1.2 Sampling Design Refreshment Sample

In 2018, we included a refreshment sample, intending to counteract the effects of attrition (see Survey attrition). Its design is explained below.

Sampling Frame Setup for Refreshment Sample

At the date of preparation of the sampling frame for the refreshment sample, the 2017 Census databases were not yet available to use as a sampling frame for households; we decided to use as a sampling frame the census of households conducted by the CMD during the 2016.

The registration lists households surveyed in person in the first round of the ELSOC survey (Original sample) and therefore consists of the blocks selected for the Original sample survey. This includes about 40,000 households in 1,067 blocks (see Stage 2: Block Selection).

Given that when generating the sample design for the refreshment sample, the registration is two years old, there may be problems of non-coverage in the registered blocks. Additionally, 179 segments have less than 20 households registered; we did not consider these segments to select the new sample. If a chosen segment was not residential, we picked a replacement in the sample selection.

The selection process of households and interviewees was again carried out in 4 stages:

Stage 1: City Selection

To select cities, we decided to use the same cities as the original sample, attending the registration conducted by the CMD. Thus, the sample design of the Refreshment Sample is equivalent to the sample design of the Original Sample up to Stage 1 of City selection.

Stage 2: Blocks Selection

The selection of blocks followed a simple random sampling among those present in the CMD registry and those selected for the Original Sample. We took this decision considering that it simplifies the data collection process by not altering the blocks to be visited (thus increasing the productivity of the survey) and allows maintaining similar representativeness concerning the original sample.

However, we discarded before selection blocks with fewer than 20 households or that were part of the pilot survey for the 2018 Refreshment Sample. This means a reduction of the universe of available blocks from 1047 to 822. Finally, since the Refreshment Sample has a target size of 1400 individuals, 402 blocks were selected.

Table 1.5 shows the registered, available and selected blocks, according to the region:

Table 1.5: Available and selected blocks, by region

Region	Registered blocks	Blocks < 20 households	Pilot blocks 2018	Available blocks	Selected blocks
Tarapacá	23	3	0	20	8
Antofagasta	24	3	0	21	9
Atacama	51	18	0	33	18
Coquimbo	73	22	0	51	23
Valparaíso	159	28	0	131	70
Metropolitana	283	32	20	231	100
O’Higgins	47	7	0	40	16
Del Maule	92	24	0	68	32
Biobío	143	34	0	109	64
La Araucanía	94	24	0	70	38
Los Ríos	22	6	0	16	9
Los Lagos	25	3	0	22	10
Aysén	11	1	0	10	5
Total	1047	205	20	822	402

Stage 3: Housholds Selection

We selected households by simple sampling among the households in the selected blocks. Given that it is defined to obtain 1,400 surveys selecting an average of 3.5 households per block is necessary to meet the target sample. For this purpose, in blocks with more than 20 households, we selected between 4 and 6 households per cluster, depending on the size of the strata and blocks. In blocks with more than 50 households, we used another sub-block generated in the registration, ensuring that this sub-block is not in the same building already selected.

The probability of selecting the same household in the block is non-zero, so a replacement was generated for that unit with the same random selection method, for which we verified the selected sample with the panel sample already defined in 2016.

Stage 4: Individuals Selection

As in the Original Sample, interviewers visited each selected household and attempted to perform the interview. The first step in the interview process is to identify the target interviewee, who is selected using a simple random sample of adults between the ages of 18 and 75 who live in the dwelling.