Racial Inequality in Education - Assignment Example

QUANTITATIVE DATA ANALYSIS 1) The specification of a research topic or theme, related to academic literature, that you will explore using the BSAS data. Secondary data is data that is collected and generated by one researcher, which could also be used by other researchers for their own aims and purposes. According to Babbie (2001: 269), analysis of secondary data consists of “data collected and processed by one researcher are reanalysed – often for a different purpose – by another”. No doubt, when undertaking this important social research there are two major regularly used methods: quantitative plus qualitative; There is frequently a great divide among the users of qualitative plus quantitative methods of research. Put just, quantitative research uses arithmetical principals plus it is statistical, so it is frequently viewed as more dependable and valid. On the other hand there is qualitative research which is based on meeting and finding out the "why" of a exacting situation. Qualitative research is interpretative rather than evocative, is non-numerical and is recognized for it's trustworthiness and auditability. The General Household Survey and British Social Attitudes Survey are some examples of many secondary data available; these data have been generated on a large scale (known as Official Statistics) collected by government researchers and officials. For this assignment, I will be analysing data from the British Social Attitudes Survey 2003; data file bsarsls06.sav using software package SPSS 11. In the course of analysing of the survey data, I have come up with a theme related to academic literature; Effects of social class and ethnicity on the quality of life. Evidence from literature shows there are inequalities in educational outcome, with the ethnic minorities and working class being disadvantaged1. However, Modood et al. (1997) findings show “ethnic groups are more likely to remain in full time education after the age of 16 than their white counterparts” (in Mason, 2000: 62). Furthermore to the findings, “minority ethnic groups are over-represented in higher education relative to the presence in the population as a whole” (Jones, 1993: 32, Modood, 1997a; Modood and Ackland, 1998; Modood and Shiner, 1994 in Mason, 2000: 62-63), as a result they are perceived as overachievers in education in regards to their percentage in the population. Mason (2000: 62-63) claims there are many reasons for this phenomena; motivation, support, compensation for previous underachievement and means of avoiding unemployment, are driving forces for ethnic minorities’ achievement in higher education2. These conclusions are supported by the 1991 Census, which found (based on a sample of 10% of the population) “more than a quarter of adult Black African…and Chinese were qualified beyond A Levels, which was double that of whites”, furthermore “Indians and African Asians were also relatively more qualified than whites” (Modood in Modood et al., 1997: 64). All in all, ethnic minority groups were substantially more qualified academically, by a wide margin, compared to those of white origin. In addition to the findings, the 1991 Census showed that on average, males are overachieving in higher education from A Levels onwards, compared to females. Statistics showed 26% of males of ethnic minorities are achieving A Levels compared to 24% of females; furthermore, 15 % of males of ethnic minorities are obtaining degrees in comparison to 11 % of females. However, it is important to bear in mind that some minority groups are over achieving, such as Indians, African Asian and Chinese, in comparison to other ethnic groups, such as Pakistanis and Bangladeshis (Modood in Modood et al., 1997: 64-66). On to a different literature field, the core findings in the Black Report published in August 1980, showed “there are large differentials in mortality and morbidity that favoured the higher social classes and that were not being redressed by health and social services” (in Smith et al. (1990), Vol. 301, Issue No. 6748: 373), those of working class background were disadvantaged in health compared to the middle class, and furthermore they were failed by the health services. The Black Report, 10 years on, claims “social class differences in mortality are widening…social class differences exist for health during life as well as for length of life”, so not only does social class affect how long one lives but also the quality of their health (in Smith et al. (1990), Vol. 301, Issue No. 6748: 373). The Black Report is supported by Green’s claims that there is “inequality in the use of the health service” which inevitably shows the “failures of the NHS”, such findings have called for “demands for higher spending on the NHS…” (Green, (ed.), 1988: 3). Analysing the root causes of the crisis, Scambler claims inequality in health “was primarily related to inequalities of material sources” (Scambler, 2002: 91). Material factors such as owning a home and car, low social class and educational achievement had significant impact in health inequality. These factors are normally associated with the working class3, thus the findings of the Black Report showed there was a correlation between health and social class (in Smith et al. (1990), Vol. 301, Issue No. 6748)4. According to Pond, the “substantial inequalities in the health of different socio-economic groups remain”, “economic inequality” was one of the root causes of inequality in health, for it is associated with poverty (in Davey and Popay, (ed.), 1993: 185). There is a correlation between income and social class which affects health. Ponds findings support the Black Report, which gives evidence that the working class suffer from more health problems that the middle class. “Long standing Illnesses” such as coronary heart disease, angina, respiratory symptoms, systolic blood pressure etc. were higher amongst people from the lower social group (in Smith et al. (1990), Vol. 301, Issue No. 6748: 374). Other findings, such as Oppenheim’s study on Poverty; the Facts, Child Poverty Action Group, (1990), London: p30 shows that large numbers of pensioners are receiving low income, they therefore have insufficient funds to pursue material gains, which has been previously mentioned to being a factor affecting health. 25% of pensioners have income below 50% of average, exclusive of housing costs (in Davey and Popay (ed.), 1993: 188). A impartial appraisal of the two forms of research require a knowledge of the history at the back the two shape of research, an in deepness look at what they are and what they together entail as social research and to find out the dissimilar methods that can be used inside qualitative as well as quantitative research. We also require finding out the power and weaknesses of equally qualitative and quantitative data so fair judgement can be made to find out if there is a correct way to assume social research. Furthermore, research on how gender-related attitudes differs crossways tribal/ethnic groups has shaped opposing results, depending upon the kind of attitudes addressed. In this research, I reconsider the literature on racial and tribal difference in three broadly defined types of gender attitudes: attitudes toward gender roles; beliefs regarding the origins and degree of gender inequality; and favourite for social action to decrease gender inequalities. I address three racial/ethnic groups in the United States: African Americans, whites, and Hispanic Americans. As research on attitudes toward gender roles has give way mixed result, research addressing attitudes inside the other two domains obviously point to better censure of gender inequality between African Americans relative to whites; research on the a variety of groups frequently mutual under the label Hispanic is too incomplete to draw any clear conclusions. Along with speak to variations across these three types of gender-related attitudes; I also sum up some other patterns evident in the literature: meeting crossways groups over time; gender gaps in gender-related approach; and discrepancy predictors of gender attitudes crossways racial/ethnic groups. 2) A discussion of your choice of at least 5 relevant variables and at least 3 hypotheses linking your chosen variables. Using the literature review, I was able to develop hypotheses for the report, using the relevant variables from the data file. Modood et al. (1997) claimed that ethnic minority groups achieved higher in the academic field compared to the white racial group, they were overachieving in regards to their population. From this information, I drew two variables; respondent’s race and highest education qualification obtained. I hypothesised that Minority ethnic groups obtain higher education qualifications than their white counterparts Furthermore in the literature, findings in the 1991 Census (in Modood et al., 1997) showed males were over achieving from A Levels to degrees in contrast to their female peers. From this I was able to draw the third variable sex of respondent and further hypothesise Males within ethnic groups achieve higher education qualifications than their female counterparts The Black Report (in Smith et al. (1990), Vol. 301, Issue No. 6748), claimed social class differences account for the inequality in health and healthcare, due to deprivation of material resources. Using this literature, I was able to identify two variables from the data file; respondents social class and how satisfied with local doctor and local GP. Middle class are more likely to be satisfied with their local doctors and GPs because they can gain better access to healthcare compared to the working class In addition in the literature, age is a factor which affects income, which in effect affects the ability to gain better healthcare and material factors needed e.g. cars, home etc. from this, I was able to obtain my third variable age and further hypothesise Those of the top age band (45+) in the working class are more likely to be dissatisfied with their local doctor and GP, compared to those of the lower age band (18-44), due to lack of material resources and access to adequate healthcare (Refer to Appendix 2, Part 1) 3) The results of appropriate statistical analyses and conclusions that you draw from your analysis. Using the hypothesis (refer to Report Part 2, Appendix 2; 1), I analysed the cross-tabulations of respondent’s race and highest educational qualification obtained (refer to Appendix 2; 2 and 3); the value of Cramer’s V was .280 with Approx. Sig of .000, implying there is an association even if it is a weak one. The Chi-square significance.000 implies the result is significant at the 0.1% level, there is less than one in a thousand chance these results happened by chance, assuming that it was a random sample. The test of significance confirms the test hypothesis; there is an association between the two variables in the wider population. I recoded the two variables into 4 dichotomous variables (refer to Appendix 3; 1, 2 and 3). The analysis of the cross-tabulation shows there is a weak association, established by the epsilon figure .225 (refer to Appendix 3; 3 for epsilon calculations). Phi value .124 confirms the weak association between the two variables, the Chi-square value is .000, indicating these figures are significant at the 0.1% level; the test hypothesis is again true (refer to Appendix 3; 3 for in-depth analysis). I hypothesised further by introducing my third variable ‘sex of respondent’ (refer to Appendix 2; 1 for hypothesis) and recoded it (refer to Appendix 3; 4 analysis in 5). The epsilon for male was .336, implying there is a stronger association, than the previous 2x2 epsilon, even if it is still a weak one. The Phi value .188 confirms the association. The Chi-square value .000 indicates the test hypothesis is true, for the result is significant at the 0.1% level, it can therefore be generalised to the whole population. The epsilon for female was .138, which is a weaker association compared to the previous 2x2 table. The Phi value of .075 confirms the weak association; the Chi-square value .077 implies there is 7.7 % chance of the Null hypothesis being true; thus the result is significant at the ten percent level. The third variable has specified the initial relationship; the results vary for male and female. The Chi-square indicates the results for male and female are statistically significant, the male being the stronger association and the female being the weaker association compared to the initial relationship. I can confirm my hypothesis supported by the literature (refer to Part 1 of Report and Appendix 2; 1) and the results produced (Appendix 2 and 3); there is an association between the variables race, the highest educational qualification obtained for male and female for the population as a whole (refer to Appendix 3; 6 for full conclusion). Again using the hypothesis (refer to Report Part 2, Appendix 2; 1), the analysis of the cross-tabulations of respondent’s social class and how satisfied with local doctor and GP (refer to Appendix 2; 2) shows the percentage of ‘very satisfied’ varies amongst the social classes. The Cramer’s V value .138, with Approx. Sig .001 indicates there is a very weak association between the variables (refer to Appendix 2; 3). The Chi-square value .001 indicates the result is significant at the one percent level; there is less than one in a hundred chances this result happened by chance. The test hypothesis has been confirmed, there is an association between the two variables in regards to the whole population. The two variables were then recoded (refer to Appendix 3; 1 and 2). The analysis of the cross-tabulation shows there is a very weak association, confirmed by epsilon value .05 (refer to Appendix 3; 3). Phi value .068 confirms the weak association with Approx. Sig .086. The Chi-square value .086 implies there is 8.6% chance of the Null hypothesis being true, one in ten chances the result happened by chance; the result is significant at the ten percent level. I introduced the third variable ‘age of respondent’ based on academic literature (refer to Appendix 2; 1) after recoding it (refer to Appendix 3; 1). The epsilon for ages 18-44 was .084 and for ages 45-64+ was .037, the previous epsilon figure for the 2x2 was .05; 18-44 had a stronger association, even though it is still low and 45-64+ had a weaker association. The Phi for 18-44 confirms the association with the value .103, the Chi-square value .086 implies there is 8.6% chance of the Null hypothesis being true; the result is significant at the 10% level. The Phi for ages 45-64+ at .053, confirms the weak epsilon association. The Chi-square value .310 implies there is 31% chance of the Null hypothesis is true, it is beyond the ten percent threshold, therefore cannot be generalised to the whole population. The third variable has specified the initial relationship. The results vary for the age bands, the ages 18-44 have a stronger association than ages 44-64+, compared to the initial relationship, but the latter cannot be generalised to the population, for it is statistically insignificant due to Chi-square value being beyond the ten percent threshold; 31% chance of the Null hypothesis being true. I can confirm my hypothesis using the results produced and the literature (refer to Appendix 2; 1); there is an association between social class and how satisfied one is with their local doctor and GP. My third variable had a partial affect; ages 18-44 had an stronger association with the initial relationship, but ages 45-64+ were statistically insignificant; therefore I cannot be confident social class and how satisfied with local doctor and GP are associated with ages 45-64+ in the population as a whole (refer to Appendix 3; 3 for full conclusion). 4) A critical discussion of the use of official statistics in social research. The purpose of official statistics is to gain understanding of the research topic by analysing coded data (data coded from survey) in numeric form, to generate correlations and patterns in the findings (Barnett, Lecture 1: 23/02/2006)5. Data is analysed in numeric form through coding using computer software such as SPSS; variables can be compared and contrasted, correlations and patterns can be found, conclusions can be drawn using statistics, which is highly reliable6. The advantages are official statistics can be generated by organisations such as the government, commercial industries etc; data can be generated for every day use. Data can be stored as data archives to be reused by other researchers to meet the goals of their own research projects, this means the research process is faster and cheaper than conducting own research from the primary phase; “research can begin at the analysis phase” (Seale, 2004: 356), in addition, due to the increase in use and sophistication of technology, such data can be generated and sent and used all over the world, thus data is easily accessible. Vast numbers of official statistics data are available to researchers, thus easily accessible. Official statistics are continuously generated, it is therefore “possible to compare trends over time” (Seale, 2004: 366). Large samples can be drawn in the population from the survey in the survey process, it is more representative of the population therefore generalisations can be made, compared to other methods; saving research funds and researchers time, data can be further analysed or reanalysed for different research aims, furthermore researchers can select the amount of data they wish to use (Babbie, 2001:269). The disadvantages of official statistics are based on the issues of analysing data for topics or themes for which the data was not originally generated; researcher must understand technical terms of original study. The researcher “using officially produced data has little control over the variables measured” (Seale, 2004: 362), or how the variables have been classified, such as age ranges etc, there are also questions about the data’s quality of research; how well the research was carried out, the skill of the original researcher. Questionnaires have a low response rate; data derived from such source could be incomplete, there are also issues regarding the validity and reliability of original generated data. Researcher may “face the limits of the original data forms” when trying to create new variables from the old data (Seale, 2004: 364), thus limiting the new data which could be generated. Furthermore, data can be manipulated, e.g. for political reasons to show policies in a favourable light, tampering of data may not be known, for raw data is not published, the aggregated ones are (Barnett: 23/02/2006). Ethnography on the other hand can be used for qualitative data analysis. The strengths of ethnography is validity; researcher is conducting the research, so will know what is being measured or what they think they are measuring. The meanings of people (meaningfulness) can also be studied; data from this can be used to achieve real things in the real world. Researcher can gain rich data of groups and social interaction through methods used, compared to other research methods because they are experimental or unrealistic, thus researcher is aware of the complexity of society. Validity is an advantage because compared to surveys and experiments, it captures in-depth data; the true picture of meanings and actions of what is being studied. There are also many weaknesses in ethnography, firstly the researcher is central to the research; biases and subjectivity is an issue, also the quality of the research is dependent on how well the researcher carries it out. Manipulation of analysis and interpretations of actions and observations may occur, due to researchers own experiences and opinions, furthermore data may reflect upon researchers own role on research (reflexity). Researcher is in the heart of the research; may get too involved (reactivity) and have a lot of stresses upon him/her. In addition, it is hard to make generalisations for each research is unique and society is changing, therefore it lacks in reliability; data is not repeatable due to the researcher and the approach of the research, for each research is distinctive (Barnett, 13/10/05). Official statistics and ethnography need different skills and methods in their research approaches. Form this assignment I learnt how to analyse data using SPSS; cross-tabulate variables, calculate Cramer’s V, Phi, Chi-square and epsilon vales. Using the literature review and the analysis of the data, I was able to verify my hypotheses. One of my hypothesis was proved wrong because it was statistically insignificant, which goes to show that there are many factors to consider before concluding and generalising on results (refer to Appendix 3; 6). Appendix 1 Part 1) Select 3 variables from the pilot questionnaire you designed in assignment 2. If possible choose ones with different levels of measurement, ideally nominal, ordinal and interval. I have selected 3 variables from assignment 2, using the pilot questionnaire based on “attitudes towards the 24 hours drinking licensing laws” survey. GENDER: is a nominal variable which has a dichotomous level of measurement. AGE: is an interval variable/ level of measurement. SOCIAL CLASS: is an ordinal variable/ level of measurement. VARIABLES Gender Age Social Class LEVEL OF MEASUREMENT Nominal Interval Ordinal Part 2) Produce a coding frame for each variable. VARIABLE LEVEL OF MEASUREMENT CODE VALUE LABEL Gender Nominal 1 2 -9 Male Female Missing Value Age Interval 1 2 3 4 5 6 7 8 9 10 11 -9 Under 18 19-23 24-28 29-33 34-38 39-43 44-48 49-53 54-58 59-63 64+ Missing Value Social Class Ordinal 1 2 3 4 5 -9 Group A Professional workers (Doctors, Lawyers etc) Scientists, Managers of large organisations Group B Shopkeepers, Farmers, Teachers, White collar Workers Group C 1. Skilled manual (i.e. hand) workers High grade e.g. Master builders, Carpenters, Shop Assistants, Nurses. 2. Skilled Manual - Low grade e.g. electricians, plumbers Group D Semi-skilled manual E.g. bus drivers, lorry drivers, fitters. Group E Unskilled manual e.g. general labourers, barmen, Port men. Missing Value Part 3) Note the level of measurement of each variable and list appropriate descriptive statistics (mode, median, mean, minimum, maximum) for each. VARIABLES Gender Age Social Class LEVEL OF MEASUREMENT Nominal Interval Ordinal DESCRIPTIVE STATISTICS Mode Mode, Median, Mean, Minimum, and Maximum Mode, Median, Minimum and Maximum Look at attachments. Appendix 2. Part 1) Explain 2 hypotheses involving the relationships between the 2 different pair of variables from the British Social Attitudes Survey data file provided. I have produced two pairs of variables using the data file British Social Attitudes Survey 2006 (bsasrsls06.sav): Variables: Respondent’s race (self rated) and highest educational qualification obtained; Sex of respondent (third variable) Hypothesis: Minority ethnic groups obtain higher education qualifications than their white counterparts, furthermore males within ethnic groups achieve higher in education qualification than their female counterparts Analysing literature based on this field, I found that Modood et al. (1997) findings show “ethnic groups are more likely to remain in full time education after the age of 16 than their white counterparts” (in Mason, 2000: 62). Even more so, in the 1991 Census, (based on a sample of 10% of the population), the study found that “more than a quarter of adult Black African…and Chinese were qualified beyond A Levels, which was double that of whites”, in addition “Indians and African Asians were also relatively more qualified than whites” (Modood in Modood et al., 1997: 64). To add to the findings from the same source, males were achieving higher in A Levels to degrees when compared to their female counterparts. Thus I was able to conclude that educational achievement is dependant on the race of a person. The level of educational achievement varies amongst the races, in other words educational achievement is affected by the variable race. Based on the literature, racially diverse groups are more likely to achieve higher in education compared to their ‘white’ race counterparts, furthermore, males are likely to achieve higher than females in the academic arena. Variables: Respondent’s social class (best estimated) and how satisfied with local doctors and GPs; Age of respondent (third variable) Hypothesis: Middle class are more likely to be satisfied with their local doctors and GPs because they can gain better access to healthcare compared to the working class, furthermore those of the top age band (45+) in the working class are more likely to be dissatisfied with their local doctor and GP, compared to those of the lower age band (18-44), due to lack of material resources and access to adequate healthcare The Black Report, 10 years on from the 1980 study, claims “social class differences in mortality are widening…social class differences exist for health during life as well as for length of life”, so not only does social class affect how long one lives but also the quality of their health due to material factors such as housing, cars, education etc (in Smith et al. (1990), Vol. 301, Issue No. 6748: 373). According to Ponds, “economic inequality” was one of the root causes of inequality in health, for it is associated with poverty (in Davey and Popay, (ed.), 1993: 185). There is a correlation between income and social class which affects health. Furthermore, 25% of pensioners had income below 50% of average, exclusive of housing costs (in Davey and Popay (ed.), 1993: 188). Therefore, they were more likely to have inadequate funds to gain access to proper healthcare. Using the literature, I was able to conclude that the level of satisfaction of the local doctors and GPs is dependant on one’s social class. Those of the lower social status group are more likely to be unsatisfied with their local doctor and GP, as they are more likely to use the NHS organisation (free healthcare), compared to someone of a higher status group, who are more likely to use private healthcare systems (personally paid for), which provide better health services. Furthermore, those of the top age band 45-64+, are also more likely to be unsatisfied, for they have low incomes which acts as an obstacle to better healthcare. Part 2) Use SPSS to produce an appropriate cross-tabulation for each pair of variables. Include appropriate percentages, measures of correlation (Phi/ Cramer’s V) and test of significance (Chi-square). The following tables are cross-tabulations for each pair of variables; they include the appropriate percentages, measures of correlation (Cramer’s V) and tests of significance (Chi-square). Note that Phi will not be used as these are large tables, bigger than 2*2 tables, thus Phi is inappropriate to use. Chi-square significance threshold will be at significant at the ten percent level. Part 3) Note what you conclude from these tables. Cramer’s V measures the level of association. 1 is the strongest association with 0 being the weakest association. Chi Square is the statistics used to measure how statistically significant the data is of the population as a whole, if the data can be generalised to the population as a whole, using test hypothesis and null hypothesis; Test hypothesis: the correlation between the variables in the population as a whole is true. Null hypothesis: no correlation, cannot generalise to the population, is false for the Chi Square is low. Respondent’s race (self rated) and highest educational qualification obtained; Analysing the rows of the cross tabulation, I can see that 14.7% of the total sample have obtained degrees (top row, last column). If there was no association between respondent’s race and highest education qualification obtained, then we would assume that 14.7% of all racial groups have obtained degrees. However analysing the table I can see that the percentage of degrees obtained vary amongst the racial groups. There is a correlation between respondent’s race and the highest education qualification obtained; 11.1% BLACK of Caribbean origin have obtained degrees, almost double that figure are Black of African origin with 23.1% 55.6% ASIAN of Indian origin have obtained degrees, compared to ASIAN of Bangladeshi origin with .0% 6.7% of ASIAN of Pakistani origin have degrees compared to ASIAN of Chinese origin, who are achieving 100% higher with 66.7% obtaining degrees WHITE of European origin are underachieving with 13.7% obtaining degrees compared to WHITE of other origin, with 33.3% achieving degrees Cramer’s V confirms there is some association even though it is a weak one; with value of 0.280 with approximate significance of .000 (Symmetric measures, 2nd row) The significance value of .000 of the Chi Square gives evidence that this result is significant at the one percent level. The Chi Square significance value is at .000 (Asymp. Sig. (2 sided)) (Chi Square tests, 4th column); this implies that this result is significant at the one percent level; this is because the significant value is less than .01. In this case the result is significant to the 0.1% level for the value is below .001. There is less than one in a thousand chances these results happened by chance, (no chance in a 1000). The correlation between the variables respondent’s race and highest education qualification obtained is a weak association. The test significance .000 confirms there is a correlation between these variables in the population as a whole, thus the results can be generalised to the wider population. The test hypothesis is true, the Null hypothesis is low thus false. Respondent’s social class (best estimated) and how satisfied with local doctors and GPs; The results from rows of the cross-tabulation shows that 28.1% of the total sample (201 respondents out of 716) were ‘Very Satisfied’ with their local doctors and GPs. If there was no association with social class and satisfaction levels, then we would expect that 28.1% of all social classes were ‘Very Satisfied’ with their local doctors and GPs. However analyzing the data, we can see that satisfaction levels are dispersed unevenly amongst the social groups; Social Class: I (SC=1) ‘Very Satisfied’ 17.5% II (SC=2) ‘Very Satisfied’ 30.5% III (non-manual) (SC=3) ‘Very Satisfied’ 26.2% III (manual) (SC=4) ‘Very Satisfied’ 27.2% IV (SC=5) ‘Very Satisfied’ 25.4% V (SC=6) ‘Very Satisfied’ 41.7% Armed Forces ‘Very Satisfied’ .0% Cramer’s V confirms there is a very weak association, with the value of .138 with approximate significance of .001. The significance value of .001 of the Chi Square (Asymp. Sig. (2 sided)) (Chi Square tests, 4th column) gives evidence that this result is significant at the one percent level, for the significant level is below .01 The Chi Square significance value is at .001 (Asymp. Sig. (2 sided)); this implies this result is significant at the one percent level, for the significance value is less than .01. There is less than one in a hundred chances these results happened by chance, (no chance in a 100). Thus the results can be generalised to the wider population, because it is below the ten percent threshold. The correlation between the variables respondent’s social class and how satisfied with local doctors and GPs obtained is a weak association, with Cramer’s V value of .138. The test hypothesis is true, for the Chi Square significance is low, thus the Null hypothesis is weak. Appendix 3 Part 1) Recode, as appropriate, the variables from the two cross-tabulations used in Appendix 2 so that you have 4 dichotomous variables. Look at attachments Part 2) Use SPSS to produce a 2x2 cross-tabulation for each pair of variables, including appropriate percentages, measures of correlation (Phi / Cramer’s V) and tests of significance (Chi-square). Look at attachments (Part 3) Comment on any differences in correlation or significance between these and your earlier tables. I recoded the variables from the two cross-tabulations in Appendix 2, to produce two new 2x2 tables. New respondent’s race (self rated) and new highest educational qualification obtained; If there was no association between the two variables, I would expect 21.0% of both racial groups to obtained degrees as their highest educational qualification. However analysing the table, I can see this is not the case; 42.3% of race of ethnic origin gained degrees as their highest educational qualification obtained compared to 19.8% of race of white origin. Value of Phi is at .124 with Approx. Sig. at .000 Value of Chi-square is at .000 Race and degree Epsilon: 42.3% - 19.8% = 22.5 22.5/100 = 0.225 Race and higher education below degree: 80.2% - 57.7% = 22.5 22.5/100 = 0.225 The epsilon value of .225 shows there is a weak association between race and highest educational qualification obtained. This is confirmed by Phi value of .124 which is also a weak association. The Chi-square value of .000 implies this result is significant at the one percent level (in this case at 0.1%). I can generalise this result to the wider population because there is one in a thousand chances this result happened by chance. Thus the test hypothesis is true and the Null hypothesis is false. Recoding the variables has changed the results. The 2x2 tables shows that 42.3% of ethnic origins gained degrees, but looking at the table from Appendix 2, I can see that the percentage of degrees obtained amongst the ethnic groups vary: More than 42.3% of ASIAN of Indian and Chinese origin obtained degrees, whilst less than 42.3% of BLACK of African and Caribbean, and Bangladeshis obtained degrees. The 2x2 tables’ show 19.8% of WHITE origin obtained degrees, however in Appendix 2; 13.7% of WHITE of European origin compared to 33.3% of WHITE of other origins obtained degrees. As we can see the educational achievement vary amongst the ethnic minority groups, by clustering them into one single category in the 2x2 tables, the percentage of racial groups who achieve highly, decreases, and those who statistically underachieve, the percentage of degrees obtained increases; the same notion applies to the WHITE race. The larger tables in Appendix 2 gives a clearer perspective of the actual data, the 2x2 tables on the other hand merge the data together, therefore the results are more vague; more general and less specific. New social class and new how satisfied with local doctor and GP; If there was no association between the two variables, I would expect 83.9% of both social classes to be satisfied with their local doctor and GP. However looking at the table I can see this is not the case; 86.1% of the middle class were satisfied with their local doctor and GP compared to 81.1% of working class. Value of Phi is at .068 with Approx. Sig .086 Value of Chi-square is at .086 (Asymp. Sig. 2 sided) Social class and satisfied Epsilon: 86.1% - 81.1% = 5 5/100 = 0.05 Social class and not satisfied Epsilon: 18.9% - 13.9% = 5 5/100 = 0.05 The epsilon value of .05 shows there is a very weak association between social class and satisfaction with local doctor and GP. This is confirmed by Phi value of .068, which is a very weak correlation. The Chi-square value of .086 implies there is 8.6% chance of the Null hypothesis being true to the wider population. The result is significant at the ten percent level; there are 1 in 10 chances the results happened by chance. Recoding the variables has slightly changed the results. The 2x2 tables shows that 86.1% of the middle class were ‘satisfied’ with the local doctor and GP, but the tables in Appendix 2 showed a different picture; SC=1 Very Satisfied 17.5% Quite Satisfied 50.0% = 67.5% SC=2 Very Satisfied 30.5% Quite Satisfied 48.4% = 78.9% SC=3 Very Satisfied 26.2% Quite Satisfied 54.3% = 80.5% The percentage level of how satisfied one is with their local doctor and GP has increased for the middle class in the 2x2 tables. However, it has decreased overall for the working class in the 2x2 tables, in comparison to the previous larger tables in Appendix 2. This is due to the merging of the varying degrees of satisfaction. In the 2x2 tables, I only used the ‘satisfied’ and ‘not satisfied’ values, so values such as ‘neither satisfied nor dissatisfied’ or ‘don’t know’ had no significance in the 2x2 tables, compared to the larger tables in Appendix 2. Therefore it has changed the percentage of satisfaction and dissatisfaction of local doctor and GP amongst the social classes. Part 4) Select a third suitable variable for each table, recode it to produce a dichotomous variable if necessary, elaborate each table by introducing the third variable. For the first relationship; Respondent’s race and highest educational qualification obtained I will be introducing the third variable ‘respondent’s sex’. Note that I would not have to recode the variable because it already has a dichotomous level of measurement. For the second relationship; Respondent’s social class and how satisfied with local doctor and GP I will be introducing the third variable ‘respondent’s age’. I will be recoding the age values into a dichotomous pair, level of measurement. Look at attachments Part 5) Comment on any differences between the values of Phi and Chi-square for each of the elaborated tables and the unelaborated cross-tabulations. New respondent’s race and new highest educational qualification obtained and respondent sex Epsilon: Male and degree: 52.2% - 18.6% = 33.6% 33.6/100 = .336% Male and higher education below degree: 81.4% - 47.8% = 33.6% 33.6/100 = .336% Male: Phi .188 Approx. Sig .000 Chi Asymp. Sig (2 sided) .000 Epsilon: Female and degree: 34.5% - 20.7% = 13.8% 13.8/100 = .138 % Female and higher education below degree: 79.3% - 65.5% = 13.8% 13.8/100 = .138 % Female: Phi .075 Approx. Sig .077 Chi Asymp. Sig (2 sided) .077 The epsilon figure for the previous 2x2 table was .225 Once the third variable was introduced, the male epsilon was higher at .336, this gives evidence that the male epsilon is a stronger association; this is in contrast to the female epsilon of .138 which is a weaker association. Male epsilon is confirmed by the Male Phi value of .188 compared to the previous 2x2 tables value of .124 (the association is stronger here). The Chi-square value has remained the same at the significant value of .000 (Asymp. Sig (2 Sided)). This means there is one in a thousand chance of the null hypothesis being true. The female Phi is weaker at .075 compared to the previous value of .124 (just as the epsilon value is weaker). This gives evidence that it is a weaker association. The Chi-square value has increased from .000 in the previous table to .077 for females. This implies there is now a 7.7% chance of the null hypothesis being true to the wider population. The result is significant at the ten percent level, one in ten chances the results happened by chance. The third variable has specified the relationship; The male results has a stronger correlation than the initial 2x2 table relationship The female results has a weaker correlation than the initial 2x2 table relationship New social class and new how satisfied with local doctor and GP and respondent age Epsilon: 18-44 and satisfied: 83.4% - 75.0% = 8.4% 8.4/100 = .084% 18-44 and not satisfied: 25.0% - 16.6% = 8.4% 8.4/100 = .084% 18-44: Phi .103 Approx. Sig .086 Chi Asymp. Sig (2 sided) .086 45-64+ and satisfied: 88.4% - 84.7% = 3.7% 3.7/100 = .037 % 45-64+ and not satisfied: 15.3% - 11.6% = 3.7% 3.7/100 = .037 % 45-64+: Phi .053 Approx. Sig .310 Chi Asymp. Sig (2 sided) .310 The epsilon figure of the initial 2x2 table was .05 Once the third variable ‘age’ was introduced, the epsilon figure changed. For the age band 18-44, the epsilon figure was stronger at the value of .084, for the age band 45-64+, the epsilon figure was weaker at .037, thus the association is stronger for the age band 18-44 in contrast to 45-64+, the Phi values confirm these findings. The initial Phi value for 2x2 was .068. For the new age 18-44 the Phi value has increased, it is stronger at .103 with the Chi-square value of .086, the same value as the initial 2x2 Chi-square value. So it too has 8.6% chance of the Null hypothesis being true, the result is significant at the ten percent level. The Phi value of the age band 45-64+ is weaker at .053 with the Chi-square value of .310, there is 31% chance of the Null hypothesis being true, which is beyond the ten percent threshold, therefore cannot be generalised to the wider population. Again the third variable has specified the relation between the initial variables; The age band 18-44 has a stronger correlation than the initial 2x2 table relationship The age band 45-64+ has a weaker correlation than the initial 2x2 table relationship Furthermore the findings for the age band 45-64+ cannot be generalised to the whole population because the Chi-square value is beyond the ten percent threshold. Part 6) what conclusions can you draw from these results? Early on in the assignment, using the academic literature (refer to Report part 1 and Appendix 2; 1) I hypothesised “Minority ethnic groups obtain higher education qualifications than their white counterparts” From here I identified two variables from the data file; respondent’s race and highest educational qualification obtained. I hypothesised further using the variable ‘sex of respondent’ “Males within ethnic groups achieve higher in education qualification than their female counterparts” Analysing the data, I found that there is a correlation between the initial variables, confirmed by Cramer’s V value .280. When recoded, the epsilon value was .225 with Phi value.124. It confirmed the hypothesis, there is an association, even if it is a weak one, and furthermore the literature review confirms that there is a correlation. The Chi-square value has remained the same, at the value of .000 in the larger and in the 2x2 tables, therefore the results can be reproduced for I am confident that the variables are associated in the population as a whole. The initial cross-tabulation for the larger tables was based ion a sample of 1412 population; the 2x2 is based on a sample population of 978, which could explain the weaker Phi in the latter cross-tabulations. When I introduced my third variable ‘sex of respondent’; The epsilon for male was .336, with Phi value .188 and Chi-square value of .000 The epsilon for female was .138, with Phi value .075 and Chi-square value of .077 The third variable has specified the initial relationship, by exhibiting that the relationship between race and highest educational qualification obtained, varies amongst males and females. The male association is stronger than the initial relationship, the female association is weaker than the initial relationship, and furthermore the male association is statistically more significant than the female correlation. From these results, I can conclude, assisted by the literature review, that there is an association between race and highest educational qualification obtained, furthermore sex of respondent also has an affect. In other words, I can be confident that race, highest educational qualification obtained are associated for male and female in the population as a whole. I also hypothesised, based on academic literature (refer to Report Part 1 and Appendix 2; 1); Middle class are more likely to be satisfied with their local doctors and GPs because they can gain better access to healthcare compared to the working class Looking at the literature, I was able to identify two variables from the data file, respondents social class and how satisfied with local doctor and GP. I hypothesised further using the variable ‘respondent’s age’; Those of the top age band (45+) in the working class are more likely to be dissatisfied with their local doctor and GP, compared to those of the lower age band (18-44), due to lack of material resources and access to adequate healthcare Through the analysis of the data, I found there is a correlation between the two initial variables; this is confirmed by Cramer’s V value .138 for the larger tables. When the two variables were recoded, the epsilon value was .05; the weak association was confirmed by the Phi value .068. The findings confirmed the hypothesis; there is a relation between race and highest educational qualification obtained even if it is a weak one. Also the literature supports the hypothesis and the findings of the results (Report Part 1, Appendix 2; 1). Initially the Chi-square was .001, in the 2x2 tables the Chi-square value was .086, however the results are still statistically significant to the whole population, because it is below the ten percent threshold, therefore the results can be reproduced. The initial sample of the cross-tabulations was based on a sample of 716; the 2x2 table was based on a sample population of 644, which isn’t a satisfactory sample size for secondary analysis, which could explain the Chi-square significance and the level of association. When the third variable ‘age’ was introduced: The epsilon for ages 18-44 was .084, with Phi value .103 and Chi-square value of .086 The epsilon for ages 45-64+ was .037, with Phi value .053 and Chi-square value of .310 The third variable has specified the relationship between social class and how satisfied with local doctor and GP, amongst the ages 18-44 and 45-64+. The 18-44 relationship is stronger than the initial relation and the results can be generalised to the population as a whole. The 45-64+ relationship is weaker and cannot be generalised to the population because it is not statistically significant, the Chi-square significance is beyond the ten percent threshold. From the analysis, I can conclude, supported by the literature review, that there is an association between social class and how satisfied with local doctor and GP, for those aged 18-44 in the population as a whole. However, due to the statistical insignificance of the results, I cannot be confident that social class, how satisfied with local doctor and GP and are associated for ages 45-64+ in the population as a whole. This goes to show that findings in literature is not always statistically significant to the population, also a large sample is needed in the data to make generalisations on the population, which could also explain the statistical insignificance of the 44-64+ ages results. BIBLIOGRAPHY Babbie, E., (2001), “The Practice of Social Research, 9th Edition”, Wadsworth: USA. Green, D. (ed.), (1988), “Acceptable Inequalities? Essays on the pursuit of equality in Healthcare, IEA Health Unit Paper No. 3”, Goron Pro-Print: Great Britain. Jeffcote, R., (1984), “Ethic Minorities and Education”, Harper and Row Publishers: London. Mason, D., (2000), “Race and Ethnicity in Modern Britain, Second Edition”, Oxford University Press: Oxford. May, T., (1997), “Social Research, Issues, Methods and Process, Second Edition”, Open University Press: Buckingham. Modood et al., (1997), “Ethnic Minorities in Britain”, Policy Studies Institute: London. Pond, C. in Davey, B. and Popay, J. (ed.), (1993), “Dilemmas in Health Care”, Open University Press: UK. Seale, C. (2004) “Researching Society and Culture, Second Edition”, SAGE Publication Ltd: London. Scambler, G. (2002), “Health and Social Change, A Critical Theory”, Open University Press: Buckingham. Smith, D.; Bartley, M. and Blane, D., (1990), “The Black Report on socio-economic inequalities in health 10 years on”, British Medical Journal, Volume. 301, Issue Number. 6748, pages 373-377. Tomlinson, S., in Troyna, B., (ed.), (1987), “Racial Inequality in Education”, Tavistock Publications: London. Lecture Notes: Greg Barnett (23/02/2006), “Introduction to Quantitative Data Analysis”. Greg Barnett (27/04/2006), “Official Statistics in Social Research”. Greg Barnett (13/10/2005), based on Ethnography Read More