• One of the earliest recorded censuses that we still have access to is the Domesday Book, which dates to the eleventh century. It was recorded on vellum and bound into book form. A 1986 effort by the BBC to replicate its scale in 1986 was recorded on laser discs, which have already become much harder to read than the much older books (p. 50).
  • The longest recorded genealogy is that of Confucius and his descendants. The “main line” can be traced for 80 generations and another chain can be traced for 86 generations (p. 53).
  • The antecedent of modern statistics was known as “political arithmetic,” which consisted of applying shopkeepers’ mathematics to demography (p. 61ff). One of its early successes was estimating the population of London using bills of mortality to extrapolate the number of living citizens based on the number of adults who died each year (p. 63).
  • When Scotland joined with England to form Grate Britain in 1707 it gained 61 Ministers of Parliament. Despite having a roughly equivalent population by the mid-1700’s the US colonies had no such representation, which was a leading grievance in the eventual revolution (p. 71).
  • Early censuses in North America and Europe made it difficult to cross-tabulate data because of the way questions were asked. Reporters took the count of how many people in each househould had various characteristics, such as being employed and being literate, but it was difficult to estimate had multiple characteristics, such as being both employed and literate. The 1840 US Census had 48 columns with convoluted combinations (p. 99). The 1850 census was revised to have one row per individual in the household, simplifying it greatly to only 13 columns (p. 102). The dimensional design made analysis much easier, and without it the later innovation by Hollerith to use punch cards and mechanical tabulation would have been unlikely. When Hollerith did introduce his system, an important feature was that the transfer of information from handwritten responses to cards “would happen only once, but counting would be repeated over and over again for all the different subgroups on which the census reported: once for men, once for women, once for those native-born, once for those foreign-born, and so on.” (p. 105)
  • One of the saddest chapters in the history of demography is that of the required registration of Jews under Nazi rule (p. 146ff.). These and similar population records were of such value that Allies targeted them for bombing. Even destroying a fraction of the records held in central repositories gave plausible deniability to resistance fighters using false papers, for example.
  • How do we know the accuracy level of a census? Modern bureaus use a technique known as “capture-recapture” estimation, which was originally developed for studying populations of wild animals who are obviously much more difficult to quanity (p. 233). It involves taking two samples from a population in short succession and estimating the total size using the amount of overlap between the samples. Today census bureaus often run a large-scale survey (although still much smaller than a complete enumeration of the population) shortly after a census to estimate under- or over-counting. The 2010 US Census was estimated to have overcounted the population by 0.01 percent, which is about as close as any group can get to perfect accuracy given the complications involved (p. 242).