Longitudinal research

Alison Smith, NRDC researcher at the Institute of Education, describes the longitudinal approach to research and considers its value for educational research and the Skills for Life strategy.

One of the limitations of most survey-style research is that it gives only a snapshot of the social context being studied. On its own, it is unable to track changes over time or to identify the cause-and-effect of, for example, an educational policy, by carrying out before-and-after studies.   

Tracking change over time

Longitudinal studies are an attempt to respond to this problem and usually take one of two forms. In one form, such as that taken by the British Social Attitudes Survey or the British Crime Survey, a sample of the population is surveyed at regular intervals and any changes over time are measured and reported. However, the sample is made up of different individuals on each occasion. This sort of study gives us a series of snapshots but, in each snapshot, we see different individuals so we cannot be sure that any changes we observe are due to differences between these individuals or to the individuals themselves having changed after exposure over time to what we are trying to measure (such as the impact of the Skills for Life strategy). 

In the other form, the panel or cohort study, the same group of respondents are surveyed on a number of occasions over a period of months, years or decades. In this form of research, we can be more confident that we are tracking the real effect of social changes, provided of course that our original cohort is representative of the population we wish to study.     

A well-known example of a cohort study is the National Child Development Study, which has been following the development of all the 17 000 children born in Britain in the week 3-9 March 1958. It is assumed that they are representative of their generation. A whole series of reports, covering a range of themes and by different authors, have been based on the data collected by this study.  

Another example, on a much smaller scale, is the NRDC's Teachers' Study. This study is collecting data from a panel of Skills for Life teachers and trainers, covering information about, for example, individual wages, hours worked, and level of education. These teachers were first interviewed in the academic year 2004-05 and will be re-interviewed at several subsequent points in time.  

Cause and effect

The opportunity to study change in the same group of individuals as a result of policy change makes longitudinal data invaluable to educational research. When evaluating educational policy, the researcher's goal is to test whether one variable has a causal effect on another (such as the effect of teacher qualifications on learners' success). Simply finding an association between these two variables might be suggestive but, unless causality can be established by looking at before-and-after data, it is rarely compelling.  

Another advantage of panel data is that it may also allow us to study the importance of time delays in behaviour, or the medium- to long-term result of exposure to learning. This information can be significant since many Skills for Life programmes can be expected to have an impact only after some time has passed. 

Bynner and Parsons

The work of John Bynner and Samantha Parsons (see page 32 of this edition of reflect), makes use of the data collected by the National Child Development Study and by the British Cohort Study, whose cohort is the 17000 babies born in one week in April 1970. Members of this cohort have been surveyed in 1975, 1980, 1986, 1991, 1996, and 2000. Some 11 000 of the original cohort are still involved. Bynner and Parsons have reanalysed the data about these individuals, including tests of their levels of literacy and numeracy, to show how these affect their life-course and life-chances. 

---

For more information about a wide range  of longitudinal research, see the websites  of the UK longitudinal Studies Centre  (ULSC) at http://iserwww.essex.ac.uk/ulsc  and of the Centre for Longitudinal Studies  at www.cls.ioe.ac.uk