Forecasting number of births and sex ratio at birth in Iran using deep neural network and ARIMA: implications for policy evaluations

Once fertility falls to a low level, the number of births declines, affecting the future of population growth and age structure. In low-fertility settings where sex preference is culturally rooted in society, the sex ratio at birth is usually higher than the normal average leading to an imbalanced age structure in the long run. Low fertility and its negative consequences have led to the implementation of pronatalist programs aimed at increasing fertility rates in Iran. In this context, the number of births and sex ratio at birth are matters of concern for policymakers. The main objective of this paper is to forecast the trends of the total number of births by gender and predict the sex ratio at birth (SRB) in Iran over 10 years (2021–2030) using two modeling approaches: Deep Neural Networks-DNNs and Autoregressive Integrated Moving Average—ARIMA. The results are compared to examine the performance of these forecasting methods. The findings from both DNN and ARIMA approaches suggest a 20.6% and 3.5% reduction in the number of births, respectively, and a changing trend within the normal range for sex ratios at birth. The results show the superiority of DNN model as compared with ARIMA for predictions. We recommend the utilization of the DNN approach and its derivations to visualize the outcomes of population policies based on accurate and long-term predictions. This approach can serve as an initial validation of policy impacts to enhance policymakers’ confidence in their proposed programs.