Stochastic Fertility Forecasting

fertility-forecastingstochastic-forecastingARMATFRF-starstructural-uncertaintyBayesianvalidationlong-run-averageexpert-judgmentARIMA

Definition

Stochastic fertility forecasting uses time-series models to generate probability distributions over future trajectories of the total fertility rate (TFR) and age-specific fertility rates, replacing scenario-based point forecasts with prediction intervals that explicitly quantify forecast uncertainty. Two distinct approaches exist: Thompson et al.'s (1989) parameterized schedule approach, which models the full age distribution of fertility through a gamma curve with three demographically interpretable parameters (TFR, mean age at childbearing [MACB], and standard deviation of age at childbearing [SDACB]) estimated jointly via a multivariate autoregressive integrated moving average (ARIMA); and Lee's constrained autoregressive moving average (ARMA) approach, which models TFR alone as an ARMA process mean-reverting toward a long-run average FF^*, sacrificing age-schedule detail to focus entirely on long-run fertility-level uncertainty. The central methodological challenge in both approaches is that the long-run average TFR operates on 30305050 year generational timescales that cannot be reliably identified from available historical data, creating irreducible structural uncertainty on top of the statistical parameter uncertainty the ARIMA/ARMA model does estimate.

Key Ideas

How It Works

Model Estimation

  1. Fix FF^* (from expert judgment or priors).
  2. Detrend the TFR series: (FtF)(F_t - F^*).
  3. Estimate ARMA(1,1)\text{ARMA}(1,1) coefficients aa and cc by maximum likelihood over the base period.
  4. Estimate σu2\sigma_u^2, the innovation variance.
  5. Simulate forward from the launch date to generate the predictive distribution.

Prediction Interval Width

The variance of the h-step-ahead forecast grows as a function of aa, cc, σu2\sigma_u^2, and (FFˉ)2(F^* - \bar{F})^2. Because a1a \approx 1 (near unit root), variance grows rapidly in the first 18\approx 18 years then asymptotes. The asymptotic width is approximately:

Asymptotic width2×1.96×σu11a2(plus F contribution)\text{Asymptotic width} \approx 2 \times 1.96 \times \sigma_u \cdot \frac{1}{\sqrt{1-a^2}} \quad (\text{plus } F^* \text{ contribution})

Bayesian Prior-Weighting

For a discrete set of FF^* values {f1,f2,}\{f_1, f_2, \ldots\} with prior weights {p1,p2,}\{p_1, p_2, \ldots\}, the overall forecast is:

P(TFRx at horizon h)=ipiP(TFRxF=fi)P(\text{TFR} \leq x \text{ at horizon } h) = \sum_i p_i \cdot P(\text{TFR} \leq x \mid F^* = f_i)

This mixes conditional predictive distributions rather than fitting a single model.

Why It Matters

Open Questions

Related

Sources