Is There a Genetic Basis of Risk, Time and Social Preferences?

By Thorsten Dickhaus, Humboldt-University Berlin

In a joint international collaboration, researchers from seven European universities are studying the extent to which a person's genetic makeup determines social, time and risk preferences.

Human beings are very diverse with respect to how many risks they take, how patient they are and how altruistically they behave towards others. These are characteristics of the human disposition which social scientists often refer to as social, risk and time preferences. This variation in preferences is related to a number of clearly definable factors, such as gender and age. But it is also related to differences in the general living environments people find themselves in and to their upbringing and personal histories. Their experiences, in particular those of early childhood, as well as their familial, social and cultural contexts profoundly shape their lives.

In fact, a large number of existing scientific studies have documented the extent to which diversity in human preferences is associated with gender, age and education, as well as socio-demographic and economic backgrounds. The observed diversity in preferences, however, could also be genetically determined. In fact, recent studies from genetics, as well as those with adoptees and with twins, have provided conclusive evidence that many facets of human behaviour are, at least to some extent, inheritable.

The goal of our study is to explore the extent to which a person's genetic makeup determines preferences. A better understanding of the deep neurobiological roots of human preferences is crucial in better comprehending human behaviour and its evolutionary origins. In order to realize this ambitious objective, our group of interdisciplinary researchers from economics, statistics and machine learning, computer science, genetics and neuroscience from Zurich, Barcelona, Berlin, Mainz, Innsbruck, Konstanz and Munich began collaborating some time ago. Currently, a so-called genome-wide association analysis, including approximately 2,000 individuals from different European cities, is in progress. This analysis investigates the relationships between genotypic information (i.e., people's genetic make-up) and phenotypic information (i.e., their preferences). The future goal of this analysis is to precisely pinpoint the loci on the human genome associated with the phenotypes. Moreover, we are also in the process of designing follow-up analyses in order to be able to confirm our genetic findings. These include identifying the neurological pathways involved — the brain regions underlying the expression of certain genes in terms of specific forms of human behaviour.

It is important to note that the relationship between the genome and the resulting behaviour is actually quite weak. In sum, environmental factors, such as upbringing, ultimately have the deciding effect on human behaviour. Moreover, two additional facts make our project a very complex endeavour. First, as one can easily imagine, even in simple decision situations people’s behaviour is extremely complex and driven by many different motives. This means that the isolation of specific behavioural motives requires precise and controlled measurement methods. Second, a very large set of different and non-adjacent positions on the human genome might simultaneously be responsible for people’s preferences.

The fact that the possible genetic effects are so weak and highly dispersed over the genome is the reason why our study requires innovative statistical methods that will disentangle relevant from irrelevant genetic positions. Thus an additional focus of our research project is the development of new statistical methodologies. The specific challenge in statistical methodology occurs because of the informational size of the human genome. This leads to a large number of simultaneous statistical hypotheses testing problems. Here, novel approaches from the field of structural statistical inference can help exploit structures on the data space or in the systems of hypotheses to be tested in a way to avoid the "curse of dimensionality" — the classical problem that sample sizes are smaller than the number of statistical parameters under consideration. To do so, we combine techniques from the fields of multiple testing and statistical machine learning. In summary, our study combines the attempt to answer a longstanding substantive question — the extent to which human behaviour is determined by the human genetic makeup — with the development of new methods for dealing with large-scale interdisciplinary datasets, as found in the fields of medicine, neuroscience and genetics.

 

(Jun.-Prof. Dr.) Thorsten Dickhaus
Humboldt-University Berlin
www.atomiumculture.eu

And the following contributors:

(Prof. Dr.) Daniel Schunk (University of Zurich and University of Mainz),
(Prof. Dr.) Ernst Fehr (University of Zurich),
(Prof. Dr.) Klaus-Robert Müller, (Dr.) Sören Sonnenburg (Berlin Institute of Technology),
(Prof. Dr.) Arcadi Navarro, Carlos Morcillo (University Pompeu Fabra, Barcelona),
(Prof. Dr.) Klaus Schmidt (LMU Munich),
(Prof. Dr.) Matthias Sutter (University of Innsbruck),
(Prof. Dr.) Urs Fischbacher (University of Konstanz).