts. Instead of collecting data among the Kuchi population in winter 2013-14 Shamzi calendar 1392, this was done in autumn of 2014 1393.
5
The Shamsi years 1392 and 1393, in which the ALCS 2013-14 was conducted, run from, respectively, 21 March 2013 to 20 March 2014 and from 21 March 2014 to 20 March 2015, respectively. Season dates are given in Annex
VIII.
6
For an optimal sample allocation a balance was obtained between proportional allocation and equal-size allocation with a Kish power allocation of I = 0.25.
7
For three provinces, the sampling frame consisted of the SDES household listings: Bamyan data collected in 2010, Ghor and Daykundi both with data collected in 2012.
8
Some of the selected EAs in rural areas comprised more than one village. In those cases only one village has been selected with probability proportional to the village size, thus creating a third sampling stage.
7
Sample weights were calculated for up-scaling the surveyed households and population to the total number of households and population in Afghanistan. The calculation was based on the official CSO population
estimate by province for January 2014 and average provincial household size derived from the survey. In view of the unequal implementation of the sample across seasons, a post-stratification adjustment was
imposed to give equal weight to the seasons. Annex IV gives an account of the background and technical details of the sampling design and
implementation.
2.7 Field operations
