Mater certa est, pater numquam: what can Facebook advertising data tell us about male fertility rates?
In: Association for the Advancement of Artificial Intelligence (Ed.): Proceedings of the Twelfth International AAAI Conference on Web and Social Media (ICWSM 2018): 25-28 June 2018, Stanford, California, 672–675
Palo Alto, CA, AAAI Press (2018)
In many developing countries, timely and accurate information about birth rates and other demographic indicators is still lacking, especially for male fertility rates. Using anonymous and aggregate data from Facebook's Advertising Platform, we produce global estimates of the Mean Age at Childbearing (MAC), a key indicator of fertility postponement. Our analysis indicates that fertility measures based on Facebook data are highly correlated with conventional indicators based on traditional data, for those countries for which we have statistics. For instance, the correlation of the MAC computed using Facebook and United Nations data is 0.47 (p = 4.02e-08) and 0.79 (p = 2.2e-15) for female and male respectively. Out of sample validation for a simple regression model indicates that the mean absolute percentage error is 2.3%. We use the linear model and Facebook data to produce estimates of the male MAC for countries for which we do not have data.