Longitudinal Data

Longitudinal data stores information on the same units over time. The recorded information might be qualitative or quantitative, and might concern different types of observational units, such as individuals or country. Since the same individuals are tracked over time, longitudinal data makes it possible to study how people evolve or change over time.
Longitudinal data can be contrasted with cross-sectional data. In the latter, data are collected at one point in time. If such studies are repeated over time, they allow describing societal or aggregated changes. However, it is not possible to understand who changed their behaviors or to describe the individual evolution of behavior over time. Moreover, longitudinal data are necessary to explore causality, since cross-sectional data only reveals associations.
In some cases, longitudinal data collections start with an event shared by all units documented, either an exhaustive population or a representative sample. Such data collection efforts are called cohorts. For example, the Swiss Transplant Cohort Study collects since 2008 longitudinal data on all solid organ transplant recipients in Switzerland. Of particular interest to life course studies are birth cohorts, like the cohort ELFE in France which started with a sample of babies born in that country in 2011. Longitudinal data collection can also be launched by a cross-sectional form of data gathering, whose individuals are then followed over time, for example the Swiss Household Panel or the Swiss National Cohort linking administrative records to individuals monitored in the 1990 and 2000 censuses. In low- and middle-income countries demographic surveillance systems also undertake longitudinal data collections. In this case, geographical areas are followed; after an initial census, survey rounds at least once a year collects all vital events (arrival, birth, deaths, departures) in the surveillance areas.
The LIVES NCCR heavily rely on longitudinal data to study vulnerability over the life course. For instance, it allows understanding how people recover (or not) from a disruptive event overt time. The LIVES NCCR collected several qualitative and quantitative longitudinal datasets. A full list of these datasets is available here: https://www.centre-lives.ch/fr/lives-data-collections A distinction can be made between prospective and retrospective longitudinal data (Scott & Alwyn, 1998). In prospective studies, a sample is followed and observed over time or at different points in time. The information is stored as the sample evolves. For instance, panel data stores the values of the same set of variables for the same sample at different point in time, resulting in several observations per individual units. Similarly, administrative data might record the situation of an individual at different points in time, for instance based on yearly data extraction.
Longitudinal retrospective data are collected by reconstituting the past life course of a sample of respondent. This might be achieved by asking the respondents to recall their past. In this case, several tools were developed by life-course researchers to minimize memory errors or bias. The NCCR LIVES actively participated in the development of life history calendars (Morselli et al., 2016). This might also be achieved using archive or other persistent trace of activity.
On top of allowing the investigation of issues of causality, according to Morselli et al. (2016), one of the main advantages of prospective data collection is the ability to ask subjective or detailed questions. It also avoids recall bias or the re-interpretation of past events. However, the process is costly and it involves keeping contact with respondents to avoid dropouts. Drop-out can end up being high, and different statistical techniques are used to control for the selectivity of those remaining under observation, without solving this issue entirely. On the other hand, retrospective questionnaire can be conducted only once and record a long-life course. However, detailed or subjective retrospective questions are generally avoided as there are more prone to recall bias.
The analysis of longitudinal data requires specific methods to describe and understand individual change over time. The NCCR LIVES is also active in the development of such method for qualitative or quantitative data.

Authors: Matthias Studer, Clémentine Rossier

References

Bernardi, L. and Sánchez-Mira, N. (2021) Introduction to the special issue: Prospective qualitative research: new directions, opportunities and challenges, Longitudinal and Life Course Studies, vol 12, no 1, 3–5.
Morselli D., Dasoki N., Gabriel R., Gauthier JA., Henke J., Le Goff JM. (2016) Using Life History Calendars to Survey Vulnerability. In: Oris M., Roberts C., Joye D., Ernst Stähli M. (eds) Surveying Human Vulnerabilities across the Life Course. Life Course Research and Social Policies, vol 3. Springer, Cham. https://doi.org/10.1007/978-3-319-24157-9_8
Scott, J., & Alwyn, D. F. (1998). Retrospective vs prospective measurement of life histories in longitudinal research. In J. Z. Giele & G. H. Elder (Eds.), Methods in life course research. Qualitative and quantitative approaches (pp. 98–127). Thousand Oaks: Sage. Park, A. & Rainsberry, M. (2020). Introduction to longitudinal Studies, Closer learning hub https://learning.closer.ac.uk/learning-modules/introduction/

Semantic network visualisation

Click to activate zoom- and drag-fonctionnality (scroll to zoom, drag nodes to move, click and hold nodes to open next level)