Nonsampling Errors in Polls and Surveys Dru Rose

Non-sampling Errors in Polls and Surveys Dru Rose C 1, L 2, S 1

Methods of Data Collection Surveys and Polls Experimental Studies Observational Studies Dru Rose C 1, L 2, S 2

Polls and Surveys Poll Survey § Few questions § Multi-choice type only § Many questions § May have branches and skips § May have a “tick multiple boxes” option § May have open-ended, write a comment questions Dru Rose C 1, L 2, S 3

Sampling Errors (random process) Non-sampling Errors Selection bias Non-response bias Self selection bias Question effects Behavioural considerations Interviewer effects Survey-format effects Transfer of findings C 1, L 2, S 4

Selection Bias: Population sampled is not exactly the population of interest. Target population (e. g. adults in NZ) Not included in sampling frame Cannot be contacted Refuse to respond Incapable of responding Sampling frame (e. g. households with a landline phone) SAMPLED POPULATION Not eligible for survey

Sources of Non-sampling Errors Non-response bias When people who have been targeted to be surveyed do not respond: Potential bias if non-respondents are likley to behave differently to respondents with respect to the question being asked. e. g. Non-respondents in an employment survey are likely to be those who work long hours. C 1, L 2, S 6

Sources of Non-sampling Errors Self-selection bias People decide themselves whether to be surveyed or not. Dru Rose C 1, L 2, S 7

Self-selection bias: phone-in or internet polls C 1, L 2, S 8

Sources of Non-sampling Errors Question effects Subtle variations in wording can have an effect on responses. e. g. “Should euthanasia be legal? ” vs. “Should voluntary euthanasia be legal? ” People are more likely to favour “voluntary” euthanasia. C 1, L 2, S 10

18 August 1980 New York Times/CBS News Poll “Do you think there should be an amendment to the constitution prohibiting abortions? ” Yes 29% No 62% Later the same people were asked: “Do you think there should be an amendment to the constitution protecting the life of the unborn child? ” Yes 50% No 39% C 1, L 2, S 11

Question Effects in the NZ Census 1986: “What is your ethnic origin? (Tick the box or boxes which apply to you. ) 1991: “Which ethnic group do you belong to? ” (Tick the box or boxes which apply to you. ) 1996: “Tick as many circles as you need to show which ethnic group(s) you belong to. ” Ethnicity Single Ethnicity European Maori Two Ethnicities European & Maori Two European gps 1986 94. 6 1991 94. 3 1996 81. 0 81. 2 9. 1 4. 0 2. 9 0. 0 78. 1 9. 6 4. 5 2. 7 0. 6 65. 8 7. 6 11. 2 4. 7 4. 5 C 1, L 2, S 12

Sources of Non-sampling Errors Behavioural considerations People tend to answer questions in a way they consider to be socially desirable. e. g. pregnant women being asked about their drinking habits may be reluctant to admit that they drink alcohol Dru Rose C 1, L 2, S 13

Sources of Non-sampling Errors Interviewer effects Different interviewers asking the same question can obtain different results. e. g. the sex, race, religion , manner of the interviewer may influence how people respond to a particular question. Dru Rose C 1, L 2, S 14

Interviewer Effects in Racial Questions In 1968, one year after a major racial disturbance in Detroit, a sample of black residents were asked: “Do you personally feel that you trust most white people, some white people or none at all? ” • White interviewer: 35% answered “most” • Black interviewer: 7% answered “most” C 1, L 2, S 15

Sources of Non-sampling Errors Survey-format effects -question order e. g. “To what extent do you think teenagers are affected by peer pressure when drinking alcohol ? ” followed by: “ Name the top 5 peer pressures you think teenagers face today. ” -survey layout -interviewed by phone or in-person or mail. C 1, L 2, S 16

Sources of Non-sampling Errors Transferring findings Taking the data from one population and transferring the results to another. e. g. Auckland opinions may not be a good indication of New Zealand opinions. Auckland sample New Zealand C 1, L 2, S 17

Non-sampling Errors • can be much larger than sampling errors • are always present • can be virtually impossible to correct for after the completion of survey • virtually impossible to determine how badly they will affect the result • good surveys try to minimize them in the design of the survey (e. g. do a pilot survey first) C 1, L 2, S 18

Surveys / Polls A report on a sample survey/poll should include: – Who carried it out and who funded it – target population (population of interest) – sample selection method – the sample size and the margin of error – the date of the survey – the exact question(s) being asked – the results – the claims (inferences) made C 1, L 2, S 19