Sample and Data Collection
The Opinions survey uses what is known as random probability sampling broken into Government Office Region, proportion of households with no car, socio-economic classification and the proportion of people aged over 65 years.
In common with most other ONS surveys, the Opinions survey draws its sample from the Royal Mail's Postcode Address File (PAF) of 'small users'. This PAF contains the addresses for approximately 27 million private households in the UK which receive fewer than 50 items of mail per day. It is the most up-to-date and complete address database in the UK. Each month 67 postal sectors are selected, with probability of selection proportionate to size. Within each sector, 30 addresses are chosen randomly giving a final sample of 2,010 addresses each month. At the start of the interview, the interviewer determines the household composition and then he or she selects the respondent from among all those aged 16 and over. This selection is performed at random using a Kish grid.
All household members over the age of 15 years are asked a set of classificatory questions but only one person per household is selected to answer the Opinions module questions. Proxy responses are not permitted on the Opinions modules. As only one person per household is interviewed, the data are subsequently weighted to correct for the unequal probability of selection that this causes. Applying these weights also gross up the data by age, sex and region to the population control totals used on the Labour Force Survey (LFS). As well as accounting for the unequal probability of selection, these weights correct for certain types of non-response bias and improve precision for most variables. Weights will be supplied at person level in each survey month and, if required, will be available at a household level and on a quarterly or annual basis.
This method of sampling and the consequent weighting affect the sampling errors of the survey estimates. The effect can be shown by calculating the Effective Sample Size which gives the size of an equal probability sample, equivalent in precision to the unequal probability sample actually used. The Effective Sample Size will vary slightly from one month to another with the proportions of interviews in different sized households. On average the Effective Sample Size of the Opinions survey is 84 per cent to 86 per cent of the actual sample of individuals. An achieved sample of 1800 individual adults in the Opinions survey is equivalent to an equal probability sample of about 1500.
When the sample for a particular month is selected by the ONS’s Sampling Implementation Unit, a letter is sent out to the sampled address approximately one week before the start of the field period. This letter explains the purpose of the survey and advises the residents that an interviewer will be calling regarding the survey. As an incentive to respond to the survey, a book of postage stamps is included with the letter. The interviewer is required to call at all sampled addresses to ensure that they are residential and not business addresses. Businesses, institutions, temporary accommodation and vacant addresses are ineligible in most instances. The Opinions Survey interviews are conducted using Blaise questionnaire programming tool on a laptop and, in most instances, the interviewer enters the respondents' replies directly into the computer. If the questions are considered to be sensitive then respondents are given an option of self completion.
From beginning to end the survey period lasts 14 weeks. This includes four weeks for developing and testing the questionnaire in Blaise (a computer-assisted survey design and processing system). The survey will be in the field for five weeks, including time for telephone reissues (the Office for National Statistics (ONS) Telephone Unit re-contact initial non-contacts and refusals). During the final five weeks of the survey period the data will be cleaned, weighted, analysed and tabulated by the research team.
