Griffith Feeney
The objectives of this exercise are to understand the general structure of census tabulations, to learn the specific content of the census publications you are using, and to begin thinking about what tabulations should be produced in the next census. While it will probably be useful for you to collaborate on this exercise, it is also important to learn something of the variety encountered in different censuses. You are therefore responsible for cooperating with other participants to ensure that this exercise is completed for a minimum of 6 different countries.
Step 1 Select a country (noting the constraint just mentioned) and obtain relevant 1990 round census publications from the East-West Center Resource Materials Collection or from the University of Hawaii Hamilton Library. Obviously there is something to be said for selecting your own country, but there are also advantages in selecting another country. If census volumes for your country are not available you will have to select another country whose publications are available. Initially, at least, you should restrict attention to national level volumes.
Step 2 Review the census questionnaire to famliarize yourself with various items. The schedule defines and constrains what tabulations are possible. For each item, note the domain, that is, the set of persons of whom this item is asked. Marital status, for example, will be asked only of persons over some lower age limit. Note which other items define the domain (age, in the marital status example). For each item, not the range, that is, the set of values that this item may assume. For marital status, for example, the range is typically (but not always) 'Never Married', 'Married', 'Widowed' and 'Divorced'. Note which items are precoded, meaning that the ranged is determined by the census schedule, and which are not. Marital status, for example, will almost always be precoded, but occupation will usually not be. For questions involving administrative units, pay close attention to the level at which responses are requested. For place of birth, for example, how precisely is place of birth identiied? First level administrative unit? Second level administrative unit? Lower level units? Put this information into Excel in the form of a list with one entry for each questionaire item and columns for each type of information, e.g., an index number column, a questionniare number column, a domain column, and so on.?
Step 3 Review the published census tabulations informally but thoroughly, noting both what tables are included and what interesting or useful tables are not included (note how open-ended the 'not included' question is!). Compare informally the tables you find with those other participants find. Which tables (if any) are common? Which are idiosyncratic? In this scrutiny, pay attention to the precise categories used in each table. A tabulation of population by age and sex ending with the open-ended age group 65+, for example, is not the same as an otherwise similar table ending with 85+. Many census items admit of great differences in the detail which may be presented. Some tables may exhaust the detail available in the schedule for one or more of their constituent items, whereas others may present very little detail. Note the domain of each table, that is the set of persons counted in the table. How does this relate to the domain of the items tabulated? unit? Second level administrative unit? Lower level units?
Step 4 Make a list of the the variables used in the tabulations, where 'variable' is understood to mean a particular set of groups defined for a particular census item. Many census items will correspond to two or more variables. Age, for example, even if a single item on the questionnaire, will invariably give rise to at least two variables, one giving single years of age and one that collapses single year groups to quinquennial (five year) age groups. Often different open-ended age groups will be used in different tabulations. A table of population by age and sex, for example, may end with the open-ended group 85+, whereas a table of population by age and school enrollment may end with the open-ended group 35+. Use Excel to make this list, putting descriptions of variables in the third column. Then add index numbers to the first column and short codes for each variable to the second column, e.g., MAR for marital status, ASY for age in single years, and so on. Make some attempt (but don't trouble yourself unduly) to use the same code as your fellow participants for common variables, such as marital status. Use the Sort facility to sort your list by code, this as a way of ensuring that you don't use the same code for two different variables. Use Undo or Sort on the index numbers to regain the original order if you want to do so.
Step 5 Make a similar list, also in Excel, also including index numbers and short codes, for the domains represented in the various census tables.
Step 6 Referring to the variable name and domain name lists of the preceding two steps, create an analytical list of census tabulations in the form
Domain by Var1 and Var2
for example,ALL by AGE and SEX
where the ALL denotes all persons enumerated in the census, AGE age in five year groups ending with an 85+ group, and SEX is (of course) male or female. Create this list in Excel as well, with a columns for table number, domain, and one column for each variable cross-tabulated in the table.Step 7 As you scrutinize these tabulations, think of how they might be extended, improved, or pruned, and consider if you can how these tables might be produced in digital form. What are the key decisions to be made in specifying census tabulations? Who should make these decisions?
Note Distinguish the content of tabulations from their format, and focus solely on the content. Note the distinction between tabulations in which geographic/administrative areas are not a variable and tabulations in which they are a variable, and consider carefully how to handle the specification of these two types of tables.
<gfeeney@gfeeney.com>
