Influence of Smoking and Drinking on Higher Education in the United Kingdom - Coursework Example

Econometric Modelling Influence of Smoking and Drinking on Higher Education in the United Kingdom Declaration ‘I hereby certify that the project submitted is my own work and that I have acknowledged the use of the ideas or words of others.’ Abstract The subject under investigation is the effects of smoking and alcohol on teenage and young adult education rates. Moreover, the effect of education on smoking and alcohol usage will also be investigated. The project was tackled using information from official statistical sources to create regressive relationships in one and two variables. The findings indicate a strong relationship between smoking and enrolment rates as well as a strong relationship between smoking, alcohol use and enrolment rates. However, there was only a casual relationship between education and smoking or alcohol use. Acknowledgements It would not have been possible for any man to go forward without mentoring and so rests the case with me. I am very thankful to {name} for their guidance and constant support in carrying out this study. The mentoring efforts provided me with constant assurance that this study could be carried out even when I encountered unchartered territory. If such help and guidance were not available, it would not have been possible to achieve the undertaking presented below. Next I would like to thank my parents for always being there when I needed them the most. Last but not least I would like to thank God for providing me with continuous support and courage to take onto things even when I do not have the strength. Table of Contents List of Figures Figure 1 - Scatter plot for smoking prevalence against young entrants for full time degreee courses 10 Figure 2 - Residuals against fitted values 11 Figure 3 - Standardized residuals against theoretical quantiles 12 Figure 4 - Square root of standardized residuals against fitted values 12 Figure 5 - Standardized results against leverage 13 Figure 6 - Scatter plot for regular smokers against those who remember taking anti-smoking lessons 14 Figure 7 - Residuals against fitted values 15 Figure 8 - Standardized results against theoretical quartiles 15 Figure 9 - Square root of standardized results against fitted values 16 Figure 10 - Standardized results against leverage 17 Figure 11 - Scatter plot of alcohol users against those who remember taking alcohol lessons 19 Figure 12 - Residuals against fitted values 19 Figure 13 - Standardized results against theoretical quartiles 20 Figure 14 - Square root of standardized results against fitted values 21 Figure 15 - Standardized results against leverage 22 Figure 16 - Scatter plot of enrolment rates against alcohol usage 23 Figure 17 - Residuals against fitted values 24 Figure 18 - Standardized values against theoretical quartiles 24 Figure 19 - Square root of standardized values against fitted values 25 Figure 20 - Standardized results against leverage 25 Figure 21 - Scatter plot of enrolment rates against smoking and alcohol usage 27 Figure 22 - Residuals against fitted values 27 Figure 23 - Standardized results against theoretical quartiles 28 Figure 24 - Square root of standardized results against fitted values 29 Figure 25 - Standardized results against leverage 30 List of Tables Table 1 33 Table 2 34 Table 3 35 Table 4 37 1. Introduction This paper will portray an insight on how education affects teenage drinking and smoking. In order to do this, I will create an econometric model to investigate if teenage smoking and drinking is affected by schooling. I will find out whether there actually is a relationship between smoking and drinking and if only one affects the other or if it can be reversed. I will need to explore this theory within a specific age group and the age group chosen for this task is between 11and 19 years. This age bracket represents young adults who are more likely to be part of an educational system while being subjected to the effects of smoking and drinking. Since this transitional period in life is guided by a number of factors that the individual is not able to fully control so there is a large likelihood to break off with education in this age range due to the detrimental effects of smoking and drinking. In addition to examining the effects of education on teenage smoking and drinking, I will probe the possibility of causal being reversed. This would indicate that there could be a relationship between teenagers that frequently drink and smoke leaving education, perhaps due to their poor performance. I will assess if and how closely correlated the relationship between the amounts of teenagers that have not chosen to go into higher education i.e. college or university is with the quantity of smokers and drinkers. To break this relationship down even further, I can go into depth by cross-referencing this relationship against a time constraint, such as to distinguish the relationship within a specific year. If this relationship is closely correlated I can evaluate why this could be due to, such as social aspects of smoking and drinking, addiction or even a lack of awareness regarding health risks associated with smoking and drinking. There are a few variables, which will have to be considered such as teenagers or young adults that have entered higher education may still smoke and drink due to family members doing the same. This factor could be complemented by peer groups that are also smoking and drinking as well as disposable income that individuals who are smoking and drinking could be earning. Lastly, another variable that will need to be considered is the teenagers and the young adult’s awareness of health risks associated with smoking and drinking. If a clear strongly correlated relationship is present in my findings, in a specific time period, e.g. a year, this could be explained in two ways. First of all, for teenagers after the age of 16, higher education is not compulsory. Consequently many teenagers are attracted to the working life of earning income. As a result some teenagers and young adults may drop out of school, find a job and start socialising with older age groups, who possibly drink and smoke socially or habitually. This encourages the teenager or young adult to gradually start smoking and drinking himself. This social habit may increase with disposable income making it easier to acquire cigarettes and alcohol. However as previously mentioned the second reason is due to the causal being reversed. Teenagers and young adults, who may be frequently drinking and smoking, could end up dropping out or being expelled due to poor performance at school. In education, teenagers are surrounded by their own peer groups and also within the same age groups. There are health classes such as PSHE (Personal Social Health Education), which teach teenagers and young adults about the health risks of smoking and drinking. Lastly I will look at the effects of smoking and heavy drinking on grades and performance at school. Brain development is on-going process during a person’s teenage years, so if a teenager is a heavy drinker or smoker, this can have long term side effects for brain growth and hence intellect. It could possibly cause the teenagers intellectual capability to be reduced, thus leading to lower grades and poor performance at school leading to renouncing education. In order to carry out further research on this matter, a number of data sets were obtained from various sources regarding smoking and alcohol usage along with educational data. The scope of this data was limited to the United Kingdom alone. An additional limit placed on this data was the inclusion of a limited age range as documented before. The contention behind this is to study the influence of smoking and drinking on teenagers and adults only. The methodology of creating relationships between smoking and alcohol usage will bank on creating single variable relationships followed by creating multivariable relationships. This will be carried out using single variable and multivariable regression analysis as well as the use of time series. The analysis will be carried out using “R” which is open source software for statistical analysis. The exact methods used to create and test relationships between variables are described in detail below. 1.1. Linking Smoking to Educational Continuation Rates It can be theorised that smoking and educational enrolment rates at higher institutions of learning would be linked to each other. The hypothesis makes sense given that teenagers who start to smoke at an early age tend to be involved in deviant behaviour given that smoking at such age is deviant behaviour itself. The socialisation patterns indicate that teenagers who tend to socialise in such circles tend to drop out of educational circles early. Therefore, it could be said that if the smoking prevalence in teenagers tends to go up then there are chances that the educational enrolment at higher centres of learning should decrease and vice versa. In order to test this hypothesis data was collected from a number of different sources including the National Health Service (NHS) and the Higher Education Statistics Agency (HESA). A report by the NHS on smoking statistics on the United Kingdom was used in order to extract the percentage of teenagers who were involved in smoking. The report classifies age segments along with the prevalence of smoking in the respective age ranges (NHS, 2011). This study chose to use the smoking prevalence in the age range between 16 and 19 years of age. The legal age to smoke in the United Kingdom is 18 years so this segment would represent teenagers who are smoking before the stipulated age range. Moreover, this age range has a high relevance to higher education enrolment which is the other variable under consideration. The transition to college can be seen to occur around the age of 16 years. As hypothesised before, if teenagers begin to smoke early then it would be apparent that their enrolment at colleges would decrease. In order to gauge the enrolment levels at colleges and other such higher centres of learning the HESA website was utilised. The website provided data for the percentage of young entrants to full time degree courses through historical record keeping for the United Kingdom domain (HESA, 2012). Since there was no involvement of any age range so data from this table was used as is. In addition to these stipulations, the data chose from either source was extracted from the year 2000 onwards to the year 2009 in order to aid the creation of a time series as well as providing a healthy sample size for analysis. The smoking prevalence among teenagers between 16 and 19 years of age was chosen as the independent variable while the enrolment rates were chosen as the dependant variable for regressive purposes. The table of data has been presented in the appendix for reference as Table 1. A scatter plot of the variable in relation to each other is presented below in order to observe the consistency of data used for this purpose. Figure 1 - Scatter plot for smoking prevalence against young entrants for full time degreee courses The plot above clearly depicts that most data lies in close proximity to each other except for a major outlier at (27,88.5) for the year 2009. The coherence of the presented data along with limited outliers means that a linear model may be able to model the situation well. The first step was to carry out linear regression in order to create a regressive model of the form: This model was evaluated in “R” to create a regression model. The coding for the regression is shown in the appendix. The resulting plots of residuals and other related checks are shown below. Figure 2 - Residuals against fitted values Figure 3 - Standardized residuals against theoretical quantiles Figure 4 - Square root of standardized residuals against fitted values Figure 5 - Standardized results against leverage The results from the regression run based on a linear model indicate that the R-squared value is 0.3714 while the adjusted R-squared value is 0.2928. This means that the resulting regression linear equation is not able to account for variability strongly. The normal Q-Q approach shows a strong fit except for the singular outlier. This indicates that the linear model may be able to explain the relationship better than expected because the current regression has been affected in part by the outlier. The simplicity of the linear regression model means that it is not able to account for the scatter of data. In order to prove this point, the entire situation was remodelled for linear regression without the outlier value (27,88.5) for the year 2009. The resulting R-square value was 0.7416 indicating a high degree of fit with significant gains in accounting for variability. 1.2. Linking Smoking Prevalence to Tobacco Classes It was next assumed that receiving lessons as per the dangerous consequences of smoking may have a deterring effect on smoking in teenagers. This hypothesis was checked out using data from the NHS from a report relating the use of smoking and alcohol in young people in United Kingdom (NHS, 2009). This was correlated to the use of lessons against tobacco. The measure used was derived from a report by the NHS regarding young people who remembered receiving lessons regarding tobacco (NHS, 2008). The previously cited report from the NHS from 2009 was not used for this measure as it did not contain any such quantification but instead it contained lessons about drugs survey results. The extracted data has been shown in the appendix as Table 2. The relevant coding has been shown in Appendix B. The resulting plots are shown in the figures provided below. Figure 6 - Scatter plot for regular smokers against those who remember taking anti-smoking lessons Figure 7 - Residuals against fitted values Figure 8 - Standardized results against theoretical quartiles Figure 9 - Square root of standardized results against fitted values Figure 10 - Standardized results against leverage The results above clearly show that a very casual relationship exists between young adults in the smoking range with lessons delivered as to the negative consequences of tobacco. The R-squared value of the linear regression model is only 0.0746 indicating a high degree of variability. The first plot in the series (Figure 6) clearly shows that the scatter is very high given the consistent repetition of values in each data segment. Based on these results it can be clearly be seen that lessons against smoking tend to have little effect on the prevalence of smoking in teenagers. This may be seen as a consequence of the low amount of time spent in these lessons against the large amount of time spent socializing with pro-smoking influences in the social circle of teenagers. 1.3. Linking Alcohol Usage to Alcohol Classes This hypothesis states that alcohol usage in teenagers and young adults can be decreased by providing them the necessary awareness regarding the dangers of alcohol usage. Data was extracted for this section in exactly the same way as it was done for the previous section based on reports from NHS. The prevalence of alcohol usage to every day was taken as the independent variable while the percentage of teenagers and young adults who remembered taking classes on alcohol usage was taken as the dependent variable. The data table is shown in Appendix C as Table 3 and this is followed by relevant coding on this hypothesis. Again linear regression was utilized in order to gauge the relationship that exists between both variables. The resulting plots are shown in the figures provided below. Figure 11 - Scatter plot of alcohol users against those who remember taking alcohol lessons Figure 12 - Residuals against fitted values Figure 13 - Standardized results against theoretical quartiles Figure 14 - Square root of standardized results against fitted values Figure 15 - Standardized results against leverage The results above clearly indicate a strong relationship given that the variables being plotted have no deviance at all from each other. This indicates that results for any particular point would be more or less just the same. The R-squared value for this regression is 0.4679 indicating a strong relationship between both variables. In terms of the bigger picture, this indicates that alcohol classes tend not to make a difference to teenagers and young adults who are using alcohol on a regular basis. This would make a lot of sense given the fact that classes against alcohol usage are only limited in their time and scope while the socialization effort behind alcohol usage is pervasive. Put more simply, teenagers and young adults spend more time socializing in circles that advocate the use of alcohol than the time they spend in classes relating to alcohol use. 1.4. Linking Alcohol Usage to to Educational Continuation Rates The next hypotheses resembles the very first hypothesis in the respect that if teenagers and young adults who consume alcohol increase, then enrolment rates at centres of higher learning would decrease and vice versa. The data was extracted using a report by the NHS concerning the prevalence of alcohol usage across the United Kingdom (NHS, 2011) along with enrolment rates from the previous data compilation (HESA, 2012). The alcohol usage rate has been set at consuming more than 21 units of alcohol in a week. Data was not available for 2003-2004 as well as 2007 so the corresponding data was excluded as well from the data set. The data is presented in Appendix D as Table 4 while the relevant coding for linear regression is shown with it. The resulting plots are shown below. Figure 16 - Scatter plot of enrolment rates against alcohol usage Figure 17 - Residuals against fitted values Figure 18 - Standardized values against theoretical quartiles Figure 19 - Square root of standardized values against fitted values Figure 20 - Standardized results against leverage The results above clearly show that a weak relationship exists between entry in higher centres of learning and alcohol use. The large amount of outliers means that the relationship has a lot of variability. This is also indicated by the R-squared value from the linear regression which is 0.005338. Overall it could be said that there is little connection between alcohol use in teenagers and young adults and enrolment rates at centres of higher learning. This variation can be explained on two different accounts. The first problem is the data itself that has values missing for 3 years leading to gaps in data. The second problem is that the data was readjusted during 2008 that tends to perform as a maxima outlier. On the more social scale, this phenomenon can be explained by the fact that regular alcohol use is only carried out by a handful of teenagers and young adults (1% from the previous section) so this small proportion has little to do with the overall enrolment rate. 1.5. Linking Smoking and Alcohol Use to Enrolment Rates Finally a multivariate hypothesis was developed based on the idea that smoking and alcohol usage would lower enrolment rates at higher centres of learning. Data extracted from previous sections was used to link up smoking prevalence and alcohol prevalence to enrolment rates using multivariate regression. The data used is shown in Appendix E as Table 5 with the relevant coding shown below it. Figure 21 - Scatter plot of enrolment rates against smoking and alcohol usage Figure 22 - Residuals against fitted values Figure 23 - Standardized results against theoretical quartiles Figure 24 - Square root of standardized results against fitted values Figure 25 - Standardized results against leverage The result of this linear regression in two variables clearly shows that a strong relationship exists between alcohol use and smoking with the enrolment rates at centres of higher learning. The R-squared value for this multivariate regression is 0.473 indicating a high confidence in the resulting relationship. This can be attributed to a small spread as shown in the first plot above as well as the strong relationship between smoking prevalence and enrolment rates. 2. Conclusions The investigation carried out above clearly demonstrates that there is a strong relationship between smoking prevalence and enrolment rates at centres of higher learning. In contrast, the relationship between alcohol usage and enrolment rates is not as strong. Also, the effect of education on smoking and alcohol use is minimal largely because of the socialisation patterns that coerce teenagers and young adults to smoke and use alcohol. The multivariate relationship between smoking, alcohol use and enrolment rates is relatively strong indicating a stronger relationship. 3. References HESA, 2012. Summary of performance indicators. [Online] Available at: HYPERLINK "http://www.hesa.ac.uk/index.php?option=com_content&task=view&id=2075&Itemid=141" http://www.hesa.ac.uk/index.php?option=com_content&task=view&id=2075&Itemid=141 [Accessed 17 April 2012]. NHS, 2008. Smoking, drinking and drug use among young people in England in 2008. [Online] NHS Available at: HYPERLINK "http://www.ic.nhs.uk/webfiles/publications/sdd08fullreport/SDD_08_%2809%29_%28Revised_Oct_09%29.pdf " http://www.ic.nhs.uk/webfiles/publications/sdd08fullreport/SDD_08_%2809%29_%28Revised_Oct_09%29.pdf [Accessed 17 April 2012]. NHS, 2009. Smoking, drinking and drug use among young people in England in 2009. [Online] NHS Available at: HYPERLINK "http://www.ic.nhs.uk/webfiles/publications/Health%20and%20Lifestyles/sdd2009/SDD_2009_Report.pdf" http://www.ic.nhs.uk/webfiles/publications/Health%20and%20Lifestyles/sdd2009/SDD_2009_Report.pdf [Accessed 17 April 2012]. NHS, 2011. Statistics on Alcohol: England, 2011. [Online] NHS Available at: HYPERLINK "http://www.ic.nhs.uk/webfiles/publications/003_Health_Lifestyles/Alcohol_2011/NHSIC_Statistics_on_Alcohol_England_2011.pdf" http://www.ic.nhs.uk/webfiles/publications/003_Health_Lifestyles/Alcohol_2011/NHSIC_Statistics_on_Alcohol_England_2011.pdf [Accessed 17 April 2012]. NHS, 2011. Statistics on Smoking: England 2011. [Online] The Health and Social Care Information Centre Available at: HYPERLINK "http://www.ic.nhs.uk/webfiles/publications/003_Health_Lifestyles/Statistics%20on%20Smoking%202011/Statistics_on_Smoking_2011.pdf" http://www.ic.nhs.uk/webfiles/publications/003_Health_Lifestyles/Statistics%20on%20Smoking%202011/Statistics_on_Smoking_2011.pdf [Accessed 17 April 2012]. 4. Appendices 4.1. Appendix A Table 1 Year Smoking Prevalence (16-19 years) Young entrants to full-time first degree courses (%age) 2000 30 84.9 2001 28 85.7 2002 25 86 2003 25 87.2 2004 26 86.8 2005 25 86.7 2006 20 87.4 2007 22 87.8 2008 23 88 2009 27 88.5 > pervalence = c(30,28,25,25,26,25,20,22,23,27) > enrollment = c(84.9,85.7,86,87.2,86.8,86.7,87.4,87.8,88,88.5) > reg.model summary(reg.model) Call: lm(formula = enrollment ~ pervalence) Residuals: Min 1Q Median 3Q Max -0.92328 -0.64659 -0.05689 0.25215 2.04226 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 92.7425 2.7040 34.299 5.71e-10 *** pervalence -0.2328 0.1071 -2.174 0.0614 . --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 0.939 on 8 degrees of freedom Multiple R-squared: 0.3714, Adjusted R-squared: 0.2928 F-statistic: 4.726 on 1 and 8 DF, p-value: 0.06145 4.2. Appendix B Table 2 Year Regular Smokers (11 - 15) Remember Smoking Lessons 2001 10 63 2002 10 65 2003 9 61 2004 9 58 2005 9 59 2006 9 58 2007 6 60 2008 6 61 > smoker = c(10,10,9,9,9,9,6,6) > remember = c(63,65,61,58,59,58,60,61) > reg.model summary(reg.model) Call: lm(formula = smoker ~ remember) Residuals: Min 1Q Median 3Q Max -2.5672 -0.2724 0.7537 0.9701 1.0746 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) -2.3582 15.6208 -0.151 0.885 remember 0.1791 0.2575 0.696 0.513 Residual standard error: 1.666 on 6 degrees of freedom Multiple R-squared: 0.07463, Adjusted R-squared: -0.0796 F-statistic: 0.4839 on 1 and 6 DF, p-value: 0.5127 4.3. Appendix C Table 3 Year Alcohol Usage Almost Every Day Percentage who remember taking lessons 2001 1 55 2002 1 58 2003 1 56 2004 1 53 2005 1 55 2006 1 54 2007 1 58 2008 1 59 > alcohol = c(1,1,1,1,1,1,1,1) > rememberalcoholclass = c(55,58,56,53,55,54,58,59) > reg.model summary(reg.model) Call: lm(formula = alcohol ~ rememberalcoholclass) Residuals: Min 1Q Median 3Q Max -8.459e-17 -6.344e-17 -3.625e-17 -2.115e-17 3.263e-16 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 1.000e+00 1.436e-15 6.962e+14 alcoholperverlance = c(28,27,27,24,31,28,26) > alcoholenrolment = c(84.9,85.7,86,87.2,87.4,88,88.5) > reg.model summary(reg.model) Call: lm(formula = alcoholperverlance ~ alcoholenrolment) Residuals: 1 2 3 4 5 6 7 0.4861 -0.4185 -0.3828 -3.2397 3.7841 0.8556 -1.0848 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 37.6322 63.1668 0.596 0.577 alcoholenrolment -0.1192 0.7275 -0.164 0.876 Residual standard error: 2.336 on 5 degrees of freedom Multiple R-squared: 0.005338, Adjusted R-squared: -0.1936 F-statistic: 0.02683 on 1 and 5 DF, p-value: 0.8763 4.5. Appendix E Table SEQ Table \* ARABIC 5 Year Alcohol Usage Enrolment Smoking Prevalence 2000 28 84.9 30 2001 27 85.7 28 2002 27 86 25 2003 24 87.2 25 2006 31 87.4 20 2008 28 88 23 2009 26 88.5 27 > smok = c(30,28,25,25,20,23,27) > alc = c(28,27,27,24,31,28,26) > reg.model summary(reg.model) Call: lm(formula = alcoholenrolment ~ smok + alc) Residuals: 1 2 3 4 5 6 7 -0.32295 -0.40248 -1.02180 -0.62173 -0.08744 0.63196 1.82444 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 101.8822 9.5427 10.676 0.000436 *** smok -0.3064 0.1626 -1.884 0.132654 alc -0.2666 0.2518 -1.059 0.349255 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 1.165 on 4 degrees of freedom Multiple R-squared: 0.473, Adjusted R-squared: 0.2095 F-statistic: 1.795 on 2 and 4 DF, p-value: 0.2777 Read More