The Mid-Twentieth Century Baby Boom and the Role of Social Interaction: An Agent-Based Modelling Approach
André Grow, University of Leuven
Jan Van Bavel, University of Leuven
Eli Nomes, University of Leuven
Studying dynamic social influence processes in social network structures is difficult with standard demographic tools. Thus, to theoretically explore the mechanism that we propose, we make use of agent-based computational modelling. This will enable us to develop and explicate theories about social behaviour and assess the implications of their assumptions by means of computational simulation and experimentation. Using input data from Belgian censuses, we simulate the diffusion of the two-child norm among social and spatial dimensions and the resulting fertility outcomes. Preliminary results show that this diffusion process can indeed account for the variety of cohort fertility trends we observe.
Around the middle of the 20th century,most western countries experienced a surge in birth rates, the BabyBoom. This was unexpected at the time, and the causes and underlyingsocial mechanisms remain unclear. In this paper we suggest that anormative shift, propelled by the power of social interaction, might have beenone of the main drivers of the observed shifts in timing and quantum offertility, namely the emergence and diffusion of the two-child norm. If newtypes of childbearing behaviour emerge at some point in the population,diffusion via social networks can translate these changes on the micro levelinto large scale changes on the macro level. At the end of the 20th century,the two-child family clearly was a well-established norm. The roots of thistrend probably go back to the 19th century fertility transition but it wasduring the Baby Boom era that the two-child family won a lot of ground. Thisnormative shift of family size lead to higher birth rates as childlessnessbecame rare. However, among groups and regions where larger family sizes werestill common, it contributed to a further fertility decline. The resulting homogenization couldtherefore explain similarities and dissimilarities across regions and socialgroups we observe in the country of our focus: Belgium (see Figure 1). Inthe north of the country, especially in rural areas like Limburg, families withmore than four children were still very prevalent during the first decades ofthe 20th century, so that a shift to the two-child norm actuallyimplied a decrease of fertility. In the south of the country and in the bigcities, childlessness and one-child families were more prevalent, so that theshift to the two-child norm caused an increase in fertility.
Figure1: Family size by province among the Baby Boom producing birth cohorts,1910-1940, Belgium.
Studying dynamic social influenceprocesses in social network structures is difficult with standard demographictools. Thus, to theoretically explore the mechanism that we propose in thispaper, we make use of agent-based computational modelling. This will enable usto develop and explicate theories about social behaviour and assess theimplications of their assumptions by means of computational simulation andexperimentation. Using input data from Belgian censuses, we simulate thediffusion of the two-child norm among social and spatial dimensions. An agent i in our model will adapt the two-childnorm at time t with a certainprobability PA depending on the shareof people in her social network SA,which is determined by spatial and educational constraints, that are alreadyadhering to this norm (1). In a second step, we incorporate fertility outcomesin the model: women have children given observed birth probabilities PB depending on age, parity, educationand region, adjusted by a multiplier MPdependent on whether they have adopted the two-child norm and whether theyalready have two children or not (2). The parameter gamma expresses the influence ofthe two-child ideal on actual fertility outcomes.
We calibrated the model using theprovince of Antwerp. With the resulting parameters, we can simulate the fertilitytrends quite well, as Figure 2 shows. When setting the gamma parameter to 0, cancellingout any effects of social influence, cohort fertility trends in all provinces havea negative slope, as the only factor driving fertility in our model thatremains is the shift to higher education and thus lower birth probabilities.However, when we turn on the effect of the diffusion of the two-child norm,the simulated fertility patterns closely resemble the observed trends: in mostprovinces we see an increase in cohort fertility among the older cohorts, andthe start of a decrease among the youngest cohorts. The former is driven by theshift from childlessness to two-child families, the latter is driven by theshift to higher education. However, the model also simulates the observeddecrease of fertility in Limburg very well, which is driven by the shift fromlarger families to two-child families.
Our simple diffusion model alreadyyields promising results. The diffusion process of the same norm could thusaccount for the variety of cohort fertility trends we observed. In the fullpaper, we will examine the parameter space to further assess the relationbetween several starting situations, the diffusion process and the modeloutcomes, as to take full advantage of the possibilities a computationalsimulation approach has to offer.
Presented in Session 1156: Fertility