Page 248 - Contributed Paper Session (CPS) - Volume 3

CPS2010 Rodrigo L. et al.
analyzing the changes in non-monitory wealth over time and across countries,
need to hold space constant. One of the biggest hurdles is the question of
whether, and to what extent, geographic boundaries change across census
years. Until now, little has been done to verify the spatial areas corresponding
to coded units in the census microdata. Even less has been done to research
spatial changes across time. Users of census microdata are limited by the
timing of censuses (typically every 5 or 10 years) and by the unit levels
identified in the data (typically administrative divisions within country).
However, given the rise in digital mapping capabilities and spatial analytical
technologies, the IPUMS census data collection has created integrated
geographical units at the first and second administrative level of geography.
The integrated geographical units take into consideration changing
boundaries, the temporal aspect of the data from multiple censuses, and the
scalar aspect by considering the different administrative levels of geography.
The work involves extensive metadata acquisition, research, and verification
(acquisition and correspondence); the creation of small-area building blocks
that cover consistent spatial extent over time (harmonization); the testing and
implementation of techniques to group spatial units to meet the 20,000
persons threshold (regionalization); and the development of GIS shapefiles
and variables (map and variable creation) (Sarkar et al, 2015).

3. Result
Preliminary results presented here include the calculation of a wealth index
using weights produced through PCA. We visualize our results with maps of
South America representing the three census rounds covered in the data
(1990, 2000, and 2010). We used GIS software Arc View 10.3 to map all our
results. Figure 1 displays the mean value for the asset index at the second
administrative level of geography; e.g. the maps represent municipios for
Brazil, Colombia, and Chile and provinces for Peru. The mean values are
divided into five groups following the natural breaks in the wealth index
distribution. Along with the maps, Figure 1 includes graphs for the national
changes in the mean asset index over time. Since space is held constant, it is
easy to analyze changes in the standard of living from one census year to
another for the different countries. Based on these maps, the asset index
shows visually an overall improvement for all countries. Colombia is the only
country where the mean asset index actually decreased over time. If we look
at the changes at the municipio level in Figure 1, the decrease seems to be
restricted to the sparsely populated eastern part of the country. The densely
populated Bogota area and the western parts of the country show an
improvement in both indices over the years. Our final paper will also analyze
the median asset index along with the mean values to adjust for such
inconsistencies.

237 | I S I W S C 2 0 1 9