The Scottish Government

« Previous | Contents | Next »

Listen

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

Comparison by awareness of the Mirror ad

None of the three Foolsspeed ads were designed specifically to address affective beliefs. However, affective belief scores were analysed to assess whether any of the changes observed over the survey could be associated with exposure to Foolsspeed advertising. This analysis was conducted for awareness of the Mirror ad, as this ad was judged to be the most likely of the three to have an impact on affective beliefs.

The analysis revealed strong evidence that the changes in affective beliefs were associated with awareness of the Mirror ad. All of the significant changes, apart from the increase in negative affective beliefs between the baseline and 2 ^nd survey, were associated with awareness of the Mirror ad at each survey stage (Table 4.22). This suggests that, although the Mirror ad was not specifically designed to influence affective beliefs, it did appear to have an impact on them.

Table 4.22: Summary of significant changes in NAB and PAB over all surveys, by awareness of Mirror ad

When PAB and NAB scores were analysed by baseline frequency of speeding, frequent speeders had the weakest negative affective beliefs and the strongest positive affective beliefs (i.e. most pro-speeding), while for infrequent speeders the situation was reversed. None of the three speeding frequency sub-groups displayed any change in PAB or NAB scores between baseline and 2 ^nd or between baseline and 3 ^rd surveys. Between baseline and 4 ^th survey, occasional speeders displayed a significant increase in negative affective beliefs.

4.5 BEHAVIOURAL INTENTIONS

Behavioural intentions were assessed using three items: '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 drive faster than 30mph'. Responses were scored separately, and an overall measure of intentions computed by taking an average of the three items (the scale for the middle item is reversed for the overall measure). Table 4.23 displays the overall measure, comparing the baseline and all three subsequent surveys.

Table 4.23: Behavioural intentions: comparison of baseline and subsequent surveys, by baseline frequency of reported speeding

SD=Standard Deviations

Intentions were unsurprisingly strongest for baseline frequent speeders at all survey stages, and weakest for infrequent speeders. There was no significant change between baseline and each subsequent survey in behavioural intentions for the sample as a whole, or for the different categories of speeder. Nor was awareness of any of the adverts at any survey stage significantly associated with any changes in behavioural intentions.

4.6 REPORTED BEHAVIOUR

This section examines reported speeding behaviour over the course of the survey. Reported speeding behaviour was measured by asking respondents how often in the last 12 months they had driven over the speed limit in three different circumstances: 'late at night or early in the morning', 'on a motorway', and 'on a road with a 30mph limit'. Responses were scored separately, and also computed to produce an overall measure of reported behaviour. Baseline and follow-up scores were analysed at each survey stage to identify any changes in the sample's reported speeding behaviour (Tables 4.24 to 4.26).

Table 4.24: Reported speeding behaviour: comparison of baseline and 2 ^nd survey

Table 4.25: Reported speeding behaviour: comparison of baseline and 3 ^rd survey

Table 4.26: Reported speeding behaviour: comparison of baseline and 4 ^th survey

There was no change in overall reported speeding behaviour or in speeding in any of the specific circumstances, between the baseline and the 2 ^nd survey. However, the 3 ^rd survey saw a significant decrease in reported frequency of speeding in the environment targeted by the Foolsspeed campaign, a 30mph road (baseline 3.82, 3 ^rd survey 3.60, p<0.05). This same decrease was also found at the 4 ^th survey (baseline 3.61, 4 ^th survey 3.35, p<0.05).

Comparison by awareness of Foolsspeed advertising

In order to assess whether there were any changes in reported behaviour which could be associated with exposure to the Foolsspeed advertising, reported behaviour between the baseline and each subsequent survey stage in turn was compared, for those who had seen the advertising and for those who had not.

Between the baseline and 2 ^nd survey, there were no changes in reported speeding behaviour by awareness of the Mirror ad. Between the baseline and 3 ^rd survey, there was an apparently significant difference between those who had seen Friends and Family and those who had not, although this was in the 'wrong' direction: the reduction in speeding on a 30mph road was found only in those who reported that they had not seen the ad (baseline 3.92, 3 ^rd survey 3.64, p<0.05). Awareness of Simon Says at the 4 ^th survey was also associated with an apparent change in reported speeding behaviour, although again this was in the 'wrong' direction. That is, those who had not seen the ad reported an apparently significant reduction in speeding on a 30mph road between the baseline and the 4 ^th survey (baseline 3.66, 4 ^th survey 3.28, p<0.05), while those who had seen the ad did not display a significant reduction (although they reported less speeding in this circumstance).

However, more sophisticated analysis (repeated measures ANOVA) revealed that there was no significant interaction between stage and awareness of the campaign. No significant interaction effect was found between awareness of the Mirror ad and reported speeding behaviour on a 30mph road (F _1,385=0.006, ns); between awareness of the Friends and Family ad and reported speeding behaviour on a 30mph road (F _1,364=0.383, ns); or between awareness of the Simon Says ad and reported speeding behaviour on a 30mph road (F _1,285=1.017, ns). This analysis suggests that the above findings, which imply an inverse relationship between awareness and reported speeding behaviour, are spurious. In other words, seeing the Foolsspeed adverts had little impact, either negative or positive, on reported speeding behaviour.

Comparison by speeding behaviour

Results were also analysed by frequency of speeding behaviour to assess whether those categorised as frequent speeders at the baseline displayed similar or different trends in reported behaviour over the survey stages, compared with occasional and infrequent speeders.

Between the baseline and the 2 ^nd survey and between the baseline and the 4 ^th survey, two significant changes occurred in reported speeding behaviour. Firstly, those categorised as infrequent speeders at the baseline displayed an increased reported frequency of speeding, in all three circumstances, at both follow-up surveys. Secondly, those categorised as frequent speeders at the baseline displayed a decreased reported frequency, in all three circumstances, at both follow-up surveys (results for the baseline and 4 ^th survey comparison are displayed in Table 4.27). This is possibly a regression effect (Markus 1979), whereby groups which have more extreme scores on a specific variable at the first stage of a survey tend towards the norm when re-measured on the same variable at a later stage.

Table 4.27: Reported speeding behaviour: comparison of baseline and 4 ^th survey, by baseline frequency of reported speeding

4.7 PREDICTIVE STRENGTH OF THE TPB MODEL

The Theory of Planned Behaviour posits that attitude and subjective norms influence behaviour through behavioural intentions. Multiple regression analyses were conducted to examine the association between behavioural intentions (dependent variable) and attitude, subjective norm and perceived behavioural control (three independent variables). The TPB also posits that control can influence behaviour both through behavioural intentions and directly. Therefore, multiple regression analyses were conducted to examine the association between reported behaviour (dependent variable) and behavioural intentions and perceived behavioural control (two independent variables). These analyses were conducted for each individual survey to assess the consistency of the model at each survey stage. Results are displayed in Tables 4.28 and 4.29, with Table 4.29 presenting the analysis using the alternative measure of perceived behavioural control (see Table 4.16). Table 4.30 examines whether inclusion of the positive and negative affective beliefs improved the predictive ability of the model.

Multiple regression analyses were also conducted to examine the TPB model's ability to predict behavioural intentions and reported behaviour over time. Table 4.31 shows the association between behavioural intentions at the 4 ^th survey (dependent variables) and attitude, subjective norms and perceived behavioural control, as measured at baseline, plus gender, social class, age and campaign awareness (12 independent variables). Table 4.31 also shows the association between reported behaviour at the 4 ^th survey (dependent variable) and baseline measures of intentions, perceived behavioural control, demographics and campaign awareness (11 independent variables).

Table 4.28: Multiple regressions examining association between:
(1) intentions and attitude, subjective norms, perceived behavioural control
(2) behaviour and intentions, perceived behavioural control

Adjusted R ² (percentage of variance explained) is shown in the top row. Betas are shown for Intentions, Attitude, Subjective Norms and Perceived Behavioural Control
* p<0.05 ** p<0.01 ***p<0.001
Where the dependent variable was intentions, between 47% and 53% of the variance was explained by attitude, subjective norms and perceived behavioural control. The amount of variance explained was highest at the second survey (53%), although subjective norms did not significantly contribute to the model at this stage. In each case, perceived behavioural control was the most powerful independent variable associated with intentions, with a higher perceived behavioural control score (equivalent to a greater lack of control) being associated with a higher intention to speed. A more pro-speed attitude was also consistently associated with increased intention to speed. At all but the 2 ^nd survey, subjective norms were also significantly associated, with stronger scores (more encouraging of speeding) being associated with higher intentions to speed.

Where the dependent variable was reported behaviour, the amount of variance explained ranged from 33% to 40%. The amount of variance explained was highest at the 3 ^rd survey (40%). At each stage, intentions and perceived behaviour control were associated with reported behaviour, with intentions being the more strongly associated. In each case higher intentions to speed and a higher perceived behavioural control score (equivalent to a greater lack of control) were associated with increased speeding behaviour.

Table 4.29 presents the same multiple regression analysis as Table 4.28 but replaces the full perceived behavioural control variable with the simpler alternative version.

Table 4.29: Multiple regressions examining association between:
(1) intentions and attitude, subjective norms, alternative perceived behavioural control
(2) behaviour and intentions, alternative perceived behavioural control

Adjusted R ² (percentage of variance explained) is shown in the top row. Betas are shown for Intentions, Attitude, Subjective Norms and Alternative Perceived Behavioural Control
* p<0.05 ** p<0.01 ***p<0.001

Where the dependent was intentions, the amount of variance explained by attitude, subjective norms and the alternative measure of perceived behavioural control improved considerably, ranging from 64% to 70%. At each survey the alternative perceived behavioural control was the most powerful independent variable associated with intentions, with a higher score (equivalent to a greater lack of control) being associated with a higher intention to speed. A more pro-speed attitude was also consistently associated with higher intentions to speed, as were stronger subjective norms (i.e. ones more encouraging of speeding).

Where the dependent variable was reported behaviour, the amount of variance explained using the alternative measure of perceived behavioural control was lower, ranging from 29% to 37%. At each survey, higher intentions to speed were again associated with increased speeding behaviour. The alternative perceived behavioural control measure was only significantly associated with behaviour at the 2 ^nd survey, where a higher score (equivalent to less control over speeding) was associated with increased speeding behaviour. The poorer performance of this model is likely to be due to the strong association between the 'independent' variables - alternative perceived behavioural control and intentions. As these two 'independents' are so strongly linked it is not ideal for them both to be included in the model. The full measure of perceived behavioural control appears to be better associated with behaviour (see Table 4.28) and is therefore used in subsequent analyses of behaviour.

Table 4.30 examines whether inclusion of the positive and negative affective beliefs (PABs and NABs) improved the model's predictive ability at each survey stage. Because PABs and NABs are closely related to attitudes, they might be assumed to influence behaviour through intentions. Intentions was therefore the dependent variable examined in these analyses. The alternative measure of perceived behavioural control was used as it displayed a strong association with intentions.

Table 4.30: Multiple regression examining association between:

intentions and attitude, subjective norms, alternative perceived behavioural control, positive affective beliefs and negative affective beliefs.

Adjusted R ² (percentage of variance explained) is shown in the top row. Betas are shown for Attitude, Subjective Norms, Alternative Perceived Behavioural Control, Positive Affective Beliefs and Negative Affective Beliefs

* p<0.05 ** p<0.01 ***p<0.001

The addition of PABs and NABs to the model made little difference to the amount of variance explained, increasing it very slightly to a range of between 64% and 71%, similar to the level without their inclusion (see table 4.29). Negative affective beliefs made a significant contribution to the model at all but the 2 ^nd survey. A higher score (more negative affective beliefs) was linked to a lower intention to speed. Positive affective beliefs contributed significantly to the model only at the baseline with a higher score (more positive affective beliefs) linked to a greater intention to speed. The inclusion of PABs and NABs resulted in the attitude measure no longer making a significant contribution at the 3 ^rd and 4 ^th surveys. This is likely to be caused by an association between attitude, positive affective beliefs and negative affective beliefs and suggests that the model would be better with one or more of these removed.

Table 4.31 overleaf presents the multiple regression analyses examining the ability of baseline measures and campaign awareness to predict speeding intentions and behaviour at the 4 ^th survey.

Table 4.31: Multiple regressions examining association between:
(1) behavioural intentions (at 4 ^th survey) and attitude, subjective norms, alternative perceived behavioural control (at baseline), campaign awareness and demographics
(2) reported behaviour (at 4 ^th survey) and intentions, perceived behavioural control (at baseline), campaign awareness and demographics

Adjusted R ² (percentage of variance explained) is shown in the top row. Betas are shown for the TPB components, awareness of Foolsspeed advertising measures, and demographic characteristics
* p<0.05 ** p<0.01 ***p<0.001

Where the dependent variable was intentions, the amount of variance explained was low, at 27%. Significant predictors of intentions to speed were attitude, perceived behavioural control, subjective norms and age. Higher intentions to speed (at 4 ^th survey) were predicted by a higher baseline measure of attitude (equivalent to more pro speeding attitudes), perceived behavioural control (equivalent to a greater lack of control) and subjective norms (i.e. ones more encouraging of speeding), and by being younger. Awareness of the campaign elements, gender and social class were not significant predictors of intentions to speed.

Where the dependent variable was reported behaviour, the amount of variance explained was slightly lower at 22%. A tendency to report speeding behaviour at the 4 ^th survey was predicted by a higher baseline measure of perceived behavioural control (equivalent to a greater lack of control), higher baseline intentions to speed, and being younger. Awareness of the campaign elements, gender and social class were not significant predictors of speeding behaviour.

Table 4.32 repeats the above multiple regression analysis for behavioural intentions, this time with positive and negative affective beliefs built into the model.

Table 4.32: Multiple regressions examining association between:

behavioural intentions (at 4 ^th stage) and attitude, positive affective beliefs, negative affective beliefs, subjective norms, alternative perceived behavioural control (at baseline), campaign awareness and demographics

Adjusted R ² (percentage of variance explained) is shown in the top row. Betas are shown for the TPB components, awareness of Foolsspeed advertising measures, and demographic characteristics
* p<0.05 ** p<0.01 ***p<0.001

The inclusion of PABs and NABs in the model increased the explained variance to 27% but neither measure made a significant contribution. Baseline measures of subjective norms, perceived behavioural control and age continued to make a significant contribution to the prediction of speeding intentions at the 4 ^th survey.

A number of conclusions can be drawn from the regression analyses. Firstly, the analyses confirm that the basic TPB is a useful model for predicting speeding intentions and behaviour, explaining between 47% and 53% of the variance in intentions and between 33% and 40% of the variance in behaviour, when analysed for each survey stage separately. These proportions are comparable to those found in other applications of the basic TPB to speeding (e.g. Parker et al 1992, Stradling & Parker 1996). Attitudes and perceived behavioural control made significant contributions to the prediction of intentions at all four survey stages, and subjective norms made a significant contribution at three of the four survey stages. Intentions and perceived behavioural control made significant contributions to prediction of behaviour at all four stages. Replacing the full perceived behavioural control variable with the simpler alternative measure appeared to improve prediction of intentions: between 64% and 70% of variance in intentions to speed was explained using this variable. However, it did not appear to improve prediction of behaviour, explaining between 29% and 37% of the variance in behaviour, compared to between 33% and 40% for the more complex PBC measure.

Secondly, the analyses suggest that the predictive strength of the TPB remains consistent when measures are taken at four separate survey stages. The ability of baseline characteristics to predict 27% of variance in intentions and 22% of variance in behaviour three years later suggests that the model's predictive strength remains reasonably robust.

Thirdly, the analyses show that affective beliefs, particularly negative affective beliefs, are associated with intentions and this lends further support to previous studies indicating that affective beliefs provide a useful extension to the model (e.g. Lawton et al 1997, Stradling & Parker 1996).

4.8 SUMMARY: CHANGES IN THEORY OF PLANNED BEHAVIOUR COMPONENTS OVER THE FOUR SURVEYS

Table 4.33 below summarises the changes which occurred in the main components of the TPB and the association, if any, with awareness of the Foolsspeed campaign. Where a change is indicated, this is in the desired (i.e. anti-speeding) direction.

Table 4.33: Summary of TPB changes from baseline to 4 ^th survey

There were significant changes in an anti-speeding direction over the campaign period in the three attitudinal components of the Theory of Planned Behaviour - composite Attitude towards the Behaviour, composite Positive Affective Beliefs and composite Negative Affective Beliefs. These changes were largely sustained between the 2 ^nd and 4 ^th surveys, and were nearly always significantly associated with awareness of the Mirror ad - i.e. they did not occur in those who did not see the Mirror ad - providing reasonable support for the conclusion that the Mirror ad had a favourable effect on Attitudes and Affective Beliefs about speeding.

There was no evidence that desired changes occurred in composite Subjective Norms or composite Perceived Behavioural Control over the campaign period. Minor changes did occur in the items which make up these two components, between the baseline and subsequent surveys, but these were either short-lived or not always in the desired anti-speeding direction. There appears, at best, only weak evidence that awareness of the Friends and Family ad had an effect on Subjective Norms, and no evidence that awareness of Simon Says had an effect on Perceived Behavioural Control.

There was no evidence of a change in Behavioural Intentions between the baseline and any subsequent survey. There was no change in the measure used for overall reported Behaviour, but there was evidence between the baseline and 3 ^rd survey, and between the baseline and 4 ^th survey, of a reduction in reported frequency of speeding in one of three different driving circumstances included in the questionnaire, 'on a road with a 30mph limit'. This change was not associated with awareness of the Foolsspeed advertising.

« Previous | Contents | Next »