DEMOGRAPHY© Volume 20, Number 3 August 1983
POPULATION ESTIMATES FROM LONGITUDINAL RECORDS
IN OTHERWISE DATA-DEFICIENT SETTINGS
Douglas L. Anderton
Department of Sociology, The University of Chicago, Chicago, Illinois 60637
Joseph Conaty
Department of Sociology, The University of Utah, Salt Lake City, Utah 84112
Thomas W. Pullum
Center for Studies in Demography and Ecology, University of Washington, Seattle, Washington
98195
Abstract-Existing methods for estimating population parameters in settings of
data deficiency do not provide techniques for analysis of commonly
available longitudinal data. In settings where complete population data is
unavailable, longitudinal data recorded for only a subset of the total
population are often available (e.g., event registers, genealogies). In this
article we present and evaluate models which derive population parame-
ters for the population subgroup underlying such longitudinal data. Using
the distribution of individual times until first recorded event within a
measurement interval, population parameters are estimated which provide
basic denominator data for analyzing event occurrence. The models which
we derive are especially suited to records which may include migration and
population growth trends. The use of the models is demonstrated and
evaluated through an application to genealogical records for a nineteenth-
century population. Possible extensions of these models and their major
limitations are also discussed.
Established methods for estimating
population parameters rely upon specific
demographic records and do not fully
exploit other, often more commonly
available, data sources for human popu-
lations. Population estimation tech-
niques designed for data-deficient set-
tings are based upon survivorship
calculations and require the use of cross-
sectional population enumerations in ad-
dition to longitudinal vital event registra-
tions or model life tables (Brass and
Coale, 1968; Coale and Zelnik, 1963;
Preston and Hill, 1980). The existing
methods which use only longitudinal reg-
istration data assume registration com-
pleteness and population closure (Lee,
1977). In addition, these methods esti-
mate only total population parameters
rather than population subgroups and do
not provide basic denominator data for
analyzing recorded events. An attractive
alternative to these methods would esti-
mate open population parameters for un-
derlying subgroups solely from longitudi-
nal records in data-deficient settings. In
addition to the demographic registrations
used by the methods above, there are
numerous records which are left behind
during the individual's life cycle. Thus,
within the context of this paper, data-
deficient refers not to a lack of recorded
information on individuals but to the lack
of reliable cross-sectional enumerations
for subgroups of interest. In this sense,
many commonplace demographic set-
tings are data-deficient. For example,
vital statistics registration records may
be used to estimate fecund women in a
population while no cross-sectional enu-
273