Combining registration-system and survey data to estimate birth probabilities.

Area of Science:

• Demography
• Statistical Modeling
• Epidemiology

Background:

• Event-history data, particularly survey data, is increasingly preferred over registration-system data in demography.
• This shift overlooks the potential benefits of integrating both data sources for more robust demographic analyses.
• Existing demographic hazard models often rely on single data types, potentially limiting accuracy and precision.

Purpose of the Study:

• To propose and demonstrate a novel framework combining survey and registration-system data for demographic hazard modeling.
• To apply this framework to estimate annual birth probabilities by parity using combined panel survey and birth registration data.
• To illustrate how integrating registration data can constrain and improve the precision of parity-specific birth probability estimates.

Main Methods:

• Development of a constrained maximum-likelihood framework for demographic hazard modeling.
• Integration of panel survey data with birth registration data.
• Utilizing the general fertility rate from registration data to constrain the weighted sum of parity-specific birth probabilities.

Main Results:

• The combined data approach successfully estimated annual birth probabilities by parity.
• Registration data significantly constrained the weighted sum of parity-specific birth probabilities, aligning with the general fertility rate.
• The variances of parity-specific birth probabilities were halved when registration data was used for constraint, indicating enhanced precision.

Conclusions:

• Combining survey and registration-system data via a constrained maximum-likelihood framework offers a superior approach to demographic hazard modeling.
• This integrated method significantly improves the precision of parity-specific birth probability estimates.
• The proposed framework has broad applicability for various demographic research questions beyond fertility estimation.