The link between climate change, food security and fertility: The case of Bangladesh

of GDP per capita on the TFR is not statistically significant. In summary, Fig 7 suggests that although economic development indicators such as crop production and infant mortality are critical in shaping fertility transition in Bangladesh, the maximum temperature and rainfall have a direct impact on the TFR and maximum temperature also indirectly affects crop pro- duction, while crop production has a direct and an indirect impact on the TFR through the infant mortality rate. The direct, indirect, and total effects of all factors are shown in Table 1. The total effect of maximum temperature on TFR is significantly negative, mainly due to the direct negative effect. Regarding maximum rainfall, despite its significant direct positive impact on TFR, the total impact is insignificant. Meanwhile, the mean temperature and rainfall have no significant direct, indirect, or total effect on TFR. The total effect of crop production is negative but insig- nificant, and this is because the negative indirect effect offsets the positive direct effect. The direct impact of infant mortality on TFR is significantly positive. Lastly, the GDP per capita has a significant negative indirect effect through the infant mortality rate, thus leading to a sig- nificant negative total effect on TFR. Short-term and long-term impacts of climate change on TFR Furthermore, we perform ECM modelling to estimate the short-term and long-term effects of climatic factors. Table 2 shows the result of the unit root test. As shown, not all variables are

stationary in their log level (e.g., GDP per capita), while they are all stationary at 5% when using the first-differenced data. Therefore, the undifferentiated variables are used for the coin- tegration test, and the first-differenced variable was used for the ECM model. Because the crop production and infant mortality are endogenous, we do not include them in the ECM model. Four cointegration vectors are preferred by Johansen’s test for the variable sets "ln(TFR), ln (maxtemp), ln(meantemp), ln(maxrain), ln(meanrain), ln(GDPPC)". Table 3 shows the results from ECM analysis. The ECM model has two parts of the results. One part is about the short-term impact, and the other part–the error correction term–is about the long-term impact in an equilibrium situation. It should be noted that when inter- preting the error correction term, all coefficients should be inverted. In model 1, the value of λ is -0.007. Statistically, it means that once the vector of TFR, GDP per capita, and the four climate factors deviate from the long-term equilibrium path, it will converge to equilibrium at a rate of 0.007. In the short term, the change in maximum Table 1. Direct, indirect, and total effects of selected factors on TFR. Variables Direct effect Indirect effect Total effect Maximum temperature -0.045� -0.387 -0.432� Mean temperature 0.056 0.086 0.142 Maximum rainfall 0.003�� 0.004 0.007 Mean rainfall -0.050 -0.187 -0.237 Crop production 0.042��� -0.074�� -0.032 Log(GDP per capita) -0.144 -2.368��� -2.512�� Infant mortality 0.052��� / 0.052���

Fertilizer / -0.008 -0.008��� Note � p<0.05 �� p<0.01 ��� p<0.001. https://doi.org/10.1371/journal.pone.0258196.t001 PLOS ONE Climate change, food security and fertility PLOS ONE | https://doi.org/10.1371/journal.pone.0258196 October 21, 2021 10 / 18