In Global Context: Mental Health Study
Indiana University . Bloomington . Indiana
Brazil Sample Design
The first Brazilian academic social survey: Pesquisa
Social Brasileira – PESB (2002)
National survey of Opinion: from June 20st to October 30th
Full probabilistic sample
Sample size: 2365
Financed by Ford Foundation
Alberto Carlos Almeida, Zairo B. Cheibub,
study employs a three stage stratified probability sample of Brazilian
In the first stage, 102 primary sampling units (PSU), or municipios,
were selected probability proportional to size (PPS).
In the second stage, 280 secondary units (census tracts)
were selected PPS within each of the PSUs.
In the third stage, households were selected at random with
one adult respondent being selected at random too.
Each PSU consists of a single município as defined
by IBGE. For the 1996 Contagem, IBGE divided
27 of these municípios
were selected with certainty as self-representing PSUs and 75 as non
self-representing PSUs. The self-representing PSUs are state capitals.
PESB represents the general population. The sample was
designed to be representative of the 5 census regions and state capitals.
To reduce costs of travel all municípios of regions
Norte and Centro-Oeste, with less than 20,000 inhabitants, were excluded.
This comes to a total of 11.9 percent of the municípios and 3.1
percent of the population excluded from the sample. The total effective
sample comes to 2010 respondents.
The PESB employs a two-stage stratification scheme where
PSUs first are explicitly stratified, or grouped, by region and type of município
(state capital/or not).
Moreover, the PESB employs implicit stratification.
Within regions, all non-self representing PSUs were further
stratified by three additional variables: (1) HH population size; (2)
urbanization rate; and (3) literacy rate.
In the sorting of PSUs using the implicit stratification
variables, we used alternating ascending and descending orders in
The explicitly stratify the sample consider the five
Grand regions as defined by IBGE: (1) Sul; (2) Sudeste; (3) Nordeste;
(4) Centro-Oeste; and (5) Norte.
Furthermore, to increase the users ability to analyze
results for the smaller regions, the PESB employs a disproportionate
sample design, sampling the three smallest regions (Centro-Oeste, Norte,
Nordeste, and Sul) at higher rates and the largest region (Sudeste) at a
Finally, we stratified PSUs by whether the sampling unit was
a capital or not. As stated in the previous section, we decided on this
design based upon the potential importance of state capitals in explaining
socio-political differences within
sample design – Logic of Stratification
2004 and SGC-MHS
In order to reduce the overall cost of this year´s PESB, a
smaller sample will be selected - 1200 interviews instead of 2500
The sample will
be selected from the original master sample selected in 2002 - 102 city
the number of interviews for self-selecting PSUs will be reduce by
1200/2500 and the number of NON-self-selecting PSUs will be reduce by
for pre-fielding and fielding activities
1022 E. Third Street, Bloomington, IN 47405 (812) 855-3841