| Article ID: | iaor20112649 |
| Volume: | 148 |
| Issue: | 2 |
| Start Page Number: | 301 |
| End Page Number: | 317 |
| Publication Date: | Feb 2011 |
| Journal: | Journal of Optimization Theory and Applications |
| Authors: | Bonneuil Nol, Fursa Elena |
| Keywords: | missing values, population, stochastic linear programme, survey data, census |
A problem, often encountered in the analysis of historical data, is the difficulty in overcoming missing or flawed data. Lotka‐McKendrick discrete demographic model, including migration, is combined with stochastic optimization to fit available censuses and vital statistics series to reconstruct missing population data, in the presence of one or two censuses. Simulations help to calibrate the method and determine error weights associated with each data series. An empirical case study is made using data from an administrative subdivision in southern Russia for the period 1863–1916.