methodology

Questionnaire  

Four type of questions are included in RCSS:

 

  • International questions from ESS

  • National questions included in the international databases (education, ethnic origin, income, religions, profession of respondent/ spouse, industry of work, profession of respondent/ spouse, region of residence)

  • National questions for Russia only

  • Methodological part а) MTMM; б) Questions for interviewer.

     

 

All interviews in RCSS are conducted in Russian language.

 

The translation of international questions is conducted by TRAPD method - Translation, Review, Adjudication, Pretesting and Documentation [Harkness, 2003]. This method implies the team work on the translation which starts from the parallel translation of two interpreter (preferable with different background), team expertise of differences and difficulties and final decision. The expert translation group in Russia includes A.V. Andreenkova, A.V. Fedotov, A.E. Bronnikova, V.G. Andreenkov, V.S. Magun, M.G. Rudnev and scholars in different sociological fields for rotating modules and invicidual questions. The decisions about final version of the translation also takes into account the translation into Russian in other countries – Latvia, Lithuania, Estonia, Israel, Georgia and Ukraine, although is not based on this comparison, so the final translation versions into Russian may differ.

 

Pretesting of the questionnaire is conducted by qualitative and quantitative methods. In rounds 3-8 the pretest was conducted by interviews with respondents from different socio-demographic groups using the full version of the questionnaire. In rounds 9-11 quantitative interviews were supplemented by cognitive in-depth interviews for testing individual new questions.

 

Since round 9 the interviews are conducted using computer-assisted program on tablet or laptop (CAPI). The program for CAPI interview is prepared by CESSI programmers (the head of the group is E.V. Obiyuch). The program includes the section of Respondent selection and Contact information, Main interview, Questions for interviewer. Visual materials (showcards) for respondents in Rounds 9-11 were used in 3 formats: printed paper, electronic for table and online on central server.

Sample design

Sample universe:

 

Available population of Russia age 15 and over in residential dwellings regardless of ethnicity, citizenship or judicial status. The sample universe includes all residents in selected households living in it during time of the survey (2-3 months). The survey includes labor migrants and temporally moved people who live in residential dwellings, students in dormitories, but does not include guests visiting for a short time. The residential dwellings are defined as apartments in multiflat houses, private houses, student’s dormitories.

 

The exclusions are: a) settlements with 'closed' visiting regime, some hard-to-reach and low populated areas of Far North, Siberia, Far East and North Caucasus; b) residents of institutions such as prisons, military areas, nursing homes, monasteries; c) people living in places of temporal residency – hotels, sanatorium, holiday hotels, hospitals; d) people who are absent from the country for than the time of the survey (2-3 months). In wave 11 fieldworks were not conducted in some territories with military actions and closed military regime.

Sample design:

 

The sample is designed by four-stage stratified cluster territorial sample model.

The similar sample model was used during all waves of the survey. However, the particular design and sample realization were different.

 

Stratification of the country into regions in waves 3-6 was based on administrative socio-economic zones, in waves 7-11 – by federal okrugs. The databases include variables for SEZ and FO for each wave.

 

In the survey of 3-6 4-stage sample were used, the primary sampling unit (PSU) was urban and rural settlements (cities/ towns/ villages), the secondary units – electoral districts, on third stage – households and fourth state – individual respondents [Andreenkov, 2009]. In surveys 7-11, 3-stage sample was employed. The primary sample unit (PSU) was electoral districts, the secondary – households and individual respondents were selected on third stage.

 

The selection of respondent within a household in surveys 3-6 was done by random distribution of different types of Selection forms (Kish procedure), in surveys 7-11 – automatic random selection of respondents were programmed in the CAPI system.

SAMPLE STAGES FOR ROUNDS 9-11

 

Stratification

Stratification of primary sample units (PSU) by geographical regions – eight federal okrugs. The total number of selection PSU was 150. 15 largest cities of Russia were defined as self-representative (Ni > N/150) – the population size of these cities were larger than the average size of PSU and they were selected with probability of 1. In total 19% of country’s population were represented in 38 PSUs in largest cities. Other 122 PSUs were selected within each strata independently with probability proportionate to the population size (15 years old and over) of each strata by systematic random selection.

Stage 1

Random selects ion of two electoral districts within each PSU with equal probability. The definition of electoral district was used according to the official information for the latest national elections to State Duma (Central Electoral Commission of Russia). The electoral districts with electoral population less than average size of PSU in 2000 adults, were united with geographically adjusted districts into one PSU.

Stage 2

Census of all households (residential addresses) in selected electoral districts either by physical visits, or available lists of geolocation data. The selection of the households was done centrally prior to the survey from total database.

Stage 3

The list of all members of the household of 15 years old and over during first visit to the household was completed by interviewers. Members of a household were listed in fixed order – firs all males from oldest to youngest, then female from oldest till youngest. The selection within a household was done automatically using random selection algorithm programmed in CAPI.

Sample size

 

The sample size for Russia in each round is about 2500 interview (refer to each round separately), except of Round 10 where sample size of 2000 interviews.

 

In ESS countries the sample size in each round is varied from 800 to 3500 for one country and depends on the sample design model and expected response rate to reach approximately equal effective sample size of 1500 cases per countries with population size more than 2 mln people and 800 cases for other countries.

Sample implementation

The data collection method is personal face-to-face interviews in respondent’s homes, in Rounds 3-6 – on paper (PAPI), in rounds 7-11 – computer-assisted interviews on tables and notebooks (CAPI).

 

Each selected sampling unit (household) is visited at least 4 times for unsuccessful contacts.

Information about response rate, detail analysis of the contact results and the comparison of sample composition with official statistical data on major socio-demographic and regional parameters is presented in Methodological reports on each round of the survey.

Fieldwork

The average length of an interview is about 60 minutes. Each round of the survey is implemented by 250-300 interviewers, their work is organized and monitored by 70-90 supervisors in different parts of the country.

The survey is conducted according to the “Declaration about professional ethics”[1]. In particular:

a) All respondents are informed about voluntary participation in the survey before the procedure of selection of respondent within a household;

b) Verbal consent of parents for the participation of the survey for underaged respondents (15-17 years old) is obtained before the interview;

c) Strict procedures of keeping anonymity and confidentiality of individual information and individual respondents are implemented. Contact information is not included in the database and kept separately in secure format for fixed period of time and it is used for quality control only. The main data are checked for information leading to potential identification of respondents using more than 15 intercorrelated parameters. Information which may lead to the violation of non-identification rules, are deleted or substituted according to the formal rules. Only large geographical units are identified and provided in the datafile to decrease the likelihood of potential identification.

 

 

[1] Declaration on Professional Ethics 2010 https://www.isi-web.org/index.php/activities/professional-ethics/isi-declaration

 

Quality control

The quality control is conducted using follow-up calls and visits, about 25% of work of each interviewer were checked.

 

The data passed logical control on:

a) ineligible codes,

b) data consistency,

c) large number of missing data in one interview,

d) interview lasted more or less than average length on 25% or more and other important parameters.

In case of ineligible or doubtful situations, an interview was directed to additional quality control by follow up call or visit. If multiple cases of violation of survey specification or standards are detected, all interviews of particular interviewer are assigned to additional quality control. If it is found as unsatisfactory or if full quality control was not possible, all cases completed by an interviewer are removed from the final dataset and the work was done again by different group of interviewers/ supervisors.

Data processing

In rounds 3-8 using pen-and-pencil data collection method, the data were entered into the database centrally by the data entry group of CESSI. The quality control was provided by double entering of 20% of interviews, logical quality control and visual checks. Starting from round 9, all data are transmitted to central database from tables/ notebooks of interviewers. The data are checked by the data processing manager taking into account the comments and information from interviewers and field manager, some checks of quality control on interviewer level are programmed to be completed automatically (average length of interview, length of individual sections or even question is compared to general average lengths, the share of missing data in individual items or in questionnaire in general, straight-lining in battery question or other similarities in filing interviews of the same interviewer). If problem is detected, the interview or the whole set of interviews of an interviewer are assigned for additional quality control.

 

The data in the final dataset are anonymized to avoid even potential possibility of direct or indirect identification of particular respondent. The rule of minimal cell size after crossing some identification parameters are used to estimate the likelihood of identification. The aggregation rule for geographical information about the place of residence in RSS is oblast level.

Constructing and using weights

 

In order to correct for sampling biases due to the sample design (clusterization) and differential non-contacts and non-response, two weights are constructed – design-weight (dweight) and post-stratification weight (pspwght).

 

    Dweight – the weight which accounts for the differences in probability of selection of units in different sample stages.

    Pspweight – post-stratification demographic weight is based on the comparison on the sample data and official statistical demographic data – National Census (NC of 2002 for rounds 3-6, NC of 2010 for rounds 6-9 and NC of 2020 for rounds 10-11) by following parameters: gender-age distribution within 8 geographical regions.

 

Since Round 8 Psgweight is constructed as joint weight (dweight x pspweight), taking into account the differences in probability of selection of sampling units and demographic comparisons with Census data.

 

In international ESS database additional weights are included:

    pweight – to take into account the differences in population size in each country. Each country is weight to its full population size.

Since Round 8 new cumulative analysis weight is included.

    Anweight - (dweight x pspweight x pweight).