1. Introduction
2. Costs of Creating and Maintaining GIS
3. Grid square database
4. Methods of allocating households or enumeration districts to a grid square
5. Advantages and disadvantages of the grid square data base approach

Cartographic presentation of small area statistics is a powerful means for visualizing the census results. For this purpose, most often, statistics of enumeration districts (ED) or census blocks are presented using the boundaries of EDs or census blocks. However, this requires digitizing the ED boundaries within a whole country which will be a very costly and time-consuming work. It would be fair to say that most developing countries cannot afford to undertake this even under the current rapid improvement of costs for acquiring digital data through such devices as GPS, digitizer and scanner.

This paper suggests that those countries, which do not have adequate amounts of resources for digitizing their ED boundaries, consider an alternative method, not a new nor a sophisticated but a less costly method, for presenting small area statistics through a grid square database approach. What is involved in this approach is (a) dividing a whole territory or a region into grid squares, each grid square being a rectangular of equal size, 100 meters (U.K.) to 1-kilometers (Japan and the Republic of Korea) and (b) allocating population data of ED to the grid squares. The paper explains how to do this allocation. Also, the paper examines the advantages and disadvantages of this approach.

Maps make it easier to present, analyze and disseminate census results. They facilitate the identification of local patterns of important demographic and social indicators. Maps are thus an integral part of policy analysis in the public and private sectors.

However, there have been many warnings and reservations for creating GIS because of costs.

• UN handbook on geographic information systems^1/ states that the initial investment and the following maintenance costs of GIS should not be underestimated. Like any new technology or organizational transformation, the introduction of GIS involves a change in routine and significant expense, not only for software and hardware but also for data purchase, training, planning and organizational restructuring.  In fact, the significant costs involved are the main reason why the section on GIS in the revised Principles and Recommendations for Population and Housing Censuses (United Nations, 1998) are worded very carefully. The indirect costs, in particular, are often underestimated and may lead to the failure of a GIS project.
• The Draft Guideline for Dissemination of this Workshop^2/ warned that "less advanced national statistical offices should not contemplate the introduction of a GIS facility before all other dissemination options have been fully investigated. The cost of implementing a GIS facility would outweigh any benefits unless a sophisticated user group exists within the country that could contribute towards the development costs associated with such a venture."
• Although digitized map of ED is the heart of constructing a GIS, digitizing the boundary of census enumeration will be a very costly undertaking, regardless of the approaches you may choose; direct data collection with GPS or geographic data conversion of hard-copy maps or aerial photographs through digitizing or scanning.

• The grid square database has been created in such countries as U.K., Japan and Republic of Korea, etc. for the purpose of presenting the data of small area.  Data from the grid square database are widely utilized, particularly for urban and rural planning and administration in national and local governments, for urban and regional studies in research institutes and for marketing in commercial fields.
• For the purpose of creating a database on the grid square basis, the whole territory of a nation or a region is subdivided into smaller areas with the same size and the same rectangular shape. The size of a grid square utilized varies among countries.  U.K. employs a size of 100 meters by 100 meters, while Japan and Korea delineate the standard Grid Square with a size of about 1000 meters by 1000 meters.
• The grid square data in U.K. were compiled by aggregating the data for all households located within each grid square, while those in Japan and Korea were compiled by aggregating the data for all enumeration districts involved within each grid square. The methodologies of the aggregation are explained below.

The most precise results will be achieved if individual households or housing units are allocated to grid cells. (Figure a) However, in order to allocate each household or housing unit, it is necessary to have a map in large scale on which the geographical location of each household is indicated precisely. Without such detailed map, it would be difficult to employ this method for creating the grid square data base, although with falling prices of GPS units one can imagine that more countries will produce such data by equipping each enumerator with a GPS during the census. However, for now, this is not a practical idea for the most of developing countries.

Allocation of enumeration districts to a grid square can have several approaches.

• One is to simply allocate enumeration districts to a grid square, if more than of its area falls into that cell (Figure b).
• On the other hand, a large enumeration district may include several much smaller grid squares (Figure c). In this case, the ED data could be assigned in total to the grid cell that contains the population centroid of the ED. The population centroid defines a representative point in the ED that should coincide with the largest population concentration in the area. Alternatively, the data can be distributed evenly across all grid cells that fall into the enumeration area.
• The centroid or representative points can also be used directly to allocate ED data to grid squares. In this method, the centroid point of each enumeration district has to be defined at the location of the center of population or at the most principal population cluster within the enumeration district. Then, data of those enumeration districts, whose centroid point falls within a grid square, are to be added to the data of that grid square. Using the digitizer or GPS the locations of the centroids of enumeration districts can be digitized and recorded in the computer-file of the enumeration districts and, then, the allocation process can be performed easily by computer.
• The centroid or the location of the central point within a census ED may be determined by finding the site of the chief settlement within the census ED on the topographical map on a large scale, such as 1/25,000.

• The grid square data can be easily aggregated into a larger area delineated arbitrarily, whereas data for villages, localities, or enumeration districts can not since they have different sizes and irregular shapes. Data for an area arbitrarily delineated can be obtained simply by adding up the data for the grid square encircled with the boundary of the area. On the other hand, in the case of using the data of villages, localities, etc. having different sizes and irregular shapes, an extensive adjustment of the aggregated data may be often needed because of the possibility of a sizable discrepancy between the aggregated area and the area actuary delineated.
• The grid square data constructed usually carry with it attributes indicating the relative location of the grid square, that is, longitude and latitude, and, therefore, a distance between two grid squares can be calculated without an actual measurement on the map.
• Such attribute of the grid square described above enable an efficient application of the computer mapping technology. For example, data for grid squares can be printed in the form of statistical map without providing additional geographical information of these areas into the computer.
• Such attributes as mentioned in (a) could also facilitate integrating population data with those from other sources which are often presented in different units of statistical area. For instance, enumeration districts might have been delineated independently among population census and other censuses such as establishment census, agriculture census, industry census and commerce census, etc. In such cases, if data of other censuses were compiled to grid squares directly or based on data for their enumeration districts, data from population census could be utilized in an integrated manner with data from other censuses on small area basis.
• The grid square data normally enable a much easier statistical comparison on the map, compared with those based on areas having different sizes and irregular shapes. Therefore, it provides an area-standardized comparison of statistical characteristics among areas concerned on the grid square map.

Against the advantages of the grid data base mentioned above, there are disadvantages, most of which arise from the characteristics of grid squares, as described below:

• A great deal of manual work may be needed for allocating each household or enumeration district to each grid square. Normally, the allocation of household or enumeration district to each grid square is made manually on the map by using a digitizer or other tools. Consequently, when the number of households or enumeration districts is large, a significant amount of manual work would be needed.
• There are often inaccuracies in the data aggregated for the grid square, particularly when an aggregation is made on the basis of the data of enumeration districts. In this case, the level of inaccuracy would rise as the size of grid square is made smaller.
• Data may not be obtained for an area perfectly identical in shape of the boundary to the area where the data are needed, because of the aggregation of rectangular shaped areas.  Therefore, in order to obtain data for the area possibly identical to the original area, the grid squares should be delineated as small as possible.
• Needless to say, data are not available for an area smaller in size than grid square.