budgeting, scheduling, design of the questionnaires, and the number and types of personnel needed to edit and tabulate the results.

If electrical accounting machines (EAM) are used, most of the editing will be accomplished manually, except for some "zero balancing" or "control checks." Data would be transcribed from the questionnaires to punch cards, which would be sorted and tabulated. Clerks would post the applicable tabulated cells to worksheets, which would then be typed and printed.

If electronic data processing (EDP) equipment is used, most of the editing would be done by computer. Data entry would be accomplished by punching cards and reading these cards into the computer. Or it may be possible to transfer the card images to magnetic tape or to enter the data directly to magnetic disk. If key-to-disk equipment is used, it would greatly reduce the number of punch cards required for data processing. In addition to editing, the computer can handle most of the control checks and can print the final tables in copy ready for printing.

The Providencia NSO had made a tentative decision to use the computer if possible. Consequently, the organization of their staff reflected this decision.

4. RECEIPT AND SORTING OF
COMPLETED QUESTIONNAIRES

The questionnaires completed in the field are sent to the Field Office Division (FOD) with two copies of Form 123 (exhibit II-6-1). The FOD Receiving Clerk verifies that all the questionnaires are received. One copy of Form 123 is retained in FOD and the other is sent with the questionnaires to the Control Operations Branch of DPD.

The DPD Control Clerk enters the identification codes (province, county, ward, and ED) in the upper right hand corner of each questionnaire. The long forms are then reviewed to determine whether the proper form was used by checking both of the following:

5. TESTING PROCEDURES

Testing the procedures is a most important step in the development of an undertaking as large and complex as a census of a country. Each step in the data processing operation should be tested, revised, and re-tested; then a final test should be made on the operation as a whole. Operations that do not run smoothly can completely upset the time schedule, can cause deficits in the budget, and can adversely affect the quality of the data results. Το wait until "live" data are available from the census itself is too late; revisions are usually too complex to be made quickly.

In the first stages of a test, the professional staff should try out the procedure, discuss difficulties, and revise as needed. In the later and final stages, the persons who are testing the procedures should include or be of

the same level as those who will be carrying out the procedure--clerks should try out the manual editing-coding procedure, key punch operators should test the punching instructions, control clerks and messengers should test the control and transmittal procedures, etc. It is desirable that potential supervisors should first serve as clerks, editor-coders, key punch operators, or verifiers. This will help supervisors to analyze problems in the management of the various activities and to check their own written procedures.

For the early testing, data from field trials and pre-tests can be used. Or test data can be compiled by creating several hundred establishments, filling actual census questionnaires for them. Care should be taken to create establishments of various types so that the procedures for unusual as well as usual cases can be tested. Some inconsistent and missing information should be built into the test. Transmittals and control procedures (and forms) should be tested also.

Time and production records should be kept for those steps in which time and production are important elements. When a procedure requires a relatively large amount of time, the staff should determine whether certain costly steps are essential to the success of the program, or whether a more economical but satisfactory alternative could be used.

5.1 Manual processing

Editor-coders should use the test data (described above) to try out the instructions for editing-coding and the procedures for complete verification and sample verification (if used). The subject matter specialists should review the work and ascertain the reasons for errors through discussions with the editorcoders. If "live" data are used in the test, the subject matter specialists should review

the number and kinds of entries that are changed or imputed. The time and production records should be examined to identify steps that are too complicated or costly. It is possible that more of the manual editingcoding can be accomplished more economically by the computer.

5.2 Key punching

Testing the key punch operation involves many of the same steps as testing the manual processing. Error reports by questionnaire item should be kept and analyzed to determine whether the instructions are unclear or whether the design of the form is faulty.

5.3 Machine processing

Most programmers will create their own test data to initially test their programs. The test data used in the manual processing and key punching should also be used. The computer programmers, systems analysts, and subject matter specialists all should be involved in the testing. Quite often situations will arise in the test data that are not programmed for or stated in the specifications.

The subject matter specialists should manually edit and tabulate the test cases, following the editing and tabulating specifications they prepared. The manual results should be compared with the edited and tabulated results produced by the computer, and differences should be resolved.

This can be

a fairly tedious and time-consuming activity, but it is well worth the effort. The type of editing changes that are made should also be reviewed for reasonableness.

Testing should cover the procedural instructions for computer control and operations personnel. Included would be such phases as correcting records, allowing for recovery of data, setting up work units, etc.