The Scottish Government

« Previous | Contents | Next »

Listen

Public Attitudes to the Environment in Scotland

CHAPTER ONE RESEARCH DESIGN AND METHOD

1.1 INTRODUCTION

The Survey of 'Public Attitudes to the Environment in Scotland' took place between February and June 2002 when four thousand adults were interviewed in private households throughout Scotland. The fieldwork was undertaken by George Street Research Ltd., using paper-based questionnaires.

The survey was financed by the Scottish Executive, in conjunction with the Forestry Commission and Scottish Natural Heritage. It was commissioned to provide nationally representative information about the views of the Scottish public towards a wide range of environmental topics with particular policy relevance in Scotland. Some areas covered in the 2002 survey were also included in a Scottish survey of public attitudes to the environment undertaken in 1991 ¹ and thus some comparisons with this study are drawn ².

This main report of the survey findings follows publication of preliminary research findings in November 2002 ³ and offers more detailed analysis of the survey data. The data will also be available from the UK data archive, accompanied by a full technical report for those who wish to undertake secondary analysis ⁴.

This introductory chapter provides information about the survey content, sampling, response and weighting and describes the structure of the report and reporting conventions. As a backdrop to subsequent chapters, characteristics of the sample are also discussed.

1.2 SURVEY DESIGN

1.2.1 Sample and interview design

The 2002 survey was designed to be representative of the Scottish population over the age of 16 and living in private households in Scotland. It was also designed to allow reliable comparison of responses for respondents in rural and urban areas. Therefore, people living in rural areas were over-sampled but figures for Scotland as a whole are weighted back to show the distribution of the population in its true proportions.

A total achieved sample of around 4,000 respondents was sought in order to allow a split interview design, whereby 2,000 respondents were asked one version of the questionnaire and 2,000 respondents were asked a second version of the questionnaire. Some elements were common to both questionnaires. The questionnaire was designed following extensive consultation with policy-makers in the Scottish Executive and key representatives from Scottish Natural Heritage and the Forestry Commission. Some questions duplicated those asked in the 1991 survey so that comparisons might be made.

Table 1.1 shows the topics asked of respondents answering versions A and B of the questionnaire. It also shows the chapter where findings on each topic are reported.

1.2.2 Sampling strategy

Households were selected at random from the Postcode Address File of residential addresses (PAF) using a multi-stage, stratified sampling approach. The first level of stratification was based on the six-fold rural/urban location definition developed for the Scottish Household Survey which combines settlement size and drive time from larger settlements (see Table 1.2). Rural categories were deliberately over-represented to ensure adequate coverage of these areas. The effects of this are apparent from Table 1.2. A second level of stratification was undertaken, based on the Scottish 'MOSAIC code'. This classifies households into distinct 'lifestyle types' which describe their socio-economic and socio-cultural behaviour ⁵. This stratification ensured that the sample was, as far as possible, representative of the population across Scotland.

After stratification, 800 sample points were randomly selected and 400 were assigned to each of sample A and sample B - denoting which questionnaire would be administered at each sample point. A random sample of eight addresses (and two substitute addresses, in the event of any of the original addresses being unsuitable ⁶) were then drawn for each sample point and issued to the interviewers. Thus, although the sample was clustered to reduce fieldwork costs, the effect of clustering on the results was minimised by the large number of sample points. The sample points were spread across the whole of Scotland as shown on the map in Appendix B.

At addresses with more than one adult, one person was selected at random, using a 'kish grid', and asked to take part in the interview.

1.2.3 Response rate

Face-to-face paper-based interviews, lasting about 40 minutes, were held in respondents' homes. Interviews were achieved with 65% of eligible respondents identified through the sampling process described above. Table 1.3 details the survey non-response. It can be seen that this comprised 15% of selected households where no contact could be made with an occupant and 9% where the initial approach resulted in a refusal to take part. In a further 4% of cases, no respondent selection was made, but the reasons are not known. At 72.5% of households, a member was selected to take part in the interview, however 3% of these individuals refused to take part and others were not interviewed because they were too busy to arrange a suitable time or did not keep an appointment (2%). A few of the selected respondents were away or ill for the duration of the fieldwork period (1%).

1.2.4 Weighting

Data used in analysis have been weighted to correct for both the over-representation of rural areas and for non-response. The weighting variable used in analysing the survey data was made up of three different components.

The first component was related to the sample design and corrected for the fact that households and individuals in the survey were selected with unequal probabilities. This was due to the rural boost, and the fact that some households contain more adults than others, but only one adult per household was selected for interview. The second component of the weight corrected for evidence within the sample of differential response rates by households and individuals and the third used post-stratification to match the sample to national population estimates (of age and sex distribution). The weighting variable was normalised to sum to the total sample size of 4,119.

The weighting did not vary according to whether questionnaire A or B was administered.

1.3 REPORTING CONVENTIONS AND TYPES OF ANALYSIS

This report describes the findings of the survey and relates findings to respondents' socio-demographic characteristics, in particular sex, age, housing tenure type, highest educational qualification and whether respondents lived in an urban or rural location.

The survey did not contain a measure of social class, and although household income was recorded, the number of missing cases was too high to permit analysis using this variable. Therefore, tenure type has been adopted in analysis as a proxy indicator for social class or income ⁹. Where those in 'social rented housing' are contrasted with 'owner occupiers', it is not necessarily because tenure type is felt to be of particular interest, but because those in social rented housing are likely to represent lower income groups in society.

Tables showing the characteristics of the sample are shown in section 1.4 (Tables 1.4 to 1.10).

1.3.1 Reporting conventions

Each chapter contains a series of tables containing the data to which the text in the chapter relates. The following conventions have been used within tables and in the text.

• All data presented in tables are weighted, but unweighted sample sizes are given - to show the true base to which each percentage in the table relates.

• Very small bases have been avoided wherever possible because of the relatively high sampling errors that attach to small numbers. In general, percentage distributions are shown if the sample size is 50 or more. Where the base is smaller, the percentages are shown in square brackets [].

• Where there are no cases in a particular cell of a table, a dash '-' is shown.

• Where there are cases in a cell, but they represent fewer than 0.5% of the sample, '0' is shown.

• Figures are rounded to the nearest whole number. Due to rounding, columns or rows which present percentages for a full distribution may sum to 99% - 101%.

• All differences reported in the text have been tested and found to be statistically significant at the 95% confidence level.

1.3.2 Logistic regression analysis

Within many chapters of the report we have used logistic regression models to assess whether there is reliable evidence that particular variables are associated with each other. Regression analysis aims to summarise the relationship between a variable of interest, for example frequency of countryside use (the 'dependent' variable), and one or more explanatory variables, such as household car availability (known as 'independent' variables). It shows how well we can estimate a respondent's score on the dependent variable from knowledge of their scores on the independent variables. The technique takes into account relationships between the different independent variables (for example, between educational attainment and age or tenure type) ¹⁰. Logistic regression is used to model variables which are dichotomous (that is have just two response categories) - for example, recycling or not recycling. An example can be found in Chapter 4 where the technique is used to investigate the characteristics of those who recycled paper and those who did not. A number of characteristics which might explain this were included in a model (age, sex, tenure type, level of education, urban or rural location and access to a car). Although simple cross-tabulations suggested that all of these characteristics were associated with recycling, the fact that many of them are also associated with one another means that it was not possible to assess whether their association with recycling was 'real' or spurious. In this case the model identified all the variables except urban or rural location as having an independent significant effect on recycling (see Figure 4.7). Put another way, this means that once the other variables had been taken into account, there was no longer a significant relationship between recycling paper and urban or rural location although the relationship with all the other explanatory variables remained significant.

Full technical details of logistic regression can be found in many textbooks on social statistics, for example Bryman and Cramer (1997) ¹¹.

1.4 CHARACTERISTICS OF THE SAMPLE

When interpreting findings by variables such as age, sex, highest educational qualification, family type ¹² and tenure type it is useful to have an understanding of how the variables are themselves inter-related. The following section describes the survey respondents in this way (using weighted profiles). The related tables are presented at the end of the section (Tables 1.4 to 1.10).

1.4.1 Sex

Just over half (52%) of all respondents were women. Women were more likely than men to be represented in the oldest two age groups. Among those aged 16-24, 49% were women while among those aged 65+, 60% were women. Women were also more likely than men not to have any qualifications. While 50% of men had a qualification of Higher level or above, fewer than four in ten women did. Women represented six in ten people living in remote small towns while in other types of areas there were not notable differences between the sexes.

1.4.2 Age

Those aged 65+ were more likely than their younger counterparts to be female, have no educational qualifications (54% had none), to live in social rented housing and to live in a household without a car (half had no access to a car).

Those in the youngest age group (16-24) were most likely to have some educational qualifications (only 7% did not) and to live in the private rented sector - a quarter did, compared with fewer than one in ten of those in other age groups. The youngest group were notably more likely than those in other age groups to live in large urban areas - over half did - and particularly unlikely to live in remote rural areas. Over four in ten of those aged 16-24 lived in large adult households, which would include those living with parents and other siblings at home, and those living in house- or flat- shares while studying or working.

Those aged 25-64 were more likely than their older and younger counterparts to have degrees or professional qualifications and to live in owner-occupied housing. Eight in ten of those aged 25-64 had access to a car.

1.4.3 Highest educational qualifications

Almost six in ten of those with no qualifications were women (57%), and three quarters were aged 45 and over. Almost half of those with no qualifications lived in social rented housing, as did just 5% of those with degrees or professional qualifications. Eighty per cent of those with degrees or professional qualifications lived in owner occupied housing.

A quarter of people living in accessible rural areas had a degree or professional qualification. At the other extreme, only 14% of those in remote small towns had such qualifications.

Access to a car increased substantially with education - rising from 55% of those with no qualifications to 88% of those with a degree or professional qualification.

1.4.4 Tenure type

Four in ten private renters were aged 16-24, while this age group comprised just one in ten of those in social rented or owner-occupied housing. Over seven in ten people aged between 45 and 64 years were owner occupiers, and a quarter of this group were social renters. A third of people in the oldest age group lived in social rented accommodation.

Almost half of those living in social rented housing had no qualifications (45%), compared with 20% of owner occupiers.

Owner occupiers were far more likely to have access to a car than renters (88% compared with around five in ten of those in the other two main tenure types).

1.4.5 Urban or rural location

Rural dwellers were less likely than those in urban areas to be social renters (17% of those in accessible rural areas and 18% of those in remote rural areas lived in social rented housing compared with 30% of those in large urban areas). While six in ten of those living in large urban areas had access to a car, this rose to almost nine in ten of those living in both accessible and remote rural areas.

Single adults comprised one in five of those in large urban areas but only 7% of those in remote rural areas. Families were more prevalent in accessible and remote rural areas than in other types of location.

Those living in accessible rural areas were most likely to have degrees or professional qualifications.

« Previous | Contents | Next »