In any type of personal interview census

or survey, it is a good policy to perform the enumerator edit of questionnaires and to require that it be done immediately after the interview takes place. The amount and kind of subsequent field edit and preliminary office edit or screening, however, is contingent on time, personnel, budgetary considerations, nature of the survey, precision sought, and other factors.

2.4 Office edit

The next step in the editing procedure is the manual office edit. This edit should be performed under rigorous conditions with precise and detailed edit instructions.

The office edit differs basically from the enumerator edit where corrections are made while in contact with the respondent. In the office edit, decisions are generally reached' on the basis of what is assumed to be the most

probable entry. For an editor to perform a complicated office edit, he must be thoroughly familiar not only with the edit instructions but also with the questionnaires, concepts, and the enumerator's instructions.

The office

edit, too, can sometimes be minimized or even eliminated, depending on the nature of the survey and the type of data processing equipment that is available.

2.5 Mechanical edit

Basically, the mechanical edit is the same as the office edit. However, in the mechanical edit, instructions are prepared for the tabulating equipment and corrections are made on the punch card or record rather than on the questionnaire. It is a good policy to do as much of the editing mechanically as is possible. The mechanical edit assures a degree of accuracy and uniformity which is

unattainable in a manual edit. The mechanical edit takes place after the card punching operation, and thus corrects punching errors as well as the errors of the enumerator (and editor-coder, if manual editing has been done). As indicated earlier, the mechanical edit may minimize or eliminate the manual edit.

The mechanical edit using EAM equipment lends itself most readily to simple uncomplicated questions. Mechanical editing on electronic computer equipment combines all of the advantages of manual editing and of mechanical editing on EAM equipment. At the same time, it has many additional refinements. In the computer edit, the correctness of an entry can be checked by examining the consistency of numerous sets of conditions simultaneously. Further, the entry can be changed automatically without recourse to the sorting and re-punching which is involved with EAM equipment. Imputations of missing characteristics can be performed on the basis of comparable data from a previous census or sample survey.

3.2 Impossible entries

Impossible entries are those which are prohibited by the coding system. For example, if form of ownership Code 1 represents government enterprise and Code 2 represents private enterprise, any other digit would be an impossible entry.

3.3 Inconsistent entries

Inconsistent entries occur when two or more items must bear a particular relationship to each other, but do not. An example would be an entry of 10 full-time production workers with a total of only 1 day of production.

3.4 Unreasonable magnitudes

This has reference not necessarily to an error, but to an entry which appears unreasonable, such as a production worker with an income far in excess of the average for that particular industry group. Such an entry may be correct, but it deserves investigation.

4. DEVELOPMENT OF AN EDITING PROCEDURE

For

In the development of an editing procedure, numerous factors must be taken into consideration. The principal consideration is whether the questionnaires are sufficiently complete so that tabulations may be derived from them. Obviously, some items on a questionnaire are more important than others. some missing, clearly unreasonable, or inconsistent entries, it may be possible to determine--with a high degree of accuracy-the correct entry from related information on the questionnaire. In other cases, depending on the nature of the item, such imputation or assignment of an entry is not readily feasible. Thus, one of the first decisions to be made in the development of an editing procedure is to determine which items create problems of imputation. For such items, the question then

arises as to what action can be taken. Where time, funds, workload, and logistical facilities permit, returning questionnaires to the field (or contacting the respondent directly) for completion of critical omissions or incorrect entries is a possibility.

Ideal conditions for returning to the field rarely exists. And often, even if they do, the nature of the question itself may prohibit such action. For example, it may be useless to send questionnaires back to the field for completion if the questions involve serious problems of recall such as "number of days worked" during a specific period. Even if the return-to-field procedure is employed, there will always be some incomplete questionnaires. Returning questionnaires, therefore, does not solve the problem of unacceptable entries but only minimizes it.

4.1 Specialization in editing

It is well-established in manual editing that many advantages may be derived through specialization (see also section 1.2 above). Specialization is nothing more than the process of taking a large and perhaps complex entity and reducing it to simple, manageable component parts. The extreme in specialization with regard to census-type questionnaires would be to have a separate editor for each item on the document. The other extreme would be a total absence of specialization, whereby each editor would edit the complete document. Somewhere between these two extremes lies the optimum. As specialization increases, control problems increase and, if carried too far, the benefits of specialization are lost. Extreme specialization in editing also reduces the overall responsibility of an editor and thereby reduces the possibility of his recognizing documents that are basically faulty.

The judicious use of specialization, however, can result in savings in time and

money. When operations are repeated again and again, decision time is reduced to a minimum,

and the editor makes the decision with little reflection. Furthermore, editors trained in

a limited number of functions learn more readily and can be trained more quickly; the investment in editors trained in only a few functions, therefore, is relatively minor and facilitates replacement of inefficient editors.

Specialization generally results in improved quality. Fewer errors are likely to be made when the task assigned is not complex, and when the operation is done repeatedly. Moreover, it often permits utilization of special aptitudes to the fullest extent.

4.2 Evaluation of editing-coding personnel

The editing-coding procedures serve the main function of ensuring "processability" of the documents. The procedures can also be used to evaluate other aspects of the program, especially during pretesting. As already indicated, poor enumerators can often be detected through editing. However, care must be exercised to make sure that the enumerator is at fault and that the error is not due to some other cause. If relatively few enumerators make a large number of errors, it may be assumed that they are incompetent. If, on the other hand, a pattern of errors occurs on a widespread scale, it may be evidence that the fault lies with the questionnaire itself. Perhaps the wording of a question or an instruction, or the sequence of questions, is misleading or confusing.

4.3 General procedural rules

Providencia decided to combine editing and coding and to use some specialization. The following rules, which are among tested methods for promoting efficiency, were adopted.

(1)

(2)

(3)

(4)

(5)

5.

All editing-coding should be done with a sharp colored pencil or fine-point pen. In Providencia, the color chosen was red. The actual choice of color is not important, except that the contrast with the respondent's entries (handwritten or typed) should be unmistakable.

When a correction is made to an existing entry, the figure being corrected should never be erased. Instead, the editorcoder should line through the figure and enter above that figure the correct one. Blank or omitted entries should also have any figure entered with a colored pencil or pen.

Verification should be done with pencil or fine-point pen of a color other than the one used for editing-coding. Providencia selected blue as the color to be used.

In verification of editing-coding, the original entry should not be erased, but should be crossed out and the correct code entered above it. This is important for quality control purposes.

The initials of the clerk or analyst performing the editing-coding steps of a given sequence should plainly appear in the margin, followed by the date. The initials and date should appear near the section of the questionnaire being processed, rather than on a separate routing or editing worksheet which may become detached from the questionnaire.

EVALUATION OF EDITING PROCEDURES

In the conduct of an industrial census, the basic philosophy underlying the editing specifications may well determine not only the quality of the data eventually published, but also the ability of the census authorities to evaluate their own editing procedures.

This evaluation aspect of editing tends to be discounted or treated as a side effect during the course of data processing, when all the emphasis quite naturally is placed on expediting the movement of good figures to final tabulation. Then, when the final posted tables are being reviewed for the published report, the thought arises, "What can we say, as

professional statisticians, about the limitations and qualifications of the data to be shown in our published tables? And more particularly, what should we be in a position to say, in text or tabular notes, about the weaknesses or shortcomings in the data as reported by respondents to the enumerators?"

The whole processing structure, but especially editing, must proceed with this objective in mind, if definitive answers to these questions are to be forthcoming at this final stage in data processing. Such evaluation concerns the criteria for accepting or correcting reported entries, for contacting the respondent in the case of dubious entries, and for estimating blank or omitted entries. Deviations from any set of criteria or procedural rules will, of course, occur in the processing of mass data, but such departures should be the exception, and for cause.

A clear understanding at the outset that strict adherence to the editing rules is to be the norm provides the setting for a valid post-census appraisal. The consistency of the editing and the type of record kept affect this ex post facto evaluation in at least two important respects:

(1) Qualifications and analytical studies of the published results.

dollars to cover eventual publication of data. Nonetheless, any given budget for conducting an economic census will yield more value, in terms of economic intelligence about a country's industrial complex, if the editing procedures provide a basis for such evaluations.

6.

ORGANIZATIONAL STRUCTURE FOR EDITING
AND CODING

The organizational structure of Providencia's NSO has, for the most part, been set up along functional rather than subject matter lines. Thus the Control Operations Branches have been given full responsibility for receiving all completed questionnaires coming into their Divisions.

Editing-coding and verification of editingcoding are performed by the Editing and Coding Branches--in the Data Processing Division for short forms and in the Manufacturing and Mining Division for long forms. In MMD, product and materials editing-coding are carried out by specialists and clerks who are closely supervised; this procedure is essential if establishments are to be assigned their proper activity codes.

The Control Operations Branches count and sort the questionnaires by subject census and transmit them in work units to the various other branches, where tests for processability, scope, and reasonableness are performed. Any rerouting of questionnaires that is necessary during the editing-coding operation is to be handled primarily by the Control Clerk in the DPD Control Operations Branch (see also section 10 of chapter V-1).

The computer will be used as much as possible to edit and code the questionnaires. However, some editing-coding is not feasible by computer (all material is not machine readable) and some would cause too many records to reject and eventually cost more in time and money.

The manual processing that is planned for Providencia primarily consists of insuring that questionnaires are processable. To be processable means that (a) the questionnaires must be complete, (b) all identification codes must be on each questionnaire, (c) all data must be in machine readable form, and (d) each entry must be in its appropriate place.

The manual editing and coding specifications refer to the manufactures long and short questionnaires. Procedures for the questionnaires for mining and for the electricity and gas industries would be similar.

The manual editing and coding sequences for the long and short forms are written as one procedure, although they were performed one at a time. The processing starts with a series of completeness checks. The basic question to be considered is: "Is the form processable?" In the case of establishments engaging 10 or more persons, only a clear answer of "Yes" to this question keeps the report in the regular processing stream. means that the tedious and sometimes complicated steps involved in coding and editing location, products, and materials are reserved for the large establishments--those that make such an investment in staff time worthwhile.

3. TESTS FOR COMPLETENESS (SEQUENCE A)

This

The purpose of this opening sequence is to determine whether the questionnaire can be processed. For this test, the NSO selected five critical items. These critical items of inquiry--number of persons engaged, total annual payrolls, cost of material, value of shipments and receipts, and nature of primary activity--appear on both the long and the short forms. For convenient reference in tabulating as well as editing, they are designated by the same item number. The lack of an entry for any one of the critical items means that the questionnaire has failed to meet the NSO processing standards. For the purpose of these tests, an entry is defined as a numeric value.

As the detailed flow chart for steps A-1 through A-5 for this sequence shows, all questionnaires that fail one or more of these