Event history

Presenter(s): Bernice Kuang

The UK Generations and Gender Survey is a new dataset with retrospective partnership histories rich with opportunities for event history analysis.  It is part of a wider, cross national data collection effort conducted by the Generations and Gender Programme, which includes Generations and Gender Surveys (GGS) from multiple countries worldwide. From the Generations and Gender Surveys, a dataset called the Harmonised Histories is derived. The Harmonised Histories is a dataset containing data across individual countries, standardised for cross-national comparison and set up for comparative event-history analysis.  Its content focuses on life histories, or life events such as having children, moving in with a partner, getting married, and getting divorced. 

Many national or regional surveys collect retrospective “life history” data (e.g., about partnership, fertility, education, or employment histories), but they differ in structures, definitions, and coding schemes. This makes it difficult to compare across countries or datasets directly. Without standardisation, apparent differences across countries could simply reflect how surveys categorise and record events rather than real social variation. For example, one country may appear to have higher marriage rates simply because cohabiting partnerships or civil partnerships were not distinguished as a separate category. The Harmonised Histories standardises retrospective life histories from different countries’ GGSs into a single, consistent, cross-nationally comparable dataset, in a format that is easy to apply event history analysis to, making it a rich source of information for comparing countries.

In practical terms, the Harmonised Histories uses the retrospective history data (such as partnership or fertility histories) from multiple countries’ standard GGSs, then re-codes and re-structures the information so that variables align across contexts and are easy for the user to analyse using event history. Examples of alignment include ensuring consistent start and stop dates or shared definitions of events like partnership or separation. In fertility histories, this might mean aligning how “birth order” is recorded or whether stillbirths are included, while in education data, it involves reconciling different school-leaving ages or qualification categories to ISCED standard.

Tour of the UK Generations and Gender Survey and Harmonised Histories

• Walk through the structure of the UK Generations and Gender Survey
• Show how the UK GGS is restructured into Harmonised History format
• Introduce the existing Harmonised History dataset, including Norway, Czechia, Estonia, Finland, Croatia and where to find and download the datasets

The UK Generations and Gender Survey (GGS) collects detailed information on individuals’ partnership and fertility histories. For each respondent, it records the sequence of cohabiting relationships, including start and end dates, marriage or divorce, and the number of children within each partnership. This retrospective structure allows researchers to reconstruct full life-course timelines of partnership and fertility events. The data are then restructured and standardised into the Harmonised Histories format, aligning variable names, coding schemes, and time definitions across countries. This process produces consistent, comparable datasets that enable cross-national event history analyses of family formation, union transitions, and fertility patterns which are particularly user friendly.

Non-Parametric and Parametric Event History Models

Practical using STATA event history analysis commands:

• This practical will demonstrate how to conduct an event history analysis using the example of transition to first birth
• Analyse the transition to first birth using a Kaplan-Meier approach
• Plot the comparison between survival curves
• Plot the comparison between cumulative hazard curves
• Analyse transition to first birth by background characteristics (i.e. education, number of siblings, and year of birth)

Event history analysis can be applied to fertility histories to understand when individuals experience key life events such as the birth of their first child. The process begins by defining the event of interest (a first live birth) and identifying when individuals enter the risk period; for women and girls, this is typically set at age 15.  Following this, the time at which individuals leave the risk period must also be identified.  This will be either after they have experienced the event (i.e. you cannot have a first birth after you have already given birth), at age 50 when biological childbearing is very unlikely, or at the time of the interview if the event has not occurred. The data are declared as survival time data using stset, which establishes the time structure and event indicator. Researchers then use Kaplan–Meier and cumulative hazard plots to visualise differences in timing across groups, followed by a parametric survival models (streg) to estimate how characteristics such as education, number of siblings, and birth cohort are associated with the rate and timing of first birth.

Competing risks model for transition to marriage/separation

Practical using STATA, manually restructuring the data for event history analysis:

• This presentation will demonstrate how to prepare the data for event history analysis with a competing risks framework, using partnership events as the example  
• Restructure the dataset into a person-year format, where each row represents one year of a respondent’s life course and indicates their partnership status at that time.
• With the data in this format, multinomial logistic regression can be applied to conduct discrete-time hazard modelling. 

 

Event history analysis can also address situations where more than one outcome may occur, such as the transition from cohabitation to either marriage or separation. The process begins by defining the start of the risk period at the formation of the partnership, when both events become possible. The duration of each partnership is calculated from its start to the time of the event or censoring. Censoring is the end of the observation period in cases when the event does not occur, such as if a partner dies or at the time of interview. The dataset is then expanded into a person-year format so that each row represents one year of an individual’s relationship history. Within this structure, each year of each individual’s observation period is labelled according to what their relationship status is in that year. A multinomial logistic regression (mlogit) is then used to estimate how factors such as education, birth cohort, or age at partnership formation are associated with the relative likelihood of marriage or separation compared with remaining cohabiting.

> Download Worksheet.

Code containing worked examples from STATA analyses based on Harmonised Histories 

About the author

Dr. Bernice Kuang is a family demography researcher at the University of Southampton, on the Generations and Gender Survey project, studying partnering, fertility, and childcare patterns across the UK and Europe.  She completed her PhD in Social Statistics and Demography at the University of Southampton, and her master’s in public health, with a concentration in demography, at Johns Hopkins University.  She has worked on various USAID and privately funded population health projects as a technical advisor in family planning and reproductive health. 

Primary author profile page

• Published on: 10 December 2025
• Event hosted by: University of Southampton
• Keywords: Critical theory | Humanistic methods | Interpretivism | Positivism | Poststructuralism | Descriptive research | Exploratory research | Explanatory research and causal analysis | Comparative and cross-national research | Cross-national research | Cross-cultural research |
• To cite this resource:

  Bernice Kuang. (2025). Event history. National Centre for Research Methods online learning resource. Available at https://www.ncrm.ac.uk/resources/online/all/?id=20861 [accessed: 14 December 2025]

⌃
BACK TO TOP