Econometric modelling to study dynamic health behaviours at an individual, household, birth cohort a local authority level.

Smoking prevalence has consistently decreased in the UK since 1974, with the latest official reports estimating that 15.1% of the adult population currently smokes. However, there are still 7.6 million smokers and the latest statistics for England suggest that 77,900 people die from diseases caused by smoking. Therefore, smoking continues to be a major public health problem in the UK, as it is for many of the OECD countries. This research proposes to expand our understanding of smoking trends by using a new conceptual and methodological approach, namely a dynamic approach. This approach uses repeated observations through time, which allows studying changes over time. In the context of smoking behaviour, this will entail studying transitions and persistence between and within smoking states and factors determining this, as well as its consequences from a health economic and policy perspective.

The research here proposed will use repeated observations from to four units of analysis to understand: 1) the role of individual and family characteristics as well as the role of social and spatial interactions on the individual's life course smoking decision; 2) the role of cultural norms, risk attitudes and risk preferences in the intergenerational transmission of smoking; 3) the role of birth cohort (or peer interaction) on smoking behaviour in order to expand the work I have been doing with SARG, and 4) the role of space on transitions and persistence between and within smoking status observed through studying smoking behaviour at a Local Authority District level in the UK.

Since smoking is increasingly concentrated among certain population subgroups, this research proposes to add subgroup analysis to the empirical models that will be performed. The main subgroups to explore are gender, socioeconomic status and population with mental health issues, but I will explore different definitions of socioeconomic status and mental health to have a deeper understanding about transitions and persistency in smoking behaviour in each of these groups and about smoking inequalities in general. Subgroup analyses will only be performed when the number of observations allow us to do so.

Smoking behaviour cannot be studied in isolation from the large number and variety of tobacco control policies that have been implemented in the UK during the last forty years. Hence, this research intends to explore the effect of these policies on smoking behaviour dynamics among individuals, households and aggregate social groups. For policy evaluation models using individual level data, this research will need to account for the fact that individuals can transition between smoking states throughout their life course, thus explore methods to use smoking status as a time-varying exposure. Policy analysis will be based on the latest developments in causal inference, in particular, exploring new matching and weighting techniques when appropriate.

The research here proposed will be expanded in two possible ways: first, by providing inputs to SARG, and second, by exploring longitudinal datasets which combine smoking behaviour with other unhealthy lifestyle behaviours such as alcohol misuse, unhealthy diet and sedentarism as contained in the Born in Bradford Study (BiB). By providing inputs to SARG, this research will enable the exploration of further health and economic costs associated to smoking behaviour dynamics, and create specific models for individuals, households, birth cohorts and local authority districts to have a holistic comprehension of smoking as a social, health and economic behaviour. By exploring BiB this research will study the relation between smoking behaviour dynamics at a household level and other unhealthy behaviours in the context of a specific cohort that has been followed since pregnancy.

The research will study smoking behaviour from a dynamic perspective, using econometric methods.

Social and spatial smoking behaviour dynamics:
For birth cohorts and local authority districts, we will create a pseudo panel to explore transitions and persistence in smoking behaviour among social units through time. Smoking behaviour trajectories by birth cohort and local authority district will be combined with individuals and households to understand the role of spatial and social interactions on individual's smoking decision and on the intergenerational transmission of smoking.

Natural history of smoking behaviour:
Using panel data models at an individual level, I will study transitions smoking state throughout an individual's life course, define patterns of smoking histories, their evolution through time as well as the role of individual, social and spatial determinants.

Intergenerational transmission of smoking behaviour:
Transitions will be defined in terms of the intergenerational transmission of smoking states using panel data containing information on both parent and children's smoking behaviour, accounting for the fact that parents may have experienced transitions in smoking states throughout their life course. This research will estimate the probability of passing smoking states from parents to children and how these probabilities have changed over time.

Effect of tobacco control policies
Use causal inference models to study the effect of tobacco control policies on smoking behaviour dynamics at individual, household, birth cohort and local authority level, exploring new matching and weighting techniques as well as time-varying exposures.

Health economic modelling and extension to other unhealthy behaviours:
Model health and economic outcomes associated with trends in smoking behaviour dynamics, and extend the approach to studying the association between smoking dynamics and other unhealthy behaviours using the BiB dataset.

This research will produce at least four academic publications, one for each level of analysis, which will contribute to our understanding of smoking behaviour trajectories and smoking inequalities in the UK as well as to existing literature on dynamic behaviour, intergenerational transmission of unhealthy behaviour and social and spatial determinants of unhealthy behaviours.

The results of this research will be presented at conferences and seminars to extend their impact beyond national academia as well as to benefit decision-makers, government and policy by providing evidence on population subgroups that persist in smoking, population subgroups that easily transition to former smoking, population subgroups that persist in never smoking and population subgroups that still transition from never to current smoking. This could help to target resources in terms of needs of cessation support, potential areas/groups where the health consequences of smoking are going to be concentrated, as well as areas/groups where good practices in terms of sustained concentrations of never smoking population can be found and, potentially, reproduced.

The results of this research will also contribute to deliverables proposed in SPECTRUM in the work packages proposed both by University of Sheffield and University of Nottingham. Through SPECTRUM this research will reach other universities in the UK and Australia as well as Public Health England, Health Scotland, Public Health Wales and the business sector. This can lead to future collaborations with the academic, public and private sector to extend the methods used in this research into other unhealthy behaviours included in SPECTRUM (alcohol and unhealthy food and drink) as well as contribute to policy making.

This research will be based on existing data sources, exploring them from a different perspective in order to expand the knowledge on smoking behaviour to a dynamic approach. It will use data available through the UK Data Service and data such as BiB Study which is available upon request. This will benefit the data owners since more publications will be produced with their data and new research projects can potentially emerge from the new approach proposed. In particular, by using the BiB data I will produce results that can be relevant for ActEarly, a recently awarded UKPRP consortium which aims to develop, implement and evaluate early life interventions intended to reduce health inequalities.

This research will contribute to the literature on policy evaluation (or causal inference) by writing academic publications and by presenting at technical conferences (like the European Causal Inference Conference or the European and International Conference on Health Economics) and in more topic-specific conferences such as the Society for Research on Nicotine and Tobacco conferences. It will also contribute to policy making through participating at SPECTRUM events and by providing inputs to SARG and relevant results to ActEarly, allowing forecasting the economic and health costs of the dynamics of smoking behaviour which is usually of interest for the annual reports from Royal College of Physicians, to which I have been contributing since 2016.

Based on my experience on working in different countries, I will seek collaborations with other institutions abroad which would benefit enormously from the research in tobacco developed in the UK as well as from the data sources used to collect information and study smoking behaviour in this country. I will contact relevant stakeholders in Chile, Spain and the US through my past work experience and research collaborations to strengthen international collaboration on research on smoking behaviour.