Survey Design


The research design for the Millennium Cohort Study (MCS) was based on the following five principles:

  1. The MCS should provide data about children living and growing up in each of the four countries of the UK.
  2. The MCS should provide usable data for sub-groups of children, in particular those living in advantaged and disadvantaged circumstances, and for children of ethnic minorities and those living in Scotland, Wales and Northern Ireland.
  3. As well data about children, the MCS should provide data about their family circumstances and the broader socio-economic context in which the children grow up.
  4. The MCS should include children born throughout a single 12-month period.
  5. All children born as members of the MCS population should have a known and non-zero probability of being included in the selected sample.

In addition, there were a number of practical constraints that influenced the chosen design:

  1. The sample had to include a substantial proportion of children born in the year 2000 and measured in their first year of life.
  2. A planning period that was very short, starting in May 2000 to prepare for fieldwork commencing one year later.
  3. A design that facilitated comparisons with earlier UK birth cohort studies, notably those that started in 1946, 1958 and 1970, was essential.

In brief, the fundamental aim of the research design, and in particular the sample design, was to ensure a proper representation of the total population, while at the same time having sufficient numbers of key subgroups for analysis. Probability (or random) methods of selection combined with stratification and clustering would achieve these aims, and would enable the vagaries of sampling to be properly accounted for through the computation of theoretically sound sampling errors.

A technical report on infant mortality has been produced in connection with principle 5 in the top list.  It investigates how many children in the birth cohort (from which the MCS was sampled) did not survive long enough to be included in the survey which took place at 9 months.  Not only would infant mortality within the birth cohort result in some cases with early deaths being missed out of MCS, but the possibility of differential infant mortality between social groups could bias the sample of survivors.