Non-response in the 1970 British Cohort Study (BCS70) from birth to 34 years

Key takeaways

content: "@ketendeNonresponse1970British2010" -file:@ketendeNonresponse1970British2010

Reading notes

Imported on 2024-05-07 21:38

⭐ Important

• & However, the original BCS70 cohort children in Northern Ireland were not followed up in these successive waves. (p. 4)
• & at age 26, the cohort child had to make their own decision to be part of the study. Recorded data for these past waves of BCS70 data did not include details of whether non-response was due to non-contact or to refusal, so analysis of this distinction was not possible. (p. 4)
• Replicate this table in some manner for the data section of BCS:
  • & Table 1: BCS70 estimated longitudinal target and observed sample, wave0 to 6 (p. 5)
• & Table 2: BCS70 summary of response in earlier sweeps (p. 6)
• & It is worth commenting that BCS70 response rates do not follow what has come to be seen as the expected pattern for longitudinal wave-on-wave responses rates. This common pattern, observed especially in annual household panel studies, sees wave2 response rates suffer the largest fall, but wave-on-wave response rates from 82.1 92.8 77.1 63.5 83 81.3 0 10 20 30 40 50 60 70 80 90 100 per cent 123456 BCS70 wave Series (p. 6)
• & 7 wave3 onwards being consistently high and at rates that are often in the 90 per cent and above range. There are a number of reasons why BCS70 did not realize this common pattern; the reasons include:  uncertainties in the funding leading to very long gaps between contacts;  the use of lower quality modes (e.g. postal surveys); and  a factor that is inherent to birth cohort studies, but not to panel surveys, the change over from parent to cohort member in commitment to the study needing to take place along the way. (p. 7)
• & The variables shown in Table 3 are those observed at birth and are used to estimate cross-sectional response at each BCS70 wave from age 5 to 34. Variables available from wave0 were mainly about the parents and the family context into which the child was born. These include the mother’s age at the birth of the cohort child, the father’s social class using the older classification of the Registrar General’s six categories, approximate indicators of mother’s and father’s level of education, number of children, marital status of parents, whether mother lived in London, whether breastfeeding was attempted and the gender and birth weight of the cohort child (p. 7)
• & The results from a multivariate logistic response model using these wave0 variables are displayed Table 4. The same independent variables from wave0 were used to predict response for every BCS70 wave. (p. 9)
• & The age of the mother at the birth of the cohort child was a significant predictor of response at all waves with families of younger mothers at birth being less likely to respond and older mothers more likely to respond at each wave. Social class of the father of the cohort child measured in 1970 was also a significant predictor of response in all waves but more so in waves when cohort members were the main respondent, age 26 and beyond. Families with a father whose social class was non manual were about 30 per cent more likely to participate in waves2 through to wave6 compared to families with a father in the skilled manual social class. Having a father from the lowest social class groups (4 and 5) was generally associated with lower response in all waves in comparison to a father in the skilled manual social class. The birth weight of the cohort child was a statistically significant predictor of response in all waves with large associations in earlier waves than in later waves. The heavier the cohort child at birth the more likely the family was to participate in the survey. This relationship continued in the same direction at age 26, 30 and 34 surveys when cohort members were the main respondent. Mothers’ and fathers’ ages at leaving school were significant predictors of responding at wave1 and wave2 (ages 5 and 10) but not at age 16 (wave3) and above. Compared with mothers or fathers who left school age 15, cohort members leaving school at other ages tended to have lower responses. (p. 10)
• & Families with more than two children (including the cohort child) were significantly less likely to participate in wave1, wave2 and wave3 and the cohort members who had more than one sibling were also less likely to participate in wave4, wave5 and wave6 than cohort members who were the only child at birth. Families in which parents were married were significantly more likely to participate in all waves and cohort members whose mothers were married were also more likely to participate in all waves. Families where no attempt was made to breastfeed the cohort child were significantly less likely to participate in all waves. Cohort members who were not breastfed were also significantly less likely to participate in surveys after age 26, when they had to decide for themselves to respond. Mothers residing in London at the birth of the cohort child were significantly less likely to participate in all waves (up to and including age 16) and the cohort children were also less likely to participate in all waves from age 26 onwards, than mothers residing outside of London. (p. 11)
• & Table 8. Wave response rates of BCS70 cohort members according to whether they participated in the age 21 years sub-study. (p. 19)