The Scottish Government

« Previous | Contents | Next »

Listen

Changing Speeding Behaviour in Scotland: An evaluation of the 'Foolsspeed' campaign

CHAPTER TWO: EVALUATION METHODOLOGY

2.1 METHOD

Foolsspeed was evaluated using a three-year longitudinal survey with a sample of 550 drivers in the west of Scotland. A baseline survey was conducted in October 1998, before the start of the Foolsspeed campaign, and took measures of respondents' demographic and driving characteristics and of the full raft of Theory of Planned Behaviour components. Follow-up surveys with the same sample were conducted in Spring 1999, Spring 2000 and Summer 2001. These follow-up surveys repeated the measures taken at the baseline survey, and also assessed response to the Foolsspeed campaign in terms of awareness, recall, comprehension, identification, involvement and perceptions of key messages.

Data were collected using a structured questionnaire administered face-to-face in respondents' homes. The administration procedure allowed the use of a fairly sophisticated questionnaire (necessary for the full measurement of TPB components), and also enabled interviewers to use prompt materials and showcards to assess awareness of and response to Foolsspeed campaign elements. Interviewing was conducted by the CSM's fieldforce of professional market research interviewers, under the supervision of the CSM's Research Co-ordinator. Interviews lasted between 20 and 40 minutes. A 10% quality check was conducted on all successful interviews to ensure that the questionnaire was administered according to instructions.

Before recruitment for participation in the baseline survey, interviewers advised respondents that the survey would last just under three years and would involve participation in two further surveys. To encourage ongoing participation and to minimise attrition, personalised mailouts were sent to respondents early each year to inform them about the next fieldwork exercise. A 5 gift voucher was used at the 3 ^rd and 4 ^th surveys as an incentive to boost response and to thank respondents for their participation.

The baseline questionnaire was thoroughly piloted. Four focus groups were conducted to develop an understanding of speeding from drivers' perspectives, to explore beliefs and norms about speeding, and to identify any further issues to be addressed in the questionnaire. Following the focus groups, a draft questionnaire was developed. Eight interviews using the draft questionnaire were then conducted by members of the CSM's market research interviewer fieldforce and observed by a CSM researcher. At the end of the questionnaire interview, the CSM researcher conducted a 30 minute semi-structured interview with each respondent to examine issues such as comprehension, interpretation of questions, ease of completion and so on. The questionnaire was then further revised. This approach ensured that the questionnaire was both relevant and comprehensible in terms of question wording, routing instructions and use of prompts. The respondent samples for the piloting exercises were recruited to match the full sample's demographic profile and driving characteristics (i.e. all respondents were in possession of a current driving licence and made at least one journey per week by car).

Questionnaire Measures

The baseline questionnaire took measures of driver attitudes, subjective norms, perceived behavioural control, intentions and reported speeding behaviour, in accordance with the Theory of Planned Behaviour (see below). The questionnaire also included questions about driving behaviour such as mileage, length of time qualified, and frequency, purpose and destination of journeys. This allowed the possibility of analysing, at a future stage, variations between different types of driver (e.g. social drivers versus those who drive for a living, heavy road users versus light road users, newer drivers versus more experienced drivers, and so on), as well as between different socio-demographic groups.

The follow-up questionnaires repeated the majority of the baseline questions (some driving characteristics questions, such as mileage and vehicle type, were omitted at the follow-up surveys in order to keep the interview length to under 40 minutes). The questionnaires also included a series of questions relating specifically to the Foolsspeed campaign. These examined recall and sources of awareness of the campaign, understanding and perceptions of the campaign's key messages, and response to the campaign in terms of identification and involvement.

Operationalisation of the TPB

The Theory of Planned Behaviour deals with particular behaviours rather than general classes of behaviours (Stradling & Parker 1996). Where the model has been applied to driving violations such as speeding, researchers have therefore usually enquired about behaviour in a specific scenario rather than about speeding behaviour in general (Stradling & Parker 1996). In the Foolsspeed survey, respondents were presented with the scenario below (adapted from Parker et al 1992) and asked to make judgements about their own views and behaviour in such a scenario:

You are driving in your car or van down a road in town. There are some shops and parked cars. It is about 2 o'clock on a fine dry afternoon. There are no other cars driving on the road. The speed limit is 30 mph.

The TPB requires that all measures be taken at the same level of specificity, therefore the corresponding intentions, attitudes, norms and control questions were all asked in relation to this scenario.

Behavioural intentions were measured using three statements: 'I would probably drive faster than 30mph myself in this situation', 'I would never drive faster than 30mph in this situation', and 'In this situation I would want to driver faster than 30mph'. Reliability analysis showed that these items were highly consistent (alpha=0.81).

Eight behavioural beliefs about the possible consequences (both good and bad) of speeding in the scenario described were used in the survey. Speeding was defined as driving at 40mph in the situation described. The eight behavioural beliefs were identified from the questionnaire piloting exercise and also from other studies (e.g. Stradling & Parker 1996, Parker et al 1996, Lawton et al 1997). Respondents were asked to indicate how likely or unlikely they judged each of the consequences (e.g. 'If I drove down this road at 40 mph I would find it difficult to stop in an emergency'). Reliability for the behavioural belief items was relatively high (alpha=0.71).

Outcome evaluations were measured using eight statements corresponding to the behavioural belief statements. For these, respondents were invited to indicate how desirable or undesirable each outcome would be (e.g. 'Finding it difficult to stop in an emergency would be…'). These two sets of items were summed using the TPB formula to produce a composite attitude towards the behaviour. Reliability for the outcome evaluation items was adequate (alpha=0.60).

Normative beliefs were measured with eleven items in which respondents were asked to indicate how much various significant others (or 'salient referents') would approve or disapprove of their speeding (i.e. driving at 40mph) in the scenario described. Again, these salient referents were identified from the piloting exercise and from previous studies (e.g. Parker et al 1992). Motivation to comply was assessed by asking respondents to indicate the extent to which they generally liked to drive in a manner of which each referent would approve. These two sets of items were summed using the TPB formula to produce a composite subjective norm score. Reliabilities for the two sets of items were high (normative beliefs: alpha=0.70, motivation to comply: alpha=0.85, when missing values were recoded to the mid-point of the scale. Missing values occurred when respondents rated a particular referent 'not relevant to me').

Perceived behavioural control was measured by asking respondents whether, in eight different circumstances, they believed they would be more or less likely to speed (for example, 'Would you be more or less likely to drive down this road at 40mph if you were running late for an appointment?'). A corresponding eight items asked respondents how often they found themselves in such circumstances when driving. These two sets of items, labelled control beliefs and control frequency, were summed to produce a composite perceived behavioural control. An alternative simpler measure of perceived behavioural control was also obtained, using two items, 'I would find it frustrating to stick to 30mph in this situation' and 'I would find it easy to stick to 30mph in this situation'. These two measures of control were used because there appears, as yet, only limited consensus on how control should be operationalised within the TPB. Reliability was high for the control beliefs (alpha=0.88) but low for the control frequencies (alpha=0.31). Reliability for the alternative measure of perceived behavioural control was high (alpha=0.80).

Some additional components, found by other researchers to have some potential to improve the predictive ability of the model (see Section 1.3), were also measured. Positive affective beliefs and negative affective beliefs are beliefs concerning the feelings which might be associated with performing the behaviour (Lawton et al 1997). Three items for each were included in the survey (e.g. positive affective belief: 'Driving down this road at 40mph would give me a feeling of pleasure'; negative affective belief: 'If I drove down this road at 40mph I would feel that I was driving foolishly'). A corresponding outcome evaluation statement was added for each affective belief statement. Reliability for the positive affective beliefs was 0.59, and 0.70 for the corresponding outcome evaluations. Reliability for the negative affective beliefs was 0.70, and 0.63 for the corresponding outcome evaluations.

Other items measured included personal norm ('It would be wrong to drive faster than 30mph in this situation' and 'I would feel bad about driving faster than 30mph in this situation'), habitual behaviour ('In this situation I would drive faster than 30mph out of habit'), and personal identity ('I like to think of myself as someone who would drive faster than 30mph in this situation'). These measures have been found in some circumstances to improve the predictive ability of the TPB (see Section 1.3).

Finally, reported behaviour was measured by asking respondents how often, in the past 12 months, they had driven over the speed limit in three different circumstances: on a 30mph road, late at night or early in the morning, and on a motorway. Questions were also asked about frequency of being flashed at by a speed camera, being stopped by the police for speeding, and receiving penalty points. The proportion of respondents reporting any frequency of these occurrences was low, and therefore the first set of questions was used as the main measure of reported behaviour throughout the survey.

2.2 SAMPLE AND RESPONSE RATE

Evaluations of mass media advertising are more rigorous if conducted with a random sample of the target population. However, because random sampling is more complex and time-consuming than other sampling methods, it was not considered achievable within the budget available for this study. The sample was therefore selected using a quota methodology with interlocking quota controls on age and sex. Respondents were recruited from Renfrew in the west of Scotland, which was selected as the study area for two reasons. Firstly, 1991 Census data show that the Renfrew population of 196,980 comprised households with similar levels of car ownership and owner-occupier status as found in the Scottish population. Further data on the affluence and deprivation of postcode sectors in Scotland (McLoone 1991) confirm that Renfrew comprises postcode sectors ranging from the most affluent to the most deprived. This sample area therefore ensured coverage of all social class groups. Secondly, in addition to its suitability in population terms, Renfrew contains a range of road types including motorway, minor country roads and roads in urban and residential areas. Thirty of Renfrew's 32 postcode sectors formed the study area; two sectors were excluded because they contained populations of less than 1,000.

The baseline sample (n = 550) was selected to be broadly representative of the 17-54 year old driving population, with interlocking quota controls on age and sex. The sample was skewed towards drivers aged 25-44 and male drivers, to reflect the core target groups of the Foolsspeed campaign. Respondents were recruited on the basis of meeting the socio-demographic requirements, being in current possession of a driving license and making at least one journey per week by car/van. Interviewers achieved the sample by recruiting door-to-door within designated streets in the 30 postcode sectors. The achieved baseline sample is illustrated overleaf (Table 2.1).

Table 2.1: Achieved sample - baseline survey

All baseline respondents were asked whether they would be willing to take part in a follow-up survey. Five hundred and twenty five respondents, 94%, indicated that they would be happy to do so, and were contacted in April-May 1999 by letter informing them that the follow-up survey would take place soon. Three hundred and eighty eight interviews were successfully achieved, equivalent to 74% of those baseline sample respondents who were willing to participate again. This was slightly lower than our anticipated response level of 82%. One of the main difficulties was in tracing and contacting the younger adults in the sample, many of whom are particularly mobile in the late teens and early twenties. In addition, 8% of the overall sample moved house between the baseline and 2 ^nd survey, and could not be traced.

Three hundred and eighty (98%) of the 388 respondents at the 2 ^nd survey indicated that they would be willing to continue to be part of the study. In addition, 84 respondents who had indicated at the baseline survey that they were willing to continue participating in the research but who had not been able to participate in the 2 ^nd survey, were included in the allocation for the 3 ^rd survey. This meant that the total allocated sample (i.e. the total number of potential respondents) for the 3 ^rd survey was 464. Three hundred and sixty-seven interviews were successfully achieved at the 3 ^rd survey, equivalent to 79% of the allocated sample of 464. This figure is slightly over the number of interviews which our original proposal indicated would be achieved by the 3 ^rd survey (360). The improved response rate in comparison to that achieved at the 2 ^nd survey may be partly attributable to this survey stage being incentivised, with each respondent who completed the interview being offered a 5 gift token. The 4 ^th survey achieved a response rate of 287, equivalent to 69% of the allocated sample of 416.

The table overleaf compares the size and gender and age distributions of the sample at each of the three stages. The 2 ^nd, 3 ^rd and 4 ^th survey columns indicate the distribution of respondents at each survey stage according to baseline gender and age categories (i.e. at the 2 ^nd survey, 19% of the sample had been in the 17-24 year old age group at the baseline, and at the 4 ^th survey 16% of the sample had been in the 17-24 year old age group at the baseline).

Table 2.2: Achieved sample at all four stages

The age distribution of the sample altered from the baseline, with the older groups being over-represented at all stages after the baseline. Younger respondents were more likely to be 'lost' to the research at each of the follow-up survey stages as a result of moving away to study or work.

2.3 SAMPLE CHARACTERISTICS

Respondents were categorised at baseline into three groups on the basis of self-reported frequency of speeding (Table 2.3). This speeding frequency score was generated from response to a question about frequency of speeding in three circumstances: late at night or early in the morning; on a motorway; and on a road with a 30mph limit. For each circumstance respondents indicated the frequency with which they carried out each speeding behaviour, in the last 12 months, on a scale of 1 to 7 (1 = Never, 7 = Almost all the time). 'Frequent speeders' refers to those who, in the last 12 months, almost always or very frequently drove above the speed limit (i.e. across the three circumstances had a mean score greater than 5.5). 'Infrequent speeders' refers to those who 'never or almost never' sped (i.e. mean score less than 2.5) and 'occasional speeders' refers to those whose mean score fell between 2.5 and 5.5 inclusive.

Table 2.3: Baseline characteristics of achieved sample at each stage: frequency of reported speeding, in 12 months prior to baseline

A fifth of the baseline sample were categorised as 'frequent speeders', while the majority (58%) were categorised as 'occasional speeders' and more than a fifth (22%) as 'infrequent speeders'. The 2 ^nd, 3 ^rd and 4 ^th survey columns indicate the distribution of respondents at each survey stage according to baseline speeding frequency (i.e. at the 2 ^nd survey, 17% of the sample had been in the frequent speeders category at the baseline, and at the 4 ^th survey 12% of the sample had been in the frequent speeders category at the baseline). The distribution by speeding frequency changed over the survey stages, with frequent speeders being under-represented in the sample at each follow-up stage and occasional and infrequent speeders being over-represented.

Frequent speeders were more likely to be male (89% of frequent speeders were male and 11% were female) and younger (34% of frequent speeders were aged 17-24, and 33% were aged 25-34).

The majority of respondents were in full-time employment (Table 2.4). Speeding frequency was associated with employment, with 81% of frequent speeders being in full-time employment compared to 63% of the total sample.

Table 2.4:Employment status

NB. Total more than 100% because respondents may fall into more than one category.

Social class is illustrated in Table 2.5. The social class profile of the baseline sample was not representative of the overall social class profile of Renfrew, which contains a greater proportion of people in social class C2DE ¹ The difference between the baseline social class profile and that of Renfrew as a whole may have been explained by our quota requirement that all respondents drive at least once a week, which is likely to have favoured people in social class ABC1, who have greater access to cars. The social class distribution of the sample did not change substantially over the follow-up surveys.

Table 2.5: Social class

Speeding frequency was associated with higher social class: at the baseline, 30% of frequent speeders were in social class AB, and 42% were in social class C1. In the total baseline sample, 25% were in social class AB, and 38% in social class C1.

Sixty percent of the baseline sample had children, and 69% were married or living with a partner. Frequent speeders were less likely to have children or be married, which may reflect their tendency to be younger.

Nearly two-fifths (39%) of the baseline sample had been qualified to drive for 9 years or less, while around a third (34%) had been qualified for 10-19 years, and just over a quarter (27%) for 20 years or more. Frequency of speeding was related to driving experience, with a tendency for speeding frequency to decrease with increased driving experience, and to increase with less driving experience. For example, half of frequent speeders (50%) had been qualified for 9 years or less, compared with just over a quarter of infrequent speeders; similarly, only 20% of frequent speeders had been qualified for 20 years or more, compared with 37% of infrequent speeders.

A quarter of the baseline sample (25%) said they were employed as a driver or did a lot of driving as part of their job (Table 2.6). Seventeen percent said they drove occasionally as part of their job, while over a third (35%) said they never did so, and around a fifth (19%) did not have a job. Males were more likely than females to be employed as drivers or drive a lot as part of their job (33% versus 9%). C2DE respondents were more likely than ABC1 respondents to be employed as drivers or do a lot of driving in relation to their employment, while ABC1 respondents were more likely to drive occasionally as part of their job. Frequent speeders were more likely than infrequent speeders to be employed as drivers, to do a lot of driving in their job, and to drive occasionally in their job.

Table 2.6: Amount of driving done as part of job, by gender, social class and frequency of reported speeding

At the 2 ^nd and 3 ^rd surveys, a repeat measure was taken of the amount of driving respondents did as part of their job, to assess whether there had been any substantial changes in respondents' circumstances. There were minor variations in the amount of job-related driving reported at both follow-up stages, the main one of which was a slight increase at the 3 ^rd survey in the proportion reporting driving "occasionally" as part of their job (23%).

The majority of the baseline sample (73%) drove only one car or van, while the remaining 27% drove more than one vehicle (Table 2.7). The majority (81%) drove their own car, while more than a tenth drove their spouse/partner's car (11%) or a car belonging to a friend or relative (13%). Just over a tenth drove a company car (12%) and a similar proportion drove a van (13%). Other vehicles such as buses, trucks and emergency vehicles were driven by 7%. Respondents who drove more than one vehicle tended to drive their own car most often (66%); more than one in ten drove a company car (11%) or van (13%) most often.

Table 2.7: Type of vehicle(s) driven at time of baseline survey. Selected responses

Table 2.8 illustrates the type of driving most frequently experienced by the sample at baseline. Male drivers and frequent speeders reported more frequency of driving 'to and from work', in rush hour, and on dual carriageways, motorways and country roads. Infrequent speeders and female drivers had higher frequency of driving 'for domestic reasons such as taking children to school or doing shopping'.

Table 2.8: Frequency of driving in different situations, by gender and frequency of reported speeding

Table 2.9 illustrates drivers' annual mileage, based on estimates calculated from two questions (see Appendix 1). Men, those who drove a lot as part of their job, and frequent speeders were more likely to report a higher mileage.

Table 2.9: Estimated annual mileage, by gender, amount of job-related driving, and frequency of reported speeding

The final table in this section reports on respondents' ratings of themselves as drivers. Each respondent was asked to indicate on a series of semantic scales a rating of him or herself as a driver in terms of a series of bipolar characteristics. Table 2.10 presents diagrammatically the mean scores given for each characteristic by the baseline sample as a whole, and by frequent and infrequent speeders.

Table 2.10: Type of driver respondents consider themselves to be, by frequency of reported speeding

On average, respondents had a positive and optimistic opinion of their driving, and were particularly likely to rate themselves as sensible, confident, safe, and considerate (mean scores 2.10, 2.16, 2.23 and 2.53 respectively, where low score indicates strong level of agreement with the characteristics). They also 'loved' driving (mean score 2.67) and considered themselves 'better than average' (2.67).

Frequent speeders, interestingly, had a different and what may be a more realistic rating of their driving. They were slightly less likely than drivers in general to rate themselves as sensible, safe and considerate (mean scores 2.56, 2.56 and 3.02 respectively), although more likely to describe themselves as confident (2.02 versus 2.16 for drivers in general). They are also less likely to describe themselves as 'never taking risks' (3.58 versus 2.80 for drivers in general) and, unsurprisingly, more likely to describe themselves as fast (3.07 versus 3.63 for drivers in general). Frequent speeders are more likely to 'love driving' but also to be 'stressed' when driving and to 'feel irritated by other drivers' (mean score 2.55, 3.10 and 4.53 versus 2.67, 2.85 and 3.98 respectively). On the 'sensible/foolish' scale, frequent speeders rated themselves as slightly less sensible and slightly more foolish (2.56) than drivers in general (2.10) and than infrequent speeders (1.64).

Overall, the demographic and driver characteristics of the baseline sample indicate that the sample was largely typical of the driving population in Scotland, although skewed deliberately towards males and 25-44 year olds to reflect the target of the Foolsspeed campaign. When demographic and driver characteristics were analysed by reported frequency of speeding, frequent speeders were found to be more likely to be male, younger and in social class ABC1; to drive for a living or drive as part of their job; to drive in rush hours and built-up areas; and to have a high annual mileage. They were also more likely to feel confident in their driving and to enjoy it, although also more likely to feel 'stressed' and irritated by other drivers. These findings were strongly consistent with other studies of the characteristics associated with speeding (see Section 1.2).

A more detailed profile of the baseline sample is presented in a CSM report (Stead et al 1999b).

« Previous | Contents | Next »