Census detail as a proxy for state capacity

1
Census detail as a proxy for state capacity
Noah Schwartz, Deborah Balk (P.I.), Melanie
Brickman, Bridget Anderson, Marc Levy Center for
International Earth Science Information Network
(CIESIN), Columbia University
Introduction Governments perform a variety of
functions, and thus government capacity is
multifaceted and somewhat difficult to measure.
Indirect measures of a governments ability to
provide physical security, foster economic
activity, secure the trust of its citizens, and
fulfill various other responsibilities have been
developed. However, researchers possess few if
any proxies for assessing a governments overall
capacity. This study was designed to investigate
whether the spatial and temporal detail of
censuses can be used as a proxy for government
capacity.

• Methods
• Spatial Resolution was regressed on Governmental
  Effectiveness and, separately, on Legitimacy.
• Filters were developed to control for extreme
  areas, populations, and population densities.

Even with these controls in place, the linear
relationship between spatial resolution and
government capacity is not strong (though it is
present). Ten percent of the variation in
Resolution can be explained by variation in
Legitimacy.
Temporal consistency is a stronger predictor of
state capacity than spatial resolution.
Hypothesis The spatial and temporal detail of
censuses increases with increased capacity.
Fig. 3 Resolution improves slightly with
increased legitimacy. Countries with small areas,
small populations, and large uninhabited areas
were excluded.
Fig. 1 All small-area states (hollow circles),
including those with low Governmental
Effectiveness, have resolution finer than 100 km.
All sparsely populated countries (orange),
including those with high GovEff, lie above on
the coarser side of the fit line. Sparsely
populated countries are defined as those in which
more than 75 of the total area contains less
than 5 people per square kilometer.
Nevertheless, mean difference tests consistently
indicate that resolution increases moderately
with increasing governmental effectiveness (Fig.
5). Associations between state capacity and
temporal resolution were explored using similar
methods to those described above for spatial
resolution. Results Spatial Resolution is a
moderately good indicator of Governmental
Effectiveness. The association is stronger for
more effective states.
Data Spatial data for 232 countries was obtained
from CIESINs Gridded Population of the World
(GPW) dataset, which represents the highest
resolution, affordably available sub-national
census units. Countries compile census data for
administrative units of varying sizes. The
smaller the units, the more detailed the census.
The spatial resolution of a census can be
measured by calculating the average size of these
units This would be the width of each unit if
all the units were square and equal in size. Data
on the temporal resolution, i.e. frequency, of
censuses from 1940 to 2004 were acquired from the
International Programs Office (IPC) of the US
Census Bureau. Many measures of government
capacity exist, pertaining to different
dimensions of governance. Forty-two indices of
governmental effectiveness and legitimacy in
political, economic, social, and security matters
were obtained from the World Bank Group, CIESIN,
and other data sources. This study focused on
focused on measures of effectiveness and
legitimacy.
In small-area countries, the census data gathered
by governments of all capacity levels have fine
resolution, simply because the territory is small
(Fig. 1). Conversely, in sparsely populated
territories, governments of all capacity levels
tend to gather coarse data. Both types of
countries could cloud a relationship between
resolution and capacity that exists in other
states. For this reason, small countries and
sparsely populated countries were excluded from
analyses or labeled distinctly.
Figs. 5 and 6 Census frequency explains about
30 of the variability in both Governmental
Effectiveness and Legitimacy.
Implications The bivariate analyses described
above have laid the ground work for the
construction of a multivariate model that may
enable researchers to infer government capacity
from simple measures of census detail.
Fig. 4 As Governmental Effectiveness increases,
the median Resolution and inter-quartile range
decrease, i.e. Resolution becomes finer and less
variable. Outliers are mainly countries with
large, sparsely-populated deserts
References Gridded Population of the World,
version 3 (GPWv3), http//sedac.ciesin.columbia.ed
u/gpw. Strategic Warning for Fragile States A
Collection of Quantitative Indicators, CIESIN.
Fig. 5 Points represent the mean resolution for
each category of Governmental Effectiveness.
Error bars represent 95 confidence intervals for
the means.
Acknowledgements Many thanks to Deborah Balk for
guidance on every aspect on this project Melanie
Brickman for constructing the dataset Bridget
Anderson and Marc Levy for help with interpreting
governance indicators and Dallas Abbott for
organizing the Earth Intern program that funded
this research.
Fig. 2 As area decreases, the median and range of
resolution decreases. All small-area states have
fine resolution, confirming Fig. 1. For states
smaller than 50,000 km2, resolution is strongly
determined by area, thereby obscuring
associations with capacity.