Presentation on theme: "Melissa Davy Office for National Statistics"— Presentation transcript:
1Melissa Davy Office for National Statistics Pseudo-cohort analysis - trends in smoking data using the General Household SurveyMelissa DavyOffice for National Statistics
2Introduction Why we carried out pseudo cohort analysis (PCA) The advantages and disadvantagesThe survey we used in our analysisThe methods we usedWork through an example
3Why is it useful to look at cohort analysis? Interested in inequalities over time and by birth cohortPeople in different birth cohorts have different experiencesCohort analysis provides a better understanding of how events change over time
4What is pseudo-cohort analysis? - panel data- same individualPseudo-cohort analysis- cross-sectional data- average experience of a given cohortFor example- aged 20 to 25 in a 1980 survey= 21 to 26 in 1981= 22 to 27 in 1982= 44 to 49 in 2004.
5Advantages of pseudo-cohort analysis Uses data that are already availableNew sample each year so no problem of non-random attritionLess burden on respondentsMore frequent data
6Disadvantages of pseudo-cohort analysis Variations in the nature of the samples surveyedLooking at average experience of the cohort limits the use of the dataRecall biasNot straightforwardSmall cell sizes
7General Household Survey Dataset goes back more than 30 yearsThe GHS covers a range of topicsRelatively large sample sizeHigh quality data source
8Extracting the data– create a database which includes all survey years– create a birth cohort variableDisadvantages- Time consumingAdvantagesValuable research toolExploiting full potential of the GHS dataMakes time series analysis easier
9Creating the dataset Long process documented in: Uren Z (2006) The GHS Pseudo Cohort Dataset (GHSPCD): Introduction and Methodology. Survey Methodology Bulletin, no 59, pp25-37.Available at:contains over 40 variables,records for over 800,000 individuals.
10The smoking example Analysis of smoking among men trends over time trends by agepseudo-cohort analysis- interaction of time and age
24Main findingsAt every age, men smoke less than the previous generation.Not due to established smokers giving up more rapidlyIn the most recent cohorts- fewer men were starting to smokebut were then giving up at a slower rate than in the past.Smoking prevalence rates may be stabilising.
25ConclusionPseudo-cohort analysis gives us a better understanding of the GHS smoking data