Integration of a National GIS
Project Within the

Planning and Implementation of a
Population Census



Eliahu Ben-Moshe
Census and Demography Department

Tel. 972-2-655 3474
FAX 972-2-655 3531
e-mail: msebm@pluto.mscc.huji.ac.il


1. General


The main objective of this paper is to present the advantages latent in integrating
the national GIS project - which was designed to supply a geographic base for
management of population statistics - within the planning and implementation of a
national population census. This will be accomplished by presenting the principal
strategies that were used by us in integrating these two projects, as well as a
description of the key phases.

In the case of Israel, integration of a national GIS and a population census enabled
the utilization of census resources to solve the problems inherent in setting up the
GIS, while at the same time, it enabled the census to take advantage of GIS
technologies to solve census problems. The integration allowed for significant
improvement of field work during the census, as it relates to covering buildings,
which improved population coverage, which is and always has been, the main
objective of the census.
A by-product of this integration is the significant saving in costs. Actually, the cost
of the GIS project was absorbed within the cost of the census budget, without
requiring an increase in funding.
On the assumption that such integration is desirable, I will attempt to show that the
earlier the project of setting up the GIS is commenced during the course of
preparations for the census, the earlier that benefits can be reaped from the
unique by-products the GIS has to offer.

The structure of this paper is as follows:
First, the paper presents a number of key problems which are present in both of
the projects and which are likely to be solved, or at least to be reduced, as a
result of integration.
Next, there is an explanation of the guiding principles in the strategy of setting up
the GIS, and the main phases of the project are described, taking into account the
schedule that was imposed and the costs involved also compared with the costs of
using alternative methods. There is an emphasis on the main advantage of
integration in terms of the GIS: the creation of a national mapping infrastructure at
the level of buildings and streets, together with a national addresses file.
Later on, there is a description of how the GIS was integrated into the different
phases of the census, and the advantages of the integration from the perspective
of the census project.
The last section is dedicated to the post-census advantages resulting from the
integration of the two projects, and several challenges for the future are
identified.

2. Central Problems Related to Both Projects


2.1 Difficulties of the GIS Project
There are two main difficulties in establishing a national (or local) GIS:
The problem of digitization
The problem of entering addresses

Anyone who deals with, or has had experience dealing with, setting up a
Geographical Information System as the infrastructure for management of population
statistics at the national level, or even the local level, would not be surprised to
learn that one of the main problems, which is a bothersome bottle-neck, is the
problem of entering the basic geographical data infrastructure in the computerized
system or, in other words, the digitization of the maps at the level of buildings and
roads. Anyone who has attempted to apply the wonders of technological development
related to geographical information systems has quickly found out that without
massive input of basic geographical data, these wonders are unattainable.
Another critical problem is that of entering the addresses, matching names and
symbols of streets and house numbers with buildings on a digital map. In most
cases, any statistical data that may be collected does not allow for geographical
identification at the coordinate level. Detailed population data may include, under the
best cases, geographical identification of the address. That is also the case
regarding data which come from such sources as property tax files, electric
company and telephone company records, etc. But once the address data is located
in a computerized system (GIS) and each building "knows" its address (or its
addresses, in the many cases when a building has more than one possible address),
any piece of external information with a defined geographical dimension can be
added to the system by linking it to the address.

I do not claim that these are the only problems that have to be faced, but they are
certainly on the critical path of the GIS system set up and are threshold problems
that need the main resources of the set-up project.

2.2 Difficulties with the Census Project (at the geographic level)
One of the main difficulties, and one of the early problems faced by planners of a
national population census, is the problem of complete mapping of all residential
buildings in the country. Identification of all buildings which may be inhabited by
people is one of the most important and most difficult tasks during the preparation
and implementation phases of the census, regardless of the actual census method
employed. If the census method is based on the drop-off and pick-up of
questionnaires by enumerators, the problem of covering the buildings is compounded
by the problem of distributing homogeneous work loads for the enumerators.
This requirement is, to a large extent, paradoxical: The planners need to pre-
estimate the anticipated results of the census by estimating the expected number of
households in each area at a relatively high level of geographical detail, so that the
enumerator's work load will be manageable (In the Israeli last census, an enumerator
could accomplish the enumeration of 275 75 households during the period of time
allocated for that). In the past, as in many other countries, before the population
census began in Israel, "a census of buildings" was conducted which served, inter
alia, to update the maps to be given to the enumerators working on the population
census. This update included not only a correction of the number of buildings
according to findings in the field, but also a listing of the addresses which helped
enumerators to orient themselves in the field during their work. This action was
very expensive because it required a great deal of particularly skilled staff and
because it lasted for a relatively long period of time.

2.3 Integration as a Solution
Since there is a clear connection between the structured problems in the two
projects, the census planners faced the question of how to come up with a combined
solution which would serve both of the projects equally well.

3. Guiding Principles in the Strategy of
Integrating the GIS in the Census


There is a definite asymmetry between the national GIS project and the population
census project. In essence, setting up the GIS can only be a relatively small part,
from all perspectives, of the planning and implementation process of the population
census. Therefore, if the subject is integration, the intention is to integrate the
setting up of the GIS project within the census project, and not the other way
around. This is not intended to belittle the size and the importance of the GIS tasks,
but these are dwarfed in comparison with the overall census tasks.
In light of this, it was decided form the beginning that the census project would be
the lead project, while setting up the GIS would be a by-product to make the mapping
procedures for the census more efficient. This strategic decision guided those
planning and implementing the project throughout the course of their work, and
enabled them to make quick decisions in the event of a conflict of interest between
the two projects.
Two main principles guided us during the planning and computerized mapping for the
census (which is almost synonymous with the GIS, thereby emphasizing the principle
noted above):
Parsimony in the requirements for accuracy and the scope of information collected
for the purpose of mapping; and
maximum standardization of the work processes and the results (the maps).
One of the team members used to remind us throughout the project that demands
need to be reduced because "any excess weight could cause the entire project to
sink." Any attempt to put more items on the maps, or to require detailed levels of
engineering accuracy, was rejected out of hand in order to reduce the project to
reasonable dimensions which would allow its implementation.
Even in a country as small as Israel, with less than 6 million residents living in
about 600,000 buildings out of a total of 700,000 buildings, a complete mapping
project of all localities is no simple task.

4. The Setting Up Process


The process of building the GIS while also integrating it within the planning and
implementation of the census processes involved a number of key stages:

4.1 Overall planning of the set-up stages and method of integration within
the census project
This stage included research into how the GIS could be utilized by the census, both
in Israel and around the world, formulation of a conception and determination of
phases of work, along with a series of experiments. The experiments were divided
into:
'Laboratory experiments' primarily designed to examine the feasibility of
various components of the planning. It should be borne in mind that in the
early planning stages, the relevant technology was still in its infancy.
Field experiments, among which were two census-related experiments: an
experiment conducted in 1991 on a locality with 20,000 residents, and the
general rehearsal for the census held in 1993 (with 60,000 residents). It
should be noted that even in the early stages, fruitful cooperation was
begun with Survey of Israel The national agency for mapping), which
continues to this day.
At the same time, during this early stage there was also a fundamental examination
of the existing inventory of maps of all types, in all the relevant media (digital,
paper, aerial photographs).

4.2 Digitization of maps (streets and buildings)
Digitization of maps was carried out by private digitization companies, and therefore,
this stage included planning and drafting of implementation requirements, in close
cooperation with the Survey of Israel.
Digitization was primarily done on the basis of current aerial photographs. In those
cases where digitized maps already existed, these were used if they could be made
to meet the requirements. Similarly, paper maps were also digitized.
Implementation requirements were that the digital maps were to be "mute", that is, they
would include streets and buildings, but no text, only a limited number of
characterizations (categories of streets and roads, type of building, etc.). An
important step was planning and implementing the procedures for receiving, checking
and entering the digital maps into the CBS computers.

4.3 Adding names to streets and adding addresses to buildings (street and
house number)
One of the best and most efficient sources for addresses can be found in the
Population Register. For the purpose of adding addresses to maps, a special
application was developed which made use of the addresses file within the Population
Register. This application, pioneering in its time, reduced the amount of errors
during the process, which was called "address ascription - application." During this
process:
The addresses were ascribed to building shapes and street lines, thereby
becoming an integral part of the geographical database; and
those addresses were then applied to their proper place on the map.
This method saved considerable work during the map editing stage prior to the maps
being printed out.
4.4 Adding administrative borders and dividing into enumeration areas
(redistricting)
For the division into enumeration areas, a special system of redistricting was
developed which utilized the problematic and not always up-to-date information from
the National Population Register concerning the size of the population in the polling
district and by address. Additional data included population estimates in statistical
areas (holding an average of about 1,200 households) which underwent automatic
corrections (standardization). The redistricting was performed in a computer-
assisted procedure. In order to divide a locality area into EAs, the area divider
received information regarding the standardized population in each polling district
within a given statistical area (polling districts are included in their entirety within
statistical areas). Division into polling districts was used as a system default; in
other words, a polling district was defined as an EA if it met the criteria of
population size. This application enabled arranging polling districts to the desired
size of about 275 households in each EA, creating additional EAs and the uniting
of existing areas, by changing the borders of the polling district. This method was
used for most of the EAs, except in those cases where prior information indicated
that the data from the Population Register was very out of date. In such cases,
division was done using the same system, but without relying on the Register data.
The result of this process was impressively accurate: Of the approximately 6,000
EAs enumerated during the census, in only about 130 cases (some 2%) were
particularly large EAs discovered, forcing field workers to split them up. Also, the
number of small EAs was less when compared to the previous census.

4.5 Generating outputs prior to field work
During the field work stage, the census project benefited from two main GIS
products:
printed maps;
addresses files.
About 6,000 maps of EAs were produced through a process which was mostly
automated. Furthermore, the system supplied maps for the higher echelons: section
maps for supervisors, and locality maps and other types of maps for the more
senior management echelons. The maps, which were totally consistent with each
other (complete matching of borders), were of a much higher quality than those we
were familiar with from previous census.
The addresses files were used during the process of controlling the enumerators'
work, in order to improve coverage of the buildings. The files were integrated into
the computerized field work system, so that each supervisor could question the
enumerators under his responsibility in cases where there was a lack of
information on any building found on the EA map.

4.6 Supporting and improving the geographic coding
The addresses file was also used as a background file for the geographic coding
which was performed during the questionnaire data entry process utilizing the
census data optical data entry system. The addresses files are what supported the
automatic coding system, during which time the addresses reported on the
questionnaire were linked with GIS addresses file. Cases where links were not
made were identified and sent for coding by the geographic experts using a
computer-assisted procedure.
4.7 Post-census updating of the GIS
With the completion of the field work, a great deal of information that had been
collected by the enumerators during the census enumeration had been stored. This
data included addresses that were not in our possession at the time the maps were
prepared, as well as many corrections and revisions, which enabled updating of the
maps as of the day of the census. This activity is now underway.

4.8 Matching updated GIS data with the census questionnaires
Another stage in the census process is the matching of addresses from the census
questionnaires with the addresses in the GIS, in order to provide coordinates and
GIS-type addresses for each household and individual in the census population
according to the building coordinates found on the GIS maps. Also during this
stage, the final linkage between the building records within the GIS and the data
from the questionnaires is performed. The result of this linkage is that each
respondent is linked to coordinates and through them, to all the geographical
information which is included, or will ever be included, in the database of the
census GIS. At the same time, each building in the GIS is linked to a population
along with its census characteristics. This stage will begin as soon as the
questionnaires file will be ready.

4.9 Producing cartographic products based on census data
With publication of the provisional census data in March 1996 (the census took
place in November 1995), maps were produced for the large localities. Today, the
process of producing other cartographic products is in progress and many maps,
which will join the various publications of the census, are currently being prepared.

4.10 Integrating the GIS into the Bureau's ongoing work
This last stage is expected to begin when the activities described above have been
completed, and will enter full gear in 1998.

5. Schedule (and Costs)


The stages involved in setting up the GIS were integrated into the census schedule.
Below is a schedule of the principal phases of the work. Naturally, with different
allocations of resources, some of the stages could have taken more or less time.
However, there is an internal dependence between the activities. For example, a
delay in the census by one year made it necessary to repeat the update on the
original maps, which required allocation of special resources.
A large portion of the activities took longer than planned. The main reason for
this was that people did not have prior experience in many of the necessary
activities, particularly considering the scope of such a large undertaking (not even
in the private professional sector).

Stage in setting up GIS
 Begin stage

End stage

Personnel
involved
 Duration of
stage (in
months)
1. Comprehensive
planning of set-up stages
and integration in the
census project, including
preparations, experiments,
formulating the content of
the GIS, and receipt and
adaptation of equipment
End of
1990
 End of
1992
5
24
2. Digitization of street
and building maps,
including drafting of
tender, receipt, and
examination of digital maps
and absorption into
system
Beginning of
1993
 Mid-1994
5
18
3. Adding of names to
streets and full addresses
to buildings (street and
house number)
Beginning of
1993
 July 1994
10
12
4. Adding administrative
borders and division into
enumeration areas
August 1994
 January
1995
12
6
5. Producing maps and
databases prior to going
out into the field
May 1995
 September
1995
10
6
6. Support and
improvement of the
geographic coding
January
1996
 August
1996
5
8
7. Post-census updating
of the GIS
January
1996
 April 1997
10
16
8. Matching updated GIS
data with the census
questionnaires
April 1997
 May 1997
5
2
9. Producing
cartographic products
based on the census data
March 1996
 Continuing
3
Continuing
Integrating the GIS into the
ongoing work of the
Bureau
Mid-1997
 Ongoing
10
Ongoing


It is difficult to relate here to the costs of setting up the GIS in absolute terms, so
as to make them meaningful in international comparison. We can say, however, that
the cost of setting up the GIS did not exceed the costs of the alternatives that would
have been required to map the residential areas prior to the census. In total, the
cost of setting up the GIS was absorbed within the census budget, without requiring
any special additional funding.

6. Advantages for the Census of Integrating
the GIS Project


One of the important characteristics of the project described here is that the GIS
was integrated within the census project from the initial planning stages of the
census (inclusive). This integration enabled the planners and implementers to gain
the maximum benefit from the investment in setting up the GIS. Actually, there wasn't
a single census phase that did not benefit from the GIS integration. I will point out
the main advantages:

6.1 Planning work procedures on the basis of standardized maps
Planning an enumerator's work on the basis of a map that has been designed in
precisely the same manner for each of the EAs, is an advantage which was possible
following production of maps from the GIS. Furthermore, the fact that the planners
could determine with certainty the contents of the maps and know that each
enumerator had in his possession exactly the same type of map, made the planning
easier. It is difficult to estimate the contribution of this type of standardization.
But after a peek at the maps used in previous census polls, which were a function
of the various types of sources used to prepare them, it can safely be argued
that in previous censuses, the quality of the maps used in some cases did not allow
the enumerator to fulfill his role as planned. This situation was prevented from
occurring in the 1995 census thanks to the contribution of the GIS.

6.2 Establishing an organizational structure for the census and distributing
uniform work packets to the enumerators
Improving the process of distributing work tasks that could be implemented at each
field level was another contribution of the GIS system. In comparison with previous
census polls, the new method described above enabled a significant savings in the
size of the infra structure which served the census field setting. Furthermore, the
very fact that almost all the enumerators were allocated reasonable work packets,
not too big and not too small, contributed to improving the work process, and in the
final analysis, did much towards improving the quality of the results.

6.3 Supervising the work of the enumerators
This was one of the main contributions of the system. The match between the
enumerators' maps and the addresses file located in the computers in the sub-
regional offices enabled efficient supervision of the enumerators' work, primarily
as it relates to covering buildings. An enumerator was required, through the use of
computer reports, to supply the supervisor with an explanation concerning each
building found on the map, either using the questionnaires from the respondents, if
the building is for residential use, or by some other reasonable explanation
concerning other uses of the building. This process not only enabled monitoring of
the enumerator's work, but also made an important contribution towards freeing the
supervisors for additional duties, which enabled additional improvement in the quality
of the data collected.

6.4 Contribution of the addresses file to the geographic coding
The national addresses files enabled automatic coding of most residences, places of
work and places of residence from five years ago that were reported in the
questionnaires.
6.5 Producing new census products
The products that have already been produced, and which are still being planned, as
a result of the census, exceed what we have known to date, both in terms of their
quality and their variety. Without examining all the advantages of producing census
products on the basis of the GIS, I would mention only the flexibility in defining the
new geographical units. In reality, the moment the data from the questionnaires is
linked to the buildings in the GIS, we can retrieve data on new geographical units
from the resolution of a digital map (from the building level and beyond). In
practical terms, a client can request data on any geographical unit he defines (of
course, subject to the limits resulting from the need to protect the confidentiality of
individual data).

We can summarize by pointing out that the contribution of the GIS focuses on
improving the ability to cover the entire population and raising the quality of the
geographical data supplied by the census. It should be borne in mind that these two
items were the key objectives that were set during the planning stage for the 1995
Census of Population and Housing.

7. Post-Census Advantages and Challenges
for the Future


The contribution of establishing a national GIS for the management of population
statistics in the context of the census is, indeed, a large part of the contribution
which the project made to the organization, but is not the entire contribution. In
essence, without the need for additional investment, the CBS has become, with the
end of the census activities, the owners of a national GIS which can serve the
Bureau in the future.
It can be argued that even if the census had not benefited from the GIS, the mere
fact that at the end of the census a computerized geographical system with the
characteristics described above was created, justifies the integration between the
GIS and the census project. But this is not the case. The contribution by the GIS
to the future management of official population statistics can not be exaggerated; its
contribution will not only be felt on population statistics, but that is beyond the scope
of this paper.

The challenges for the future are not insignificant:
The first challenge lies in the establishment of an updating process which will
enable us to maintain good matching between the contents of the GIS and the situation
in the field at any given point in time. A second challenge is matching the GIS to the
needs of the other CBS units. We are only beginning to face these two challenges.

The third challenge is to help the organization provide GIS-based services to an
ever-growing population of data consumers. This challenge has already been
seriously considered in our plans for distributing the census data. From the
experience we accumulate, we hope to learn ways to market the non-census
geographical statistical data.

As a result of the GIS the statisticians may face new professional challenges. On
the one hand, we will have to develop tools which will enable geographical sampling
based on the GIS. It may be correct to say that in terms of the sampling approach
this is not an earth-shattering innovation, but without the ability to take population
samples from maps, the existing sampling methods work mainly on the basis of
various lists from which the samples are taken. Addition of the spatial component is
likely to lead to interesting developments. A second area is the integration of the
spatial component in statistical analysis. This does not mean displaying statistical
analysis on maps, but this relates to what is referred to as "geo-statistics": fully
incorporating the spatial component within the analysis. Preliminary steps have been
taken around the world in this field, and Israeli researchers have played an
important role in this. But there is still much work to be done, and we have only
just begun.



Copyright © 1997-1999 The State of Israel. All rights reserved.
See "Terms of Use" for the conditions under which this service may be used.