POF's major working methods:
POPULATION POLLS
Household Interview
Telephone Surveys
Data Sources
Weekly nationwide mass polls are conducted with representative samples in 100 settlements in 44 regions, territories and republics in all Russian economic-geographic areas. Interviews are conducted in the respondents' homes. Sample size: 1500 respondents. The margin of statistical error does not exceed 3.6%.
Additional weekly representative polls in Moscow (sample size: 600 respondents, 100 of whom are included in the national sample).
Chart figures are shown as a percentage of the whole sample separately for Russia and Moscow.
Techniques
Population polls in Russia and Moscow are conducted by a uniform questionnaire which consists of closed-ended and open-ended questions.
In response to the closed-ended question, the interviewee is asked to choose the most suitable response alternative – one or several – from those offered. In some cases, the interviewer shows the respondent a card listing response alternatives.
Closed-ended questions with a single response option are the most common variant. This and other situations are stipulated in parentheses following the question (e.g. «any number of responses allowed» or «up to three responses allowed»).
If the respondent is shown a card listing the choices, this condition is specifically stipulated in the questionnaire, and the number of choices is always stated («one response allowed per card»; «any number of responses allowed per card», etc.). When responding to an open-ended question, the interviewee is asked to word it in free form. The interviewer simply reads the question without offering any choices. If an open-ended question suggests only one response alternative, only the question type is specified (open-ended) - e.g. «the respondent gives the names himself», etc.
If the interviewee is supposed to give more than one response, both conditions are noted in parentheses after the question («any number of responses allowed,» «the respondent gives the names himself»).
Processing Open-Ended Questions
All responses are broken down by thematic groups and subgroups. An opinion is taken as the unit of analysis (one response may contain several opinions). Thus, responses given by some respondents can be included into different thematic groups according to their content. Data Representation (Closed-Ended and Open-Ended Questions)
The chart figures are shown as a percentage of the whole sample separately for Russia and Moscow. The most typical opinions are provided for open-ended questions.
Household Interview
The Public Opinion Foundation conducts mass polls in respondents' homes in the form of face-to-face interviews and through completion of questionnaires that vary in length and complexity (the samples represent the total Russian population or certain specific areas). Almost all urban or rural residents have equal chances of being sampled. If a respondent is away at the time of the survey, he or she can be visited again for an interview.
PENTA Household Polls
PENTA is a method of conducting Russia-wide population surveys. When using this method, only five days elapse from the date on which the question is formulated to the date on which the poll's findings are obtained, hence the name PENTA.
PENTA polls are conducted across Russia in the form of face-to-face interviews with respondents in their homes.
The OMNIBUS-type questionnaire consists of questions on one or more issues and can also include custom-designed questions.
PENTA polls have been conducted by the Public Opinion Foundation since the fall of 1992.
The findings of all of our surveys are stored in the POF data base. Since fall 1996, PENTA survey data has been published on the Internet.
Organization of a Poll
Questionnaires are prepared weekly on Monday or Tuesday and emailed to the regions on Wednesdays. Fieldwork is carried out on Friday night, Saturday and Sunday from 10 a.m. to 9 p.m.
The respondents are selected so their views most precisely represent Russian public opinion. The major selection principle is that every Russian citizen has a chance of being interviewed.
The data is emailed from the field back to Moscow on Monday and data processing is conducted on Tuesday and Wednesday. Simple cross-tabs can be provided to clients on Wednesday night, and fully processed data is emailed to the customer on Thursday.
Data Source
Weekly nationwide population polls are conducted with representative samples in 100 settlements in 44 regions, territories and republics in all Russian economic-geographic areas. Interviews are conducted in the respondents' homes. Sample size: 1500 respondents. The margin of statistical error does not exceed 3.6%. Additional weekly representative polls are conducted in Moscow (sample size: 600 respondents, 100 of whom are included in a national sample).
Chart figures are shown as a percentage of the whole sample separately for Russia and Moscow.
Questionnaire
The questionnaire consists of two standard sections. The first asks about social and political issues, while the second includes social and demographic questions: - gender (2 choices);
- age (open-ended question);
- education (5 choices);
- occupation (12 choices);
- title, type of employment (5 choices);
- per capita family income (open-ended question).
Other questions are related to current issues.
The interviewer reads out a questionnaire, and in the course of the interview, shows cards with response alternatives or pictures. The duration of the average interview is 30-40 minutes. Respondents sampled in previous POF polls are not interviewed again.
Description of Weekly Russia-Wide Poll Sampling
1. The Sample Universe
The sample universe is the urban and rural population of Russia, aged 18 and above.
Owing to the current situation in Chechnya, which makes it impossible to conduct polls there, this region is excluded from the sample frame. Homeless people are also excluded (households are taken as the unit of selection, and interviews are always conducted at the respondents' homes).
Adult residents of Moscow are sampled separately.
2. Sample Design
Multi-stage stratified territorial random sampling is used in our polls. 2.1. The Sample Frame
The sample frame has two specific features:
The first feature consists of two parts. Each one represents the Russian population and can be used separately. If the number of respondents does not exceed 1500, it is better to use one of the sample parts. When the number of interviews exceeds 2000, it is more appropriate to use the whole sample.
The Public Opinion Foundation uses the advantages of this sample frame in the following way. Russia-wide weekly population polls are conducted alternately with each half of the sample in even and odd weeks. Some questions are asked in two successive polls. This allows us to gather representative data across Russia and gain more accurate estimates from the repeated questions asked over the past two weeks, which corresponds to the whole sample.
In all sampling points, the interviewers have an equal workload. In addition, the application of this sample frame allows the polls to be conducted twice a week, each one with its own sample part. In this case, the interviewers' workload is not great, since each sampling point is included in only one poll.
The second feature of the sample frame is the separate sampling of Moscow residents. The estimated sample size for Russia is 1500 respondents, 86 of whom are Muscovites. To analyze data separately for Moscow, the sample size is increased to 600, with the whole sample thus amounting to 2014 respondents. Disproportion in the sample frame is compensated by weight coefficients.
2.2. First Selection Stage
The primary selection units at the first stage of the sample are administrative districts (raions), as well as towns subordinated to the government of a federation subject and not related to any district, that is situated on the border of two or more raions, (for convenience, further referred to as «areas»).
Administrative districts are grouped into strata (according to sampling theory requirements, strata are supposed to be homogeneous where possible and consist of similar areas) and do not strongly vary in population size. The strata are formed within economic-geographic regions of Russia (at the time the sample was designed, Russia had not been divided into Federal Districts): every stratum includes administrative districts within one region. For Kaliningrad region, which isn't part of the 11 economic-geographic regions, a separate stratum was formed.
The following features are used as characteristics of administrative districts for the formation of strata:
Two or three of the listed features are used for stratification of every economic-geographic region, which are specific and most significant for a particular district. The stratum integrates districts with similar features.
A total of 98 strata have been formed. Two strata substantially exceed the others in population size. These are Moscow (8,389,700 residents) and St. Petersburg (4,728,200 residents). One stratum, which includes only the districts within Kaliningrad region, has the smallest population (951,400). The number of residents in the remaining strata varies insignificantly – from 1,063,500 to 1,681,500. The average size of these strata is 1,387,800 residents, so the margin of bias in stratum size from the mean does not exceed 23.5%.
Nine of the 98 strata include only one district. This can be a district with a large regional center (regional capital), or a large regional center by itself. These 9 strata are self-representing, and no selection is made there at the first stage. Every district (city) that constitutes a self-representing stratum is included in the sample with a probability equal to one and represents only itself. Among self-representing strata, we included the two largest cities, Moscow and St. Petersburg.
From every remaining non-self-representing 89 strata, one administrative district is selected, representing all districts within the stratum. The selection is made with a probability proportionate to the population size in the region (Goskomstat [State Statistics Committee] data from January 1999 was used).
The sample thus includes 98 administrative areas (or towns at the republican, territorial, regional or district levels).
Moscow and St. Petersburg are included in both sample parts, as Russia cannot be represented without these cities. The remaining 96 strata are equally distributed. Each sample part thus includes 50 districts.
The total sample size is allocated across the strata proportionate to population size within the strata, except for Moscow, where the sample size was disproportionately increased to 600 respondents.
2.3. Second Selection Stage
At the second stage, the selection units are polling stations. In every administrative district selected in the first stage, three polling stations are selected, with the exception of Moscow and St. Petersburg, where 16 and 10 polling stations, respectively, are selected. The selection of polling stations is made with a probability proportionate to the station size.
Polling station size is measured by the number of households situated in the polling area. As official data on the number of households in a polling area is not available, it is calculated based on available information. For cities, one of two calculation methods was employed depending on available information.
I. If the number of voters in the station is known, the number of households is calculated as follows:
number of households = (urban population size/number of voters in a town)*(number of voters in a polling station /average size of a household)
II. If the number of voters in a polling station is not known, the following formula was employed:
number of households = urban population size /( number of voters in a town * average size of household)
In large cities that are divided into districts, a calculation was made for each individual district. In rural settlements where the population size and the number of polling stations are known, households were calculated using the same two methods. If the data for individual rural settlements is not available, a calculation is made for the whole district population, which is equal to the selection of rural settlements with an equal probability.
The following data sources were used for sampling. The population of urban settlements (as well as city districts) was taken from the annual Goskomstat bulletin. In our calculations, we used data for January 1, 1999, as at the time we were designing our sample, the data for 2000 had not been unpublished. The number of voters in a polling station area, city or district, as well as the number of polling station in a town or district, was obtained from decrees issued by city, district, region, krai or republican administrations containing a description of the polling stations prior to the presidential (March 2000) or parliamentary (December 1999) elections. The average size of households in the federation subjects was taken from Goskomstat material on the recent micro-census returns (1994). Data on the average household size for individual cities or districts is not available in POF.
The selection of polling stations (with a probability proportionate to a station size) is made using a systematic selection method, with an implicit stratification being employed. Prior to selection, the stations within each administrative district are arranged as follows: First come city polling stations, with cities, if they are several in the district, arranged in decreasing order of population size. Next come rural polling stations. Rural stations and city stations are arranged in order of their geographical location, i.e. bordering stations are listed next to each other (the numbering of polling stations usually corresponds to this order). Systematic selection thus ensures a relatively equal allocation of sampled stations across an administrative district, cities with varying population size, and villages.
A total of 378 polling stations in 202 residencies, including 80 polling stations in Moscow, were selected and sampled.
Weekly Russia-wide polls, conducted with a half-sample, cover 189 polling stations in 102 settlements, including 40 in Moscow.
The planned size of a sample for an administrative district is allocated proportionally between all selected polling stations within the district. From 8 to 11 respondents are interviewed in one polling station area, depending on the district, except Moscow, where 15 respondents are interviewed in every polling station.
2.4. Third Selection Stage
In the third stage, the selection units are households. The selection of households in a polling station area is made by a random route method. The interviewer canvasses households located in the polling station area and selects households for the survey, observing a predetermined interval (selection interval), which ensures random systematic selection. Given this method, the list of households is not made in the office, but is formed by the interviewer himself during the survey.
The application of the route method for regular polls (Russia-wide polls are conducted weekly) is in many ways simplified. The route starting point and the order of canvassing becomes less important. When conducting a poll, the interviewer canvasses only a small number of households in his area. In the next poll, he continues canvassing from where he left off and keeps doing so until all households in a polling station area are included in the route. When the work is finished in one area, the next one is selected (the second stage selection is repeated), and the interviewer moves there. The interviewer normally works in one area for three to six months.
Since the interviewer canvasses all households in the area, each of them has a chance of being sampled.
The Order of Canvassing Households
The order of canvassing households is described in a special guideline and depends on the description of a polling station area. The descriptions of polling stations areas in large cities usually provide street names and house numbers. In this case, the interviewer canvasses houses in accordance with the description. Calling on apartments within one apartment block is done in increasing number order; if apartment numbers are not available, the interviewer moves from left to right up the stairs.
If a street is located within a polling station area, the description may lack house numbers. In this case, the interviewer starts from the head of the street, walking down one side, then the other, without missing any buildings, including those in courtyards.
If the description lacks the street names (this happens sometimes if a polling station area includes the whole residency), the interviewer must list the streets on his own canvass. He can list one or several streets in the residency and update his notes as he moves along. The list drawn up by the interviewer is passed to a poll supervisor along with the questionnaires. If any streets in the area are unlisted, this fact must be noted.
Polling stations where street names are not specified are the most difficult for a random route design. In this case, it is up to the interviewer to chose the order of canvassing. This circumstance is not very important, for in the course of several successive Russia-wide polls, the interviewer must canvass all streets and households in the area, which implies that all of them have a non-zero probability of being sampled.
Household Selection Interval
The route method implies that not every household is selected. Rather, there must be a predefined interval (selection interval) between two successive households. The magnitude of the selection interval affects the distance between various households where the survey is conducted. When the households are close together, there is a higher probability of getting similar answers to the questionnaire, given that the interviewees live in similar conditions, may work at the same enterprise, or belong to the same social group (e.g. buildings populated mostly by employees of a single company, as well as luxury housing). The similarity of interviewees' answers, which depends on the sample frame, augments the sample's statistical error caused by an increase of the design effect.
Increasing the selection interval reduces the correlation of interviewees' answers while complicating the interviewer's job. In POF's weekly Russia-wide polls, a variable selection interval is used depending on the number of households (apartments) per building. The following table shows the relationship of the selection interval and the number of apartments per building:
The interviewer's guidelines contain a set of detailed rules on using the selection interval. It is only after an interview has been successfully conducted that the corresponding household is added to the total number. In case of absence or refusal to answer the questions, the interviewer turns to the next apartment. Thus, the interval between households where an interview was obtained either equals the selection interval or exceeds it by several households. The selection interval is set up so that no more than five interviews can be conducted within an apartment building (except large buildings with more than 100 apartments).
Applying the variable selection interval results in deviations from an equal probability of household sampling. Small apartment buildings have more chances of being included in the sample. However, in practice, these deviations are insignificant, since a single area often contains buildings of the same type.
Picking a Route Starting Point
When the interviewer first comes to an electoral district, the first building in the district description becomes his starting point. In case the district is being revisited with a survey, the interviewer starts at the household where the last interview was taken. This is the household to which the selection interval is added. Practically every household gets an opportunity to be included in the sample, since the selection interval depends on the number of households in a building, so that in case interviewees are unavailable, it becomes equal to one.
2.5. Picking an Interviewee in a Household
One person per household is picked to be interviewed. The interviewee is chosen according to a given quota. There is an interlocking quota for age and sex and a separate one for education (higher or otherwise).
Quotas are used in the selection of interviewees because in the weekly Russia-wide polls, only two days are dedicated to fieldwork. It is impossible within this short period to properly revisit interviewees who were unavailable when visited the first time. Repeated visits are particularly difficult in remote localities where interviewers must travel. Under such circumstances, unavailability reaches an average of 50%, and sampling deviations tending towards an increase in the percentage of retired people and women become systematic. Therefore, the use of quotas in the last stage of selection is inevitable, given the little time available, not to mention any financial limitations. Nonetheless, experience shows the results attained in this way are quite acceptable. The main negative aspect in the use of quotas is deviation from the principles of the probability sample design.
2.6. General Sampling Parameters
The full Russia-wide sample comprises 203 urban and rural localities from 63 federation subjects. The sample includes 378 electoral districts, including 80 Moscow electoral districts. The entire sample size is 4028 people. Of them, 3000 represent a proportional Russia-wide sample (including 172 people residing in Moscow), and 1028 make up an additional sample for Moscow.
The full sample consists of two parts, or half-samples. Each half-sample independently represents the population of Russia. The weekly Russia-wide population polls are carried out in turns on half-samples. Each half-sample includes Moscow and St. Petersburg, which are separate federation subjects. In addition, the first half-sample comprises 102 localities from 42 subjects, while the second comprises 98 localities from 44 subjects. Each half-sample includes 189 electoral districts, including 40 in Moscow. The planned size of a half-sample is 2014 people. Of them, 1500 represent a proportional Russia-wide sample (including 86 people in Moscow), and 514 make up an additional sample for Moscow. As a result the, the total half-sample for Moscow is 600 people. Except for Moscow, 8 to 11 people are interviewed in every electoral district. In Moscow, 15 people are interviewed per district. This proportion is kept both for the full and the half-sample.
Description of the GeoRating Project Sample
Population Polls in 69 Federation Subjects
Polls are conducted in 69 federation subjects. Interviewees are 18 years of age or older. The sample size in every subject is 500 interviewees.
In all 69 subjects, the general principles of sample design were applied. Household sampling was done using territorial three-level stratified sampling. Households were selected in three stages. Administrative districts were picked in the first stage, polling stations in the second, and households in the third.
1. First Selection Stage: Administrative Districts
Stage 1 Sampling Units
Administrative units make up the basis of the Russian Federation's organizational division. They also include cities of republican, krai, region, and okrug subordination. According to all Goskomstat bulletins, such cities are regarded as immediate federation subjects with no subordination to the region where they are located.
The sampling is based on the territorial principle. Its design is largely affected by the geographical location of a given polling station. This is why, prior to the first selection stage, all cities of republican, krai, region, and okrug subordination were placed together with their corresponding region. As a result, first stage selection units were established: these are administrative districts, including all settlements located on their territory.
In some Federation subjects, there are cities and towns that cannot be considered to belong to a single district, which are located on the boundary between two or more districts (they are most numerous in the Sverdlovsk region). Such cities and towns are included in the sampling independently and are not considered different from districts.
Sometimes a city and localities subordinated to it occupy an extensive territory comparable to a district. The boundaries of such territorial formations are even reflected on maps (in Murmansk region, for instance). Such cities with their subordinated localities are regarded as separate selection units.
The Minimum Number of Interviews per District
The number of administrative districts and other selection units regarded as equivalent to them varies according to the federation subjects: the Jewish Autonomous Region has 5 such districts, whereas the Altai krai has 60. In federation subjects with few districts, all of them make up the sample. Where districts are numerous, a selection is performed. It must be decided in each case how many districts are included in the sample for each federation subject. This decision is based on two principles:
First, the minimum number of interviews per district is defined. The smaller this number, the more accurate the poll results will be. Yet the cost of the poll grows accordingly: expenses needed to send an interviewer to a district are divided by the number of interviews in the district, thus determining the average cost of an interview.
Obviously, it is inexpedient to send an interviewer to a district for just 3 or 4 interviews. There is a specialized technique that allows calculation of the optimal number of interviews per district. That number assures the accuracy results for a given sample size. In the sample design for the GeoRating study, the minimum number of interviews per district was defined as 20 (4% of the federation district sample size). Yet for some districts, exceptions were made, so that their sample size was 10 or 15 interviews.
Selection of Districts
Districts in federation subjects were selected according to the following scheme. First a sample size was calculated that would correspond to each district if the sample were distributed among all districts proportionally to population size. Districts where no less than 4% of the federation subject's population reside and whose share of respondents is 20 or more were automatically included in the sample.
Prior to the selection, the districts were broken down into small groups. The main criterion was the proportion of the urban vs. rural population. (Numerous studies demonstrate the distribution of respondents' reactions to all kinds of questions correlates most of all with this parameter.) The districts' geographical location and distance from the center were also considered. In some federation subjects, the ethnic composition of the districts' population was used as a stratum-forming feature.
As a result, districts that were similar according to these parameters were grouped into strata. And since the selection was conducted on a per-stratum basis, all types of districts are represented in the sample. Districts were selected in the strata with a probability proportional to their population size. This selection method guarantees equal probabilities for each household to be included in the sample. The number of selected districts per stratum depended on the stratum size. It was determined by the principle: 20 respondents (4% of the sample) per every 4% of the federation subject's population.
The 69 federation subjects where the study took place contained 1648 districts. Among them, 850, that is, more than a half, entered the sample. In addition, the selection included 62 cities of republican, krai, region, and okrug subordination, as well as the cities of Moscow and St. Petersburg, as independent selection units.
2. Second Selection Stage: Polling Stations
In each district included in the sample at the first stage, polling stations were selected. The principle behind the polling station selection on the whole followed that of district selection. Every polling station had a minimum number of 10 interviews assigned to it (although for some polling stations, it was reduced to 5). Sampling was made with a probability proportional to the population size. Localities with 10 and more interviewees, according to population size, were included automatically. In other localities, random selection was used.
The sample size in large cities depended on the percentage of the entire population of the federation subject residing in the city: one percent of the population corresponds to a sample of 5 people. The sample size thus calculated for a large city was rounded off to 5 in order to facilitate the interviewers' work. Sampling of urban polling stations presented no difficulties, thanks to Goskomstat's bulletins, which are published annually on the population of all cites.
It was harder to sample villages, because POF lacks a full list of population statistics for them. This is why villages located in the districts included in the first stage selection were selected mostly at the regional level. When reliable information on rural population size was available, they were selected with a proportional probability. If there was no such information, selection was done with an equal probability. An extra stage, that of rural electoral district selection, was frequently used. Electoral district lists are more often available and are based on relatively up-to-date data. Electoral districts do not vary substantially in size, so they can be selected with an equal probability. A rural electoral district usually consists of one large or several small villages. After the electoral districts are selected, polling station sampling is not a difficult task.
In large cities, where the sample size was several tens of people, an extra selection stage for electoral districts was also conducted. This makes it possible to systematically distribute the sample in the city area. In each electoral district, 10 respondents were interviewed.
The total number of polling stations in a sample equals 1954. They comprise 67 republican, krai or region centers, including Moscow and St. Petersburg (which are not only separate federation subjects, but also subjects of the Moscow and Leningrad regions), 609 cities and towns, 315 urban-type settlements and 963 villages. Note that among the 2709 urban localities in 69 federation subjects, 991 were included in the selection (i.e. more than a third). 3. Selection of Households
The selection of households in a polling station area was made by the random route method. According to this method, every interviewer is given a route that he must follow to select households to be interviewed. Usually, the route follows a street selected randomly from a complete list of streets in the locality. In villages with only one street, the route follows that street.
Households are selected observing a predefined interval (selection interval). The selection interval is defined to avoid a situation where the households interviewed are all concentrated within one small area, so they should occupy a certain part of the locality's territory. On the other hand, the interviewer must not walk more than 1 to 1.5 km while following the route. The selection interval depends on the type of buildings prevailing in the area. In cities with large apartment buildings, it is defined as several tens of households. In cities with mainly private houses, as well as in villages, the selection interval is 2 to 4 households.
The route assignment given to the interviewer includes:
In cities where the sample size exceeds 10 interviewees, several routes are defined in different parts of the city. If an additional selection stage is conducted (electoral district selection), routes are defined within the districts. In this case, the route must include not one, but several streets within a single electoral district.
The route method implements the standard random systematic selection procedure. All steps the interviewer takes to select the households are strictly regulated and can be verified. The interviewer cannot select a household at his own discretion, but can only call on households defined in the route assignment.
4. Selection of Interviewee in a Household
One person per household was picked to be interviewed. The interviewee was picked according to a given quota. There was an interlocking quota for age and sex, and a separate one for education (higher or otherwise). The age quota comprised three gradations: a) 18 to 24, b) 25 to 54, c) 55 and older. The education quota comprised two gradations: a) higher or incomplete higher education, b) secondary education or less. 5 General Sampling Parameters
Polls were conducted in 69 federation subjects, which comprised 135, 471,100 people, i.e., 93,3% of the total population of the Russian Federation. Interviewers were 18 years old and above. The total sample size was 34500 respondents, 500 respondents in each federation subject.
The sample encompassed 1954 localities, among them 676 cities and towns, 315 urban-type settlements and 963 villages.
The statistical error for each subject did not exceed 5.5%. Statistical error for the total result in all 69 subjects did not exceed 1%.
Telephone Surveys
Telephone surveys imply immediate computer processing of the data. The advantage of this method is that it is much cheaper (about half as expensive) and faster than population polls conducted in a residency or place of work.
Currently, POF is implementing a new telephone survey project to survey experts in a completely automatic mode with the possibility of respondent registration (CATI).
Telephone Polls in Moscow (TELEFOM)
Telephone polls are conducted to study Muscovites' attitudes on political parties, movements, or leaders, and on issues related to city life.
To build a sample, a random digit generator is used. A computer picks out prefixes (the first three digits of phone numbers) for district automatic telephone exchanges from a complete list of greater Moscow exchanges and randomly adds four digits to each prefix. Repeats are eliminated from the list of numbers before it is given to interviewers.
Polls are conducted on Mondays and Tuesdays over the whole day, using office and home numbers.
Respondents interviewed are people 18 or older who reside in Moscow. Their sex, age and education are also checked so the sample frame is in line with the demographic makeup of Moscow's adult population.
Questions concern presidential, parliamentary and Moscow mayoral elections. Since November 1999, the questionnaire has included five questions on other topics.
Up to fifty interviewers are active in such polls, each of them conducting about twenty interviews.
EXPERT POLLS
Experts are selected on the basis of competence. The expert group's size and representativeness depend on qualitative rather than quantitative parameters.
Expert polls are not anonymous and they presuppose the respondent's active participation in the discussion of issues. Experts are allowed to express their opinions at length.
Basically, an expert poll serves to refine hypotheses, make forecasts or amplify the interpretation of phenomena under study. Open-ended questions dominate such polls, and multiple-choice answers are only used to estimate the level of certitude or agreement and disagreement with the opinions of other experts.
Data Source
Phone interviews with representatives of regional legislative and executive bodies, regional media, scholars, professors, NGOs, private researchers and advisers, expert council members, etc. .
POF publications present the results of some 100 formalized and 30 standardized interviews conducted weekly with 130 experts from 28 regions of Russia. .
.
Experts Types
There are four types of experts:
Type 1
Regional political elite
This group comprises representatives of provincial legislative and executive government bodies.
Type 2
Local political elite
This group comprises representatives of provincial capital city legislative and executive government bodies.
Type 3
Intellectuals
This group comprises university professors, scholars, NGO representatives, and private researchers and advisers (lawyers, auditors, pollsters, PR companies, etc.). These people are not members of local legislatures or government, but they may serve as members of expert boards, think-tanks and other task groups that assist the authorities.
Type 4
Journalists
This group comprises people working for regional media, such as TV and radio stations and periodicals, who may be loyal to the regional authorities, neutral, or in the opposition.
Experts are selected by POF regional branches according to special instructions.
Polls are conducted using an automated telephone interview system (CATI) in several forms:
- a blitz-poll, with experts answering questions by selecting answers from readymade sets (formalized interview);
- free conversation with topics suggested in the form of open-ended questions (standardized interview).
The board of experts covers 28 regions of Russia. The regions of Belgorod, Bryansk, Vladimir, Volgograd, Vologda, Voronezh, Ivanovo, Kaliningrad, Kaluga, Kemerovo, Nizhny Novgorod, Orenburg, Penza, Rostov-on-Don, Ryazan, Samara, Saratov, Sverdlovsk (Yekaterinburg), Tula, Ulyanovsk are represented, as well as Altai, Krasnoyar, Krasnodar, and Stavropol krai, the city of St. Petersburg, and the republics of Bashkortostan, Tatarstan, and Udmurtia.
In-Depth Interviews
An in-depth interview is a face to face conversation between an interviewer and respondent. It is used in:
- expert polls;
- questionnaire preparation for population polls (non-directed interview);
- analysis of poll data to determine the motivation behind answers and broaden the set of answers (directed interview).
Method
In-depth interviews serve to uncover respondent's underlying motives. The topic for discussion is given by the interviewer, and the conversation loosely follows a preset topic-guide, with question order and wording typically depending on what the respondent says.
The important thing about this method is that unlike focus group discussions, the respondent's statements are not influenced by others.
DISCUSSION FOCUS GROUPS (DFG)
A focus group involves a discussion of issues by a group of respondents.
Field of Application
POF applies the focus group method in:
- political marketing (image analysis for political leaders, parties, movements, organizations, and their programs);
- marketing of goods and services (study of advertising efficiency);
- marketing of non-profit organizations (efficiency evaluation of social projects and programs, pilot test of hypotheses concerning phenomena new to society);
- development of methods and tools for collection of large files of primary social information;
- the processing of poll data and analysis of findings; here, focus groups help elaborate the motivations behind certain answers and broaden the range of answers.
Technique
Ten people participate in the Discussion Focus Group (DFG) in each city. They include men and women who voted either for Vladimir Putin, Gennady Zyuganov, or some other candidate, or `none of the above' in the last presidential election.
The DFG is a roundtable guided by a moderator who has a topic-guide that is not shown to the discussion participants. The conversation is taped and transcribed records make up the basis for the report.
After an introduction and announcement of the issue, it is moderator's task to make all participants contribute to the discussion without going off-topic.
Group discussion actualizes the associative links in participants' minds. Because they communicate with each other, and not just with the moderator, new information can be obtained that is often impossible to secure in a face-to-face interview. The duration of focus groups normally does not exceed two hours.
Participants and Their Selection
The number of focus groups to be conducted and the number and composition of the participants depend on the research task. One study typically requires two to eight focus groups with eight to twelve participants, each. In POF focus groups, we usually have eight to ten participants.
Depending on the research tasks and customer demands, respondents can be selected by any socio-demographic criteria. They are recruited from the POF polling database, as well as by means of a snowball sampling.
Research Tasks
Issues discussed in DFG include the most topical Russian events of the past week that were covered on television. The aim is to explore the public's reaction to these events.
DIAL-METER FOCUS GROUPS (DMFG)
Dial-Meter Focus Groups (DMFG) – are a method of studying respondents' opinions and judgments through the use of modern electronic equipment.
Fields of Application
Television:
- analysis of TV broadcasts to discover the pattern of `strong' and `weak' spots in them;
- viewer rating forecasts to determine `prime' moments in time.
Advertising:
- analysis of television commercials under various aspects (spots can be shown several times);
- perception analysis of advertising posters.
The testing of design solutions:
- comparing different design solutions in architecture, fashion, etc., including those still under development;
- evaluation of sections of video and computer presentations.
Political consulting:
- public speaking coaching for candidates;
- fine-tuning of political advertising;
- campaign blitz-analysis and correction in the run-up to elections;
- analysis of political leaders and candidates' taped speeches and debates. Determining spots of high and low ratings helps improve the contents and style of speeches.
Potential customers:
- producers and managers of TV production companies;
- research departments of advertising agencies;
- advertising company executives, analysts and market researchers;
- image-makers;
- advisors.
Technique
Participants in a DMFG normally number 40-48 (a minimum of 20, maximum up to 100) people of both sexes, representing different ages, education levels and political views. Depending on the research tasks, they are selected at random or by certain criteria.
Each DMFG participant is provided with a five-point rating device to rate what they see on the screen.
A computer continuously takes and records the readings from the devices, systematizes and generalizes them and matches them to the video material.
Electronic registration of participants makes it possible to divide the audience in groups according to the research tasks. After the moderator instructs them and sets them their tasks, the video to be tested is shown, and participants use their devices to express their reaction to what they see in terms of like/dislike, interest/disinterest, agreement/disagreement, and so on.
Participants and Their Selection
Respondents are recruited from a constantly refreshed database of regular POF poll participants or by means of a snowball sampling. Their socio-demographic composition depends on the research task. Equal representation of all presidential candidates' supporters is a key requirement.
After the DMFG is finished, a discussion takes place in which five male and five female respondents participate, among whom are supporters of Putin, Zyuganov, another candidate, and `none of the above'. This interview makes it possible to disclose the motives behind the evaluations and compare verbal and nonverbal reactions.
During the DMFG, participants are shown the same video three times and asked to use their rating devices to register the degree to which they find the episodes interesting/uninteresting, whether they agree/disagree with what they see, and finally how much they trust/distrust it.
Representation
Signals from the rating devices are synchronized with the videotape. They are transformed into curves that are shown (individually or in various combinations) over the picture, visualizing the changing reaction of the audience.
For analysis, the video material is cut into fragments according to the ups and downs of averaged rating curves. The reasons for reaction change are derived from the episodes' contents, which form the core of the analytical commentary presented to the customer along with the videotape.
The commentary contains tables showing the averages for liking, agreement and trust parameters in each of the four subgroups in the audience (i.e. Putin voters, Zyuganov voters, other candidates' supporters, and `none of the above' voters).
Oscillations (indicated as strong, mean, or weak) show the difference between the minimum and maximum values of the three curves.
ÒV MEASUREMENT
The analysis of TV broadcasting is based on evening (6 p.m. to 1 a.m.) broadcasts containing information and news round-ups on the three leading national channels: ORT, RTR and NTV. Broadcasts are divided into one minute segments, and each segment is regarded as a separate broadcast with an audience of its own.
For every segment, a rating can be calculated based on the numerical relation of its audience to the population. The measure of TV impact is GRP (Gross Rating Points) describing the `accumulation' of an audience.
A week's totality of segments that mention or feature one political figure forms a `quasi-campaign', which affects viewers by reminding them of the politician's name and attracting attention to him or her. The GRP of this totality shows the intensity of the impact of the `quasi-campaign' on the audience.
TV-viewing data is collected by Gallup Media using peoplemeters, devices that register viewing behavior. Peoplemeters are installed in 1200 homes in the eighteen largest cities of Russia. These families, a universe of about 3200 persons, are representative of large Russian cities with populations of 400,000 or more