THE DEVELOPMENT AND TESTING OF A RESPONDENT RATING SCALE FOR OPINION AND MARKET RESEARCH INTERVIEWERS

620 35 9MB

English Pages 202

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

THE DEVELOPMENT AND TESTING OF A RESPONDENT RATING SCALE FOR OPINION AND MARKET RESEARCH INTERVIEWERS

Citation preview

The Pennsylvania State College The Graduate School Department of Psychology

THE DEVELOPMENT AND TESTING OF A RESPONDENT RATING SGAIE FOR OPINION AND MARKET RESEARCH INTERVIEWERS

A Dissertation by John Marshall Broun

Submitted in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY June 1951

Approved

y A

p

p

' r

Associate Professor of Psych o

v

e

d ___________ Professor of Psycholo,

Aporoved^^Siy

rj

'

Head, Department of Psychology

ACKNOWLEDGMENTS

The writer wishes to thank Dr. Lester P. Guest for his interest, guidance, and criticism during the development and testing of the Re­ spondent Rating Scale and during the writing of this report. Dr. Guest's research on interviewer-respondent relationship was particularly helpful in directing the writer's interest and in formulating the problem.

The

assistance of Dr. Guest in soliciting the aid of the National Opinion Research Center is also appreciated. The complete cooperation of the staff of the National Opinion Re­ search Center made this study possible.

Mr. Paul Sheatsley, Eastern

Representative, made available his office facilities and the entire field 3taff of National Opinion Research Center.

Mr. Herbert Stember, Study

Director, greatly aided in the development and pre-testing of the Re­ spondent Rating Scale items and in his constant correspondence with the writer.

Sue Smithers and *^yra Schuss of the office staff assisted in

the administration of the Respondent Ratings Scale.

The writer ac­

knowledges, too, the contribution of the 91 interviewers and 1276 re­ spondents throughout the country, who participated in the study. The writer expresses his sincere appreciation to Dr. Lester P. Guest, Dr. Joseph R. Hilgert, Dr. Albert K. Kurtz, Dr. Bruce V. Moore, and Dr. Kinsley R. Smith, for serving on his advisory committee and for their critical reviews of the report.

Dr. Albert K. Kurtz's valuable sug­

gestions for statistical treatment and for improving and clarifying the exposition are greatly appreciated.

iii

TABLE OF CONTENTS Chapter I

Page INTRODUCTION

1

A. Purpose of the Study B. Review of Literature II

PROCEDURE A. B. C. D. E. F. G. H.

III

25

Construction of the RespondentRating Scale Pre-testing of the Respondent Rating Scale Administration of the Respondent Rating Scale The Sample The Interviewers The Questionnaire Coding of the Data Statistical Treatment of the Data

REPORT OF THE RESULTS A. Testing the Major Hypotheses B. Comparing the Respondent Rating Scale and Questionnaire Results C . Factual Data and Supplementary Data

IV

SUMMARY

25 28 29 31 34 36 37 37 43 47 70 95 114

A. Purpose of the Study B. Important Procedures C. Important Results V

1 A

CONCLUSIONS

U5 116 122

A. Major Conclusions B. Minor Conclusions BIBLIOGRAPHY

122 125 128

APPENDICES Appendix As Appendix Bs

Sample Respondent RatingScales, Interviewers’ Instructions and Questions, and Codes Chi-square Tables

132 144

LIST OF TABLES Table Number I

Relation between the Number of Surveys for NORC in Which Interviewers Participated and Re­ spondents’ Indications of Whether Interviewer Took a Lot of Time to Explain Purpose of Survey

II

Relation between Interviewers’ Refusal Rates and Respondents’ Indications of How Much They En­ joyed the Interview

III

Relation between Interviewers’ Refusal Rates and Respondents' Indications of Usefulness of Opinion Surveys

IV

Relation between Interviewers' Refusal Rates and Respondents' Indications of the Way Thqy Felt IXirlng Interview

V

Relation between Interviewers' Refusal Rates and Respondents' Indications of Whether Other Things Were Discussed During the Interview

VI

Ratios of Usable Answers to Eligible Questions for Ten Items from the Respondent Rating Scale

VII

Relation between the Number of Non-legitimate "Don’t Know" Answers Given by Respondents and Their Indications of How Much They Enjoyed the Interview

VIII

Relation between the Number of Non-legitimate "Don't Know" Answers Given by Respondents and Their Indications of How Long the Interview Took

IX

Relation between the Number of Non-legitimate "Don't Know" Answers Given by Respondents and Their Indications of the Usefulness of Opinion Surveys

X

Relation between the Number of Non-legitimate "Don't Know" Answers Given by Respondents and Their Indications of the Suitability of the Time for the Interview

XI

Relation between the Number of Non-legitimate "Don’t Know" Answers Given by Respondents and Their Indications of How Th^y Felt During the Interview

Table Number

Page

XII

Relation between the Number of Non-legitimate "Don't Know'1 Answers Given by Respondents and Their Indications of Whether the Interviewer Talked With the Respondent or Didn't Talk Much About Anything Except the Survey

65

XIII

Relation between the Number of Non-legitimate "Don't Know” Answers Given by Respondents and Their Indications of Whether the Interviewer Went Right Into the Interview or Took a Lot of Time to Explain the Purpose of the Survey

66

Relation between the Number of Non-legitimate "Don't Know" Answers Given by Respondents and Their Indications of Whether the Interviewers Repeated the Questions Trying to Get an Answer or Went Right on to the Next Question when They Said They Didn't Know

67

XV

Relation between the Number of Non-legitimate "Don't Know" Answers Given by Respondents and Their Indications of Whether the Interviewers Made Them Feel Free to Give the First Answers that Came Into Their Minds or Made Them Feel They Should Think Carefully Before Answering

68

XVI

Relation between the Number of Non-legitimate "Don't Know" Answers Given by Respondents and Their Estimations of Interviewers' Opinions

69

Relation between Respondents' Answers to the Question "Do you expect the United States to fight in another war within the next ten years?" on the Questionnaire and the R. R. Scale

70

Percentage of Respondents Answering The Identical Question the Same Way or Differently on the Questionnaire and The R. R. Scale by Age, Educational Level and Economic Level of Respondent

72

Relation between True Interviewing Time and Respondents' Indications of How Long the Interview Took

74

Relation between Interviewers' and Respondents' Indications of Whether the Interviewer Went Right Into the Survey or Took a Lot of Time to Explain the Purpose of the Survey

75

XIV

XVII

XVIII

XIX

XX

vi

Page Relation between Interviewers’ and Respondents1 Indications of Whether the Interviewer Talked With Respondent or Didn’t Talk Much About Other Things During the Interview

76

Relation between Interviewers’ Ratings of Respondents’ Cooperativeness and Respondents' Answers to Ten Items from the R. R. Scale

77

Relation between Interviewers’ Ratings of Respondents’ Frankness and Honesty and Respondents* Answers to Nine Items from the R. R. Scale

84-

Relation between Interviewers’ Ratings of En­ joyment from the Interview and Respondents’ Answers to Ten Items from the R. R. Scale

89

Percentages of Respondents' Answers to Five Items from the Respondent Rating Scale ty The Time of Day the Interview Was Started

97

Percentages of Respondents' Answers to Five Items from the Respondent Rating Scale by Sex of Interviewers

98

Percentages of Respondents' Answers to Five Items from the Respondent Rating Scale by Age of Interviewers

99

Percentages of Respondents' Answers to Five Items from the Respondent Rating Scale by Educational Level of Interviewers

100

Percentages of Respondents' Answers to Five Items from the Respondent Rating Scale by Sax of Respondents

102

Percentages of Respondents' Answers to Five Items from the Respondent Rating Scale by Age of Respondents

103

Percentages of Respondents’ Answers to Five Items from the Respondent Rating Scale ty Educational Level of Respondents

104

Percentages of Respondents' Answers to Five Items from the Respondent Rating Scale ty Respondents’ Economic Level

105

vii

Table Number XXXIII

Page Percentage of Different Numbers of Nonlegitimate "Don’t Know" Answers Obtained by Interviewers ty The Time of Day the Interview Was Started and by Four Interviewer Character­ istics

107

Percentages of Different Numbers of Nonlegitimate "Don’t Know" Answers Obtained by Interviewers by Four Characteristics of the Respondents

109

XXXV

Percentages of Different Numbers of Non-usable "Don’t Know" Answers Obtained by Interviewers by Five Types of Interviewers' Ratings of The Interview and the Respondent

no

XXXVI

Ratios of Usable Answers of Eligible Questions for Four Respondent Characteristics and Three Characteristics of Interviewing

112

XXXIV

CHAPTER I INTRODUCTION

A.

PURPOSE OF THE STUDY

Opinion researchers generally recognize the importance of the interviewers' work, but the problem of evaluating their work has al­ ways been difficult.

Several studies evaluating the work of intervisu­

ers have been reported.

I-lost of these studies have shown that inter­

viewers can and frequently do bias respondents' answers, but very few studies have attempted to find the cause of the bias or to measure the relationship between interviewer and respondent while the answers were being solicited.

The present study developed and tested a respondent

rating scale for measurement of the interviewer-respondent relationship. A t present, there are two common methods of estimating interviewers' proficiency,

subjective ratings by supervisor;.’- personnel and objective

ratings from the returned questionnaires.

The latter usually consist

of a check on the number of "don't know" answers obtained, number of usable answers, legibility of writing, fulfillment of quotas, or some combination of these.

All studies of interviewer proficiency have been

handicapped by the lack of any checks on the validity of respondents' answers.

Verification of respondents' answers is not possible since

opinions m y

vary from moment to moment; written expression, verbal ex­

pression, and overt behavior may not necessarily be the sane. Recognizing both the importance of the interviewer-respondent relationship and the fact that study of it has been somewhat neglected, the writer developed and tested a new method of assessing the relationship

between interviewers and respondents.

Respondents could indicate good

or bad interviewer-respondent relationship by their answers on a rating scale.

Three areas influencing interviewer-respondent relationship

were measured by respondent ratings: 1. Characteristics of interviewers' work 2. Qualities of the questionnaire 3. Respondent feelings at the time of the interview. The following steps were taken in exploring this problem: 1. The development of the Respondent Rating Scale (hereafter re­ ferred to as R. R. Scale) for opinion and market research interviewers 2. The administration of the R. R. Scale to the respondents in a regular national poll 3. The comparison of results from the R. R. Scale with criteria of interviewer competence. Six major hypotheses were tested in this study:

In rating charac­

teristics of interviewers' work, qualities of the questionnaire, and their feelings at the tine of the interview— j-. 1. Respondents will give better” ratings on the It. P.. Scale when interviewed by interviewers who were judged as best rather than mediocre by supervisory personnel. 2. Respondents will give better ratings on the It. R. Scale when interviewed by more experienced rather than less experienced inter­ viewers .

wIn consultation with researchers the better answers for the R. R. Scale items were selected. They are marked with an asterisk (*) on the R. R. Scale, Appendix A, page 135.

3

3. Respondents will give better ratings on the R. R. Scale when interviewed by interviewers who have lower rather than higher refusal rates.

4 . Respondents who were among the last three interviewed by an interviewer will give better ratings than those who were among the first three interviewed on any one survey. 5. Respondents who give more usable answers to free response questions will give better ratings on the K. R. Scale than those who give fewer usable answers. 6. Respondents who give fewer "don't know" answers not considered as legitimate answers will give better ratings on the R. R. Scale than those who give more "don't know" answers. To check the above hypotheses it was necessary to have a regular survey as vehicle for the study.

Through the cooperation of national

Opinion Research Center (hereafter referred to as IIORC) the R. R. Scale was used on one of their periodic national polls.

Respondents' ratings

on the R. R. Scale were compared with the following criteria*": 1. A subjective numerical rating for the past year's work, given the interviewers by NORC personnel 2. The actual number of surveys for NORC in which interviewers participated 3. A ratio of the number of refusals to the total number of inter­ views completed on this survey 4. A ratio of the total number of usable answers to the total number of questions where there should have been an answer

"Each criterion is more fully discussed in Chapter III.

K

5. The number of "don't know" answers obtained which were not considered legitimate. In order to test the fourth major hypothesis the ratings given by respondents who were among the first three interviewed were canpared with those who were among the last three interviewed by any one interviewer.

If any or all of the above hypotheses were proven

correct it would give additional support to the use of such criteria for selection of interviewers.

It would also indicate that an

R. R. Scale might be used as a selection tool.

If none of the

hypotheses were proven correct it would question the validity of the criteria and the R. R. Scale as measures of interviewer-respondent relationship. B.

REVIEW OF LITERATURE

On surveying the literature, the writer noted a paucity of pub­ lished material on the exact topic, on the spot measurement of inter­ viewer-respondent relationship.

However, many publications dealt with

closely allied topics, which aided in the development of the R. R. Scale.

The writer selected those which dealt with closely related

problems and grouped them into two categories:

Interviewer Problems

and Rating Methods. 1. Interviewer Problems This category includes a review of important publications concern­ ing interview bias, selection, and training, a. Interviewer Bias In 1929 Rice (38)* published one of the first critical studies of

•Figures in parentheses refer to numbered bibliography beginning on page 128.

5

interviewing.

The causes of the destitute condition of 2000 boneless

non wore determined by 12 interviewers.

The nen were randomly assigned

to the interviewers for questioning, but ever;- interviewer obtained significantly different responses to the sane questions.

For example,

interviewer A, an ardent prohibitionist, found liquor the major cause of downfall, while interviewer 3, a socialist, found industrial condi­ tions nore often the c a u s e . Blankenship (3) reported,

in 194-0,

three interviewers, each representing a

the results of a study using different church and age group.

Ety- checking the interviewer's questionnaire against interviewer obtained, he found evidence of bias.

the results each

Each interviewer r e ­

ceived a larger proportion of answers similar to his own questionnaire replies. Two years later, Stanton and Baker

(4-9) published the results of

a study to test three problem areas: 1. The bias of results even when interviewers were properly trained and warned of the dangers 2. The effect of bias on completely or incompletely learned materials 3. The amount of individual differences in bias from interviewer to interviewer. Using undergraduate students in a laboratory situation they found: 1. Bias took place. 2. Incompletely learned material had a higher rate of bias. 3. There were no significant results showing individual differences in the amount of bias. Friedman (20) reported an experiment in which she repeated as

6

nearly as possible the aforesaid study.

Her results showed no statis­

tically si,gnifleant bias and therefore pub

generalisation of Stanton

and Baker's study in question. In 194-2 Katz (2S) published the results of a study done in Pitts­ burgh, Pennsylvania.

Using two groups cf interviewers, middle class

white collar interviewers and working class interviewers, he found a difference in results from comparable samples of respondents.

The

white collar interviewers found a greater incidence of conservative attitudes among the lower income groups and less support for isolation­ ist sentiments than did the working class interviewers.

The differences

between the interviewing groups increased when union members or their relations were interviewed.

The results of the moi-e experienced inter­

viewers were more nearly the same as those of the working class inter­ viewers . Studying the same problem areas, Williams and Cantril used, two groups of interviewers.

(54) also

No significant differences in results

on political questions were obtained by Negro and white interviewers working in Harlem, New York.

However, there was none evidence that the

Negro respondents tended to hide some answers from the white inter­ viewers . Frcidberg, Vaughan, and Evans

(IB) of the Psychological Corporation

reported that a minimal amount of bias was obtained from interviewing when a large number of interviewers did a comparatively snail number of interviews.

They seemed to suggest that if there wore bias effects,

they tended to cancel out when a large group of interviewers wore used. Robinson and Rohde (39) used four groups of interviewers in an ex­ tensive anti-Semitism poll in New York. City.

They reported

that:

r-f

I

1. Anti-Serai tic views tended to be withheld from interviewers with Jo*rich appearance or none. P.. Anti-Sor.iitic resooncoe were wore freouent iron the lov; cconor.iic and educational groups than iron the middle or high proups. There wa s an indication

bha i lower economic and educational groups ;;cre

more influenced by the interviewre r s 1 aopearc.ncos or n a m e s , but that they ’./ore also .’ .ore prone to spree with stereotypes.

Jims direct q*aostions

revealed more anti-Somitisr. tl'.an did indirect questions. In 194.7 Shapiro and Jiborhart (4-7) reported results of a chec3: for bias in open ended interviewing.

They compared each interviewer 1s re­

sults with the group results and found that one of the four intei’vicwers studied had significant deviations on several iteas.

They checked sev­

eral points, among which were: 1. Number of "don't know" answers obtained to certain questions

2 . Humber of multiple response ansv/ers to a specific question 3. Number of non-ascertained ansv/ers to some factual data itcr.is. They found interviewer differences,

but made no interpretation.

Reporting a study in which he used recordings of a standard inter­ view, Guest (24.), had one coached respondent jive set answers to a "roup of interviewers.

He showed that a larje majoi’ity of the errors that

interviewers made could not be detected from the returned questionnaires, Jhus he suggested that present methods of judging interviewers 1 worl: from the returned questionnaires night not be accurate.

There was

little relationship between the respondent ratings and judges' of the recorded interviews.

opinions

Nevertheless, it was suggested that further

refinement of the rating scalo used by the respondent and more cases would probably yield more significant results.

Using a toot re-ter.t study, Crerpi (12) in 194-9 studied the effect of interviewing on the respondent.

He found n significant shift awry

from r.o opinion and suspected the following five possible biases tlmt the interviewer night cause: 1. Pre-amble bias- The introductory phrases before the questions 2. Topical bias- Dwelling on one side of an issue 3. Oversimplification bias- Ashing either-or questions as if they were the only alternatives 4-. Salience bias- Intensity of respondents' feelings not indicated by question answers 5.

Commitment bias- ''Pushing" for answers when the respondents do

not know answers. Checking on classifications of factual data items, hosteller (4-1) reported that second interviewers' classifications correlated .63 with first interviewers' classifications of economic status.

The changes

in classifications were mostly toward higher economic classifications. Ho did not call this bias, but pointed cat that an btiiervlewor when working on a quota assignment "ray be tempted to classify a nearly borderline ind.ividua 1 in a category which will best suit his assign­ ment. " In 194-9, Blankenship (5) reported a follow up study checking on classifications of standards of living.

His follow up interviewers

agreed with the original classifications 93 per cent in terms of over all results.

In contradiction to hosteller's report he found that most

of the reclassifications made were toward a lower standard of living. Blankenship interpreted his results as showing a distinct bias in the usual ouota-control samoling method.

In a study using student interviev/ers, Uyatt and Campbell (55) chocked the possibility of interviewers' expectations as a cause of bias In a study about the l?4fi presidential campaign. consistent or statistically significant biases.

They found no

Where there was cone

indication of bias, interviewer expectation of respondents' opinions seemed to be a greater source of bias than the interviewer's own opinion. The experiment might differ from other results due to: 1.

Use of student interviewers who were relatively inexperienced

and untrained P. Study of only one community 3. Elimination of all dishonest interviewing. Attempting to study how attitude structure expectations distorted answers given by respondents, Smith and Hyman (44-) found that inter­ viewers ' expectations were much more important than their ideologies in producing biases.

Little evidence was found showing experienced inter­

viewers (one year or more) caused less bias due to expectation.

It was

suggested that inexperienced interviewers night be less likely to feel the pangs of conscience or to understand the very strict iml.es govern­ ing recording of answers. Fisher (16) studied the recordin.g operation to note the relation­ ship of bias and recording ability.

In general, he found that inter­

viewers tended to record more statements which conformed with their own attitudes.

Interviewers who were good on a recording test tended to

obtain a larger percentage of free response answers,

more free response

answers were obtained near the end of the interview than at the begin­ ning, indicating that practice effect was more important than fatigue. Ho differences in bias of experienced versus non-experienced inter1 v iewers were found.

10

In another laboratory study Guest and Iluckols (2?) reported a small tendency for interviewers to bias respondents' answers to fit their own point of view by inproper recording of reEDor.ce s .

P..esuits

of the kinnesota Clerical fest showed no relations with the errors made, but interviewers with higher intelligence neened less likelv to :.iake •i

biased errors.



v

Although errors did exist in the form of incorrect re­

cording, any one interviewer tended to nake errors which canceled out as far as direction of error was concerned. The review of pviblications on interviewer bias s u w e s t s that inter­ viewers can bias respondents' ansv/ers.

Up to the present tine interviewer

bias does not seen to influence results of surveys too greatly, but it nay become worse and from a theoretical viewpoint it would be best to eliminate it.

There is also an indication from the publications that

more research might explain the causes for the bias, so that it can be controlled or eliminated.

Several of the methods used to detect bias

and suggestions given in the articles were particularly useful in the development of the P. R. Scale.

The use of respondents rating inter­

viewers' work, suggested by Guest, was particularly important in the development of the R. R. Scale, b. Interviewer Selection In order to have good interviewing, it is important to select com­ petent individuals to do the work.

Selection procedures for most poll­

ing organisations are still very ragged or non-existent.

However, the

problem is very important and as Connelly and Harris (11) have stated, "the original selection of the Interviewer is at least as important as one full days'

training."

The above mentioned authors published more concrete suggestions

11

about interviewer selection than anyone else.

They told of MOP.C 1s

usual procedure of selection, which consists of locating applicants by writing five or si:-: newspapermen, school officials, librarians, social workers, or employment agencies for names of recommendcec.

Those recon-

nendces v;ho are fairly well educated and interested in the work beyond the point of wages could then be hired.

In addition, the authors sug­

gested the possibility of hiring a cross section of the population, which would supposedly neutralize bias.

This, of course, would be im­

possible, since some interviewers of very limited education would be hired who could not do the work. Connelly and Harris further suggested that polling organizations needed better methods of locating applicants. be selecting new members entirely by mail.

One possible method might

by sending these members

somo initial assignments, a check, might be made of their interests and how they handled detail.

Ivon an estimate of interviewer-respondent

rapport might bo gained from the quality of their free response answers. If they proved to be efficient, they could be given personal training, .’his selection method would be useful in helping to exit e:rpenses.

The

writers also advocated tho use of interviewers with strong feelings rather than those with neutral or no strong views.

This view was sub­

stantiated by reports of studies which showed that interviewers w ith no opinion on a question reported more of their respondents with correspond­ ing views than did either pro or con interviewers. On the contrary Uechsler (52) commented that interviewer effect on the respondent was an important topic that was often overlooked.

The

tone of voice and method of presentation, he wrote, might easily in­ fluence the renlies a resoondent gave.

He cautioned against the uso of

12

Interviewers having strong; opinions since such an interviewer would be likely to influence his respondents’ answers without trying; to do so. In an article describing a series of meetings h.old for interviewers of Kid-Continent Surveys, interviewers.

Borg

(6)

suggested cone qualifications for

Sane of the na jor qualif icat Ions given ’.;ero as follows:

1. Genuine love for people regardless of the respondents'

economic

status in life 2. A thoroughly honest approach in handling every assignment 3. A n above average intelligence,

sprinkled with a conscientious

attitude 4. A 100 per cent neutral attitude on any subject when interviewing 5. The ability to recognise differences between direct and evasive answers in open end questions 6. The ability to be adaptable under any circumstances, among any group of people 7. A master at readable longhand writing, utilising good abbre­ viations which are recognisable to others S.

A friendly, unhurried attitude with respondents, giving suf­

ficient time for each question to be answered thoughtfully, but yet conducting the interview in a businesslike manner 9. The ability to follow instructions to the letter 10. The capability of reading all questions exactly as worded 11. The ability to be brief. In her article about interviewer problems, Andrews qualities which should be found in interviewers.

The main personal

characteristics which she stated as desirable were: must—

(l) suggested

The interviewer

13

1. like people 2. be able to adjust to all cultural classes and all economic brackets 3. have a pleasant personality /f. be sound in health 5. write legibly 6. have a genuine knowledge of his territory 7. be honest in record keeping S. be well but simply dressed 9. be attractive looking, without physical defects. Eleanor P. Clarkson (10) in her article on honesty of interviewers, suggested qualifications which interviewers must have as follows: The prospective interviewer must— 1. like people 2. be the kind of person you'd leb in your home 3. have a sense of humor J+. seen dependable 5. have a sound reason for wanting the job 6. fit in with the other interviewers 7. have a clear idea of the way she can organize her time to do an adequate job S.

be the type of person that you as a supervisor would like to

know better. She further stated that in selection you should never hire anyone with obvious marital difficulties, poor health, pre-school children (unless there is a well worked-out arrangement for their care), or anyone whose husband is not in sympathy with her working.

(The author is Field

Director for Division of Header Research of the McCall Corporation and evidently uses only feiaale interviewers in a narrower area of opinion polling than is covered by sone other polling organisations.) According to Blankenship {4 ), the interviewers should be personally selected by one or more members of the central office staff.

‘ fhe use

of correspondence for selection was frowned on as completely inadequate. Interviewers must be selected by personal interview, during which the staff members should determine whether the applicant is intelligent, e::trover ted, honest, healthy, and capable of following directions.

Per­

sonal references should be checked and the interviewer should be chosen for the particular type of survey being undertaken. At the Central City Conference on Public Opinion Research, 1946, Maloney (36) suggested the necessary qualifications for an interviewer as: She should— 1. be old enough to "know her way around" 2. have a permanent residence-stability 3. like people

4 . not be opinionated 5. be able to understand written instructions. lie suggested that the first step for hiring part time interviewers should bo contacting,

through correspondence, likely people who could

recommend interested, suitable prospects,

bchool officials were sug­

gested as good contacts. llaloney went on to describe the ideal interviewer as a married woman, thirty seven years old, neither adverse nor too aroused about politics, and able to understand and follow instructions.

15

Oilman II. K. Smith (36) commented on lialoney*s renarks by suggest­ ing screening of extremists on political and religious subjects.

Ho

told of Opinion Research Corporation's screening questionnaire which was sent before an applicant was seen.

lie added that Opinion Research

Corporation recruited their interviewers through the rir.il and that re­ commendation of other interviewers brought in the majority of new workers. oac

gs

lie said lawyers and newspapermen provided pood initial con-

.

Smith and Maloney disagreed on the amount of education an inter­ viewer should have. years of college.

Snith preferred interviewers with at least two Due to the increase of opinionated views with edu­

cation, Maloney contended that "not too intellectual" interviewers were better. Selection of interviewers was discussed by Reed, Parker, and Vitriol (37).

Some sources of interviewers they suggested were:

letters of application, newspapers, academic institutions, w o m e n ’s organisations, and key people in the community.

They further suggested

that there is a slight trend toward a "core or cadre" of full time interviewers.

To them the most important point for selecting inter­

viewers was to avoid extremes.

Interviewers should liavo a clean and

pleasing appearance, not be too aggressive or backward, be free from obvious racial, political, and religious prejudices, and have a sym­ pathetic attitude toward people. To obtain more information about the methods of selection used by opinion and market research organizations, the writer corresponded with several prominent researchers. Although there didn't seem to be any general methods of interviewer

16

selection reported, several objective qualities were suggested as im­ portant.

-jovrles (7) reported that Joe Belden and Associates prefer

women, ages twenty-five through forty-five, who are housewives and have at least a high school education.

Keier (32) stated that the lav;a Poll

generally prefers to have in tiervi ewers with expe rienco and at least a high school education: he regarded age and sox of lesser importance. Gaffin (22), of Ben Gaffin and Associates, on the other hand, reported a preference for college graduates and interviewers with no previous experience. hoot of the researchers reported that subjective methods of selection are used.

An application form is generally used and some

sort of screening interview is held with the apolicants.

Dodd (14;

reported that an application blank "supplemented, wherever possible with personal conferences with prospective interviewers11 is used by 'Washington Public Opinion Laboratory.

The selection at Foote, Gono

and Gelding, reported by Udow (31), consists mainly of a screening interview by o girl who has been an interviewer herself. familiar with the problems she is assigned interviewers.

Since she is

hie duties of selecting new

Cannol (9) stated that interviewers for Garvey r.oscarch

Center "are selected on the basis of personal interviews designed to evaluate the personality of the prospective interviewer...11

lie further

stated tiiat they try to select interviewers who have an educational background in the social sciences and with some past experience.

Ac­

cording to I.eier (32) the Iowa Poll prefers certain objective qualities in their interviewers and also evaluates the previous experience, religious and political affiliation or preferences.

Although Gaffin

mentioned a ^reference for two objective criteria he stated that

(22)

17

"Probably the most important factor is the appearance of intelligence and social ngreeability which the interviewer manifests toward the supervisor at

.he hiring interview."

Since his organization lias been established for over twenty-five years Gross.ley (13) reported that lie has ulerty cf applicants.

These

applications are processed by ::.enns of forms and put aside for possible try outs.

. ’ J upervisors

then meet with the an licants when interviewers

are needed in their area.

bhile bowles

(7 ) ri'yectol

several objective

criteria, he also uses training meetings as op or tun it. os to uwced out" interviewers "who are undesirable due to physical defects, poor appear­ ance or inaptitude." Host of the researchers reported that an analysis of performance is used for a validation of the selection systems, but no validation studies iiave been done.

In general, researchers feel that they hire

good interviewers with better than average ercpectancy. A review of material concerning interviewer selection showed that several objective criteria for selecting interviewers are used, but practically everything done in selection is a matter of opinion or judgment.

It would be impossible to list all the characteristics of

good interviewing suggested by all supervisors.

Many cacable research

men vary ’widely in their judgments of good and bad intcrviowers.

On

the whole, no real studies of Interviewer characteristics have been re­ ported, not because of incompetence among researchers, but because where do not seem to be any criteria for adequate evaluation.

T .Jhen solec'tion

systems have been validated, it has been done on a subjective basis, c. Interviewer Training Good interviewing depends not only on selection ’.rat on adequate

braining.

As Blankenship (4) stated,

"the interviewer who chows great­

est potentiality cannot Bo good work without proper training.” Ike amount and type of training p.ve!i intervio\/ers varies iron practically none, oneopt oreli: ;innry instructions for a cnocific job, to a completo orientation of the company consisting of several weeks in the home office petting the feel of survey work and special training in interviewing procedures and techniques. Blankenship further stated that interviewing is more of an art *

than a science.

However, he suggested general principles which should

be included in the training of interviewers.

home of thorn were as follows

The interviewer should— 1. make a few trial calls before beginning his assignment 2. be taught the proper time to approach the type of respondent rcauired 3.

be taught to obtain the exact sample quotas

4.

be taught to follow instructions

3.

betaught to ask questions word for word as

o.

betaught to be neutral during the interview.

Connelly and karris

on the cuestionnaire

(11) also suggested that there was room for

improvement in the quality of current interviewing and that training which night at first sees financially prohibitive night in the long m m be more economical.

Reporting IIOEC’s training program., the authors ad­

mitted that often the training program was only partially fulfilled. Each new interviewer was to receive one full day of training by a super­ visor.

But, in the past the training had rarely been more than three,

and in come cases only one hour of field worl: under supervision.

UOP.Cc

training is supplemented by a manual, Interviewing for h'QRC (/+6), which

each interviewer is to master, and by detailed specifications for each survey.

Additional training throuyhout the working ncriod is achieved

by the use of letters from the home office which offer suyyccticns for each interviewer's work. In 1947 none changed its procedure so that each new interviewer roc elves more hours of field work and completes five interviews after observing at least one interview by the supervisor.

After each inter­

view, examples of pood and bad interviewing are discussed.

In addition

each trainee completes specified amounts of interviewing and deals with, specific problems set up by the trainer. Kay and Schlick (19) reported that the role-practico method of training was peculiarly successful in developing skill and insfyht in the interviewer.

'./hile studying social conflict in a community of

nixed ethnic groups,

the authors reruired interviewers to obtain in­

formation regarding respondents'

intimate, private attitudes.

The

interviewers were trained by practicing with one another until they were adopt at several roles, representing different approaches and how to act with different types of respondents.

After classroom work -they

were supervised in practice interviewing in the community. The basic instructions which ITORC uses as a guide for interviewers wero reported by Williams (53).

The instructions serve as background

and reference material during the personal training period.

They in­

clude an Introduction, followed by a short description of methods and means to use in selecting respondents, a description of economic group­ ings, how to make the approach and introduction, what sort of manner and attitude to maintain during the interview, and additional specific information on aye and economic classifications, place of interviewing, and the use of substitutes.

20

Maloney (36) suggested proceeding with training of prospective interviewers in the following manner: 1. At a group meeting, present a general introduction about opinion research,

importance of interviewing, and techniques and methods es­

sential to successful interviewing. 2. After the group meeting tali with trainees individually and go over a typical questionnaire with them. 3. Take trainees,

two or three at a tine, into the field and let

them observe the trainer doing an interview.

Have each trainee con­

duct several interviews and uoint out corrections on the spot. A study comparing the work of trained versus untrained interviewers was recorded by Rugg (Al)•

Interviewers from the staff of HGRC and

American Institute of Public Opinion (AIP0) were used for the compari­ son.

AIP0 interviewers were secured and supervised by mail, roceiving

a basic manual to read before beginning work, but having no personal training or supervision.

The IJORC interviewers received a manual of

instructions, personal supervision, and trial interviews.

Comparable

assignments were given to both groups and the returned questionnaires were rated.

There was no marked difference between the judged quality

of work of the two groups. two groups,

Where there was some difference between the

it was in favor of the trained Interviewers.

It can be noted from this section that training of interviewers generally was done by supervisors. reported to be advantageous.

The use of prac tice interviews was

Most organizations supplemented inter­

viewers1 training with a general manual and specific instructions for each survey.

Host evidence leads to the assumption that great gains

in interviewer efficiency might be effected by additional training.

21

Obviously the training programs wotild be more efficient if exact qualities of good interviewing were I m o u n .

The P..

Scale will at­

tempt to Isolate some of these qualities. 2. Rating Methods Systematic recording of observations and judgments on a scale of units or values may be classified into four categories:

rating scales,

rani: order method, paired associates method, and forced choice tech­ nique.

The rank order and the paired associates methods necessitate

judging all the persons

bo be rated.

It is impossible to meet this re­

quirement in the opinion research situation since each resiiondent (judge) meets only one interviewer.

Thus the rating scales and forced, choice

technique are the ones available for use in this study. Some traits are more amenable to rating than others; Ilollingworth, Miner, and Shen (27, 33, A3) reported work on this problem.

It seemed

according to Ilollingworth, that close agreement among raters was more likely on traits such as efficiency, perseverance, quickness, judgment, clearness, and will.

On the contrary he reported poor agreement on

courage, unselfishness, integrity, cooperativeness, and kindness. Miner found higher agreement for traits such as scholarship, leadership, and intelligence and low agreement on traits such as punctuality, judi­ cial sense, and tact. There are two major types of rating scales which would be useful for an R. R. Scale: 1. A number of phrases of varying degrees of a trait, characteris­ tically arranged in order to form a descriptive scale 2. Shorter descriptive phrases printed at appropriate positions under a straight line, usually called a graphic rating scale.

The two types are very similar, but the former easier to use with untrained judges.

Of all possible methods the graphic type is most

popular and most satisfactory according to Freyd (19).

he listed the

advantages as simplicity and comprehensibility, freedom, from direct quantitative terras, and ratings as finely discriminated as the rater wishes.

host of these advantages also apply in the case of the d e ­

scriptive scale. The reliability and validity of ratings have been studied frequently. Generally the reliability is increased by increasing the number of judges used.

Authors stiggested varying numbers of judges for .good reliability

obtained.

Bradshaw

(8) reported

that from five to twenty-one judges

should be used, Symonds (50) recommended eight judges, and Hugg (4-0) suggested at least three.

Furiey (11) has shown that the re3j.abdl.ity

of a rating may be increased by subdividing a characteristic and having the judges rate each sub-characteristic. On Array rating tests the Staff of the Adjutant General's Office (4-7) used the forced choice technique to offset markedly negatively shewed, leotolcurtic distributions with low validitv. x

t>
lied to tests involving one degree of freedom.

This correction,

’21own as la t e 1c correction for continuity, is particularly important when applying the chi-square test with only one degree of freedom and when the uncorrected chi-square is just above a customary confidence limit.

However, Guilford (26) states that "there is nrobably nothing

to be gained by applying Y a t e 1e correction when there is more than one degree of freedom, and under these conditions the correction becomes complicated." In the present study the cariter lias always attempted to avoid small theoretical cell frequencies by combining adjacent categories.

In some

instances it seemed unwise to name combinations so small theoretical frequencies were necessitated.

fhe number of degrees of freedom were

always large, nov/ever, and. the chi-squares were far above or below the customary confidence limits. 'fhe chi-square value empresses the other than chance differences between the observed and expected frequencies,

fhe probability of ob­

taining differences as great as those between the observed and expected frequencies per category on the bases of chance alone can be estimated from the sampling distribution of chi-square. Sampling distributions of the chi-square statistic are not normal unless there are thirty or more classes or categories.

The distribution

for less than thirty categories varies markedly depending upon tho number

40

of classes.

Tables of probability values for categories ranging fron

two tlirough thirty can be found in most statistics boohs.

Those tables

arc commonly set up with a series of probability values (r) developed in terns of degrees of freedom (df) with the chi-square values in the >f the table. The degrees of freedom for any chi-square is equal to the number of classos or categories for which hypothetical frequency values can bo freely assigned.

Thus the degrees of freedom is ecus! to tho total

number of categories minus the number of constraints imposed on tho data when computing the hypothetical frequencies. In this study the degrees of freedom for each chi-square van the product of the number of categories in each variable minus one. Peters and VnnVoorhis

As

(34) have pointed out, the determination of tho

degrees of freedom depends upon the meaning of the problem.

The’’ state

that if one is "talking about the general case in which the samples may vary in every respect except 1-1, the number of degrees of freedom, is (nr — l ) , where I! is the total population and n' is the total number of colls."

Peters and VnnVoorhis further state that "workors have

followed for all purposes fisher’s lead in using (k - l)(r - l) indiscriminatingly.11

This is the correct method when tho marginal totals

are to remain constant from sample to sample, but in this study’ there was no reason to believe that they would.

Hence the only constraint on

tho data, was the total number of cases (II) . After computing the chi-square statistic the 'probability of obtain­ ing such a value was found by reference to a chi-square distribution table.

Entering the table with the correct degree of freedom, the

probability value of obtaining a given chi-square was road directly.

a

Some of the contingency tables in this study contained sore than thirty-one cells, thus more than thirty degrees of freed on.

Peters and

VanVoorhis (34) have cited Fisher’s proposal for obtaining probability values for chi-squares when the degrees of freedom arc greater than u.n*v-hrr

Assuming that

2df - 1

V

nay be treated as a normal deviate about

as the mean it is possible to obtain a probability value

from a table of normal distribution.

this procedure van used in the

present study whenever there v;ere more than thirty-one cells in a con­ tingency table. The computed chi-square value enables one to estimate the proba­ bility of obtaining such distributions by chance.

Then the value is

low, there is very little probability that such cell frequencies were obtained by chance alone; therefore, there is probably a reliablo association between the variables of the contingency table.

The amount

of association may be indicated by a contingency coefficient (C).

1/hen-

ever the probability of obtaining such distributions by chance uas be­ low .05 in this study, a coefficient of contingency was computed by means of the following formula:

C =

i n

The letter C indicates the contingency coefficient, Xf~ indicates chisquare, and IT the total number of cases in the contingency table. The coefficient of contingency varies between zero and one and may roughly be interpreted in the same manner as a proauct-monent coefficient of correlation.

It does not tell the direction of association and is

limited by the number of categories into which the distributions are

49

divided.

The direction of association mist be determined by inspection

of the frequencies in the contingency table. A contingency coefficient of one can bo obtained only vhen the num­ ber of classes or categories is large.

Tor exanole, when data is ar­

ranged a 10 x 10 Table, C cannot exceed .949, when a 4- x 4 Table, C cannot exceed .066.

Discussions of the limitations of C and tables

shoving the maximum value may be found in several statistics boohs (26, 34, 35).

Care in interpretation of G values must be tahcn parti­

cularly where the number of categories is low. All statistical computations v/ere cheched by recomputation by another person.

The following additional checks were also made:

1. The sums of the expected frequencies for each row and each column

ere checked with the obtained marginal frequencies.

2. 'fhe products of the computed cell square contingency values and the e j e c t e d frequencies were checked with the squared differences between the observed and expected frequencies. 3. The suras of the cell square contingency values were checked by several additions, each time in a different direction. 4. The significance of the chi-square values were checked by canparison to a table showing the minimum values necessary for tho .05 and .01 level at various degrees of freedom. 5. The computed C values were checked against a conversion table for converting chi-square values to C when II is constant.

(This was

used only as a check since II varied slightly in some cases.)

CHAPTER III REPORT OF THE RESULTS The results of this study are divided into three sections accord­ ing to the data treated.

The first section treats the data concerning

the six major hypotheses.

These data compare the frequencies of

respondents' replies to every item on the R. R. Scale with the follow­ ing: 1. A subjective numerical rating for the past years' work 2. The actual number of surveys for NORC in which interviewers participated 3. A ratio of the number of refusals to the total number of inter­ views completed on this survey A. The serial position of the interview in relation to the total number of interviews done by one interviewer 5. A ratio of the total number of usable answers to the total num­ ber of questions where there should have been an answer 6. The number of "don't know" answers obtained which were not con­ sidered legitimate answers. The second section of the results treats the data that compare respondents' answers to R. R. Scale items with answers to specific questions on the questionnaire.

This section presents the following

comparisons: 1. Respondents' answers to an identical question on the question­ naire and the R. R. Scale 2. Actual interviewing time and respondents' indications of how long the interview took

A4

3. Interviewers’ and respondents’ indications of how much time the interviewer took to explain the purpose of the survey 4. Interviewers’ and respondents' indications of whether other things were discussed during the interview 5. Interviewers' ratings of respondents’ cooperativeness and respondents' answers to ten items on the R. R. Scale 6. Interviewers’ ratings of respondents’ frankness and honesty and respondents' answers to ten items on the R. R. Scale 7. Interviewers’ ratings of enjoyment from the interview and respondents' answers to ten items on the R. R. Scale. The third section of the results treats factual data from the questionnaire and supplementary to the major hypotheses.

The following

data are presented: 1. Percentages of respondents' answers to five items on the R. R. Scale by the time of day the interview was started and by three inter­ viewer characteristics (sex, age, and educational level) 2. Percentages of respondents’ answers to five items on the R. R. Scale by four respondent characteristics (sex, age, educational level, and economic level) 3. Percentages of different numbers of non-legitimate "don't know" answers by the time of day the interview was started and by four inter­ viewer characteristics: a. Sex b. Numerical rating for the past year c. The actual number of surveys for NORC in which interviewers participated d. The refusal rate. 4. Percentages of different numbers of non-legitimate "don't know" answers by four characteristics of the respondents (sex, age, educational

level, and economic level). 5. Percentages of different numbers of non-legitimate "don’t know" answers by five types of interviewers' ratings: a. Interviewers’ ratings of respondents' cooperativeness b. Interviewers’ ratings of respondents' frankness and honesty c. Interviewers’ ratings of their enjoyment of the interview d. Interviewers' indications of how much they talked about other things during the interview e. Interviewers' indications of how much explanation of the purpose of the survey they gave to the respondents. 6. Ratios of usable answers to eligible questions by: a. Four respondent characteristics (sex, age, educational level, and economic level) b. Three characteristics of interviewing: 1. The time of day the interview was started 2. The length of time to complete the interview 3. The size of town where the interviewing was done. Results for the first two sections are primarily in the form of chi-square tests.

In the appendix all contingency tables are shown

with the obtained and expected frequencies in each cell and the chisquare for the table.

When the chi-square is significant at the .05

level or better, a contingency coefficient is shown. When the chi-square value computed for any one table is shown to be significant at the .05 level or better there is probably some re­ lationship between the variables in the contingency table.

It is

possible to state that at least some of the relationship is based on other than chance factors.

A chi-square value significant at the .01

level indicates that the differences between obtained and expected cell frequencies could be expected by chance alone only 1 time in 100. When the chi-square statistic is significant at the .05 through .02 level it is marked with a single asterisk (*).

When significant at the

.01 level or better it is marked with a double asterisk (**).

If the

4.6

chi-square value for any table is significant at the .05 level or

better, the table is presented in this chapter and in the appendix. If a table does not have a chi-square value which is significant at the .05 level or better, the table is shown in the appendix only. Tables presented in this chapter show the cell square contingency value for each cell, the chi-square for the table, the degrees of freedom, and the contingency coefficient.

In each cell where the obtained fre­

quency is larger than the expected frequency a plus sign (/) is placed below the cell square contingency value.

The direction of association

may be found by noting the trend of the plus signs through the table. For each table in this chapter a table in the appendix shows the obtained and expected frequencies for each cell instead of the cell square contingency value.

Die table number and page number for corre­

sponding appendix tables are found in parentheses beneath each table in this chapter. In all chi-square tables the symbol X2 indicates the chi-3quare value for the table, df indicates degrees of freedom, and C indicates the contingency coefficient. Results which are not tested ty the chi-square technique are pre­ sented in two other forms.

Data for one of the major hypotheses is

necessarily in ratio form and is presented without tests for significance, since no suitable test was available.

Factual data from the question­

naire and supplementary data are presented in percentage form without tests for significance of differences.

The percentages are run in the

direction of what seemed to be the casual factor.

47

A. TESTING THE MAJOR HYPOTHESES The tables and Interpretations presented in the following section show the association between the items an the R. R. Scale and the major criteria.

Five of the six major hypotheses are tested by the chi-

square technique while one criterion is compared to the R. R. Scale in ratio form.

If the association was shown to be other than chance by

the chi-square technique, a contingency coefficient was computed to show the amount of relationship between the item and criterion.

No suitable

test of significance could be found for the data in ratio form. 1. A Subjective Numerical Rating for the Past Year's Work Supervisory personnel from NGRG give each interviewer a subjective rating on each survey in which she participates.

The average rating

obtained during the year preceding this study was compared to ten items on the R. R. Scale and in no case was a chi-square value obtained which indicated a significant relationship.

A comparison of every R. R.

Scale item with the average numerical rating for the past year’s work can be found in Table A, Appendix B, pages I46-I50. 2. The Actual Number of Surveys for NORC in Which Interviewers Parti­ cipated One item from the R. R. Scale showed significant association with the number of surveys in which interviewers participated.

All of the

other items on the scale did not show a significant relationship.

A

comparison of every R. R. Scale item with the number of surveys in which interviewers participated can be found in Table B, Appendix B, pages 151-157. Whether the interviewer went right into the su rve y without telling the respondent much of what the survey was about or whether he took a

lot of time to explain the purpose of the survey showed a significant association with the number of surveys in which interviewers participated. Table I shows the test for the significance of this relationship. Table I Relation between the Number of Surveys for NORC in Which Interviewers Participated and Respondents' Indications of Whether Interviewer Took a Lot of Time to Explain Purpose of Survey R. R. Scale Item 7. Respondents' indications of whether interviewer took time to explain purpose of survey

Number of Surveys

0

100 or 50-9( p) overTotal 10-19 20-29 30-49

1-3

4& 5

6-9

Want right into questions

.19 .40 /

1.26

2.08

5.30

.79

3.40 /

.12

.3* 713

Took a lot of time to explain purpose

.47 .53

.26

.86

5.19 /

.02

4.84

.00

.05 440

No answer on R. R. Scale

.06 .02

3.04 /

2.95

1.52 /

5.60

.08

60

81

185

128

Total

198 X2’j it..

-;h

One of your ballots has boon stnmocu 'i'ntarviedcr *o f.nootioi naira 1. 1'lcn.so fill out this ou estionnc i re with your own opinions ri; ht after you finish reafin thou a specifiontiono for the first time — before on oh tain arr, inf A r, ;;. 'drf it in thu ar.clo ijJ af. .• ’■a-’...i-.ir.:sr»'nvulon::, to the .;■ fori: of 'iur.' 1:or a.li.u Lely.

and a il i !. back

1 Kb "ij..*» an Ifii'• co a.i;rya "at,

•or.a ■o',i. "aprp."cb

hu t-~.

I:--on r a.f-'la tally on your

’uola shoot of any rer—

an to ,r^nt you. an i r . l . a .

At f!’: -r.d if 2rour

-pr--..,yf .o 1.a 1 any a” '*’' i'ci .!i ~n r 't'O1' i.'a•* total nur v.^:r of refuse In ‘/ou •-.co\ ’it in f 1' o;ic.ce . •,c:

Do not ask this of f a n residents, but do ask it of rural non-farm and urban residents. Simply enter the appropriate amount on the appropriate 3.'no,depending upon v;hether tlie respondent owns or rents his hone, bo remainder of the Factual Data can be filled out from your own observations, so this point hand the respondent the pi.nl: Survey 283 form in its envelope, and while he is answering those questions, complete your recordinp of the Survey 202 Factual Data, at

I TDD 9 : Tlie I etronoiitnn Districts with populations of more than one million are the following: be’.: York, Doc ton, Philadelphia, Pittsburgh, Chicago, Detroit, Cleveland, St. Louis, Baltimore, Lor Angeles and San Francisco. All sub­ urban and central city interviev;s within those eleven D.l.'s should be circled Code 1. All other metropolitan District interviews, city and suburban, should be cir­ cled Code ?, ITED 1 0 :

Describe tlie respondent's hor.c address as completely as nossible. If he lives in an apartment or otlv. r multi-family house, list hie apartment . umber or describe its location ("2nd floor-rear" or v/hatever ). Tn rural areas, ••!vo the name or number of the read, the nostril route, or other such idontif ication. T iPD 1 2 :

Fill in here the tine the interviev: itself ended — that is, the current time — not the time you 1< eve the house after the respondent fills out Dio ;Dnl form.

Di'DDD 1J:—1 [>:

Circle the one code v/hicl. yen fool best describes the rosnondont’s cooperativcness and hjc frankness, Daturally, you can't Do certain -A the honesty with which he answered, but let us have your imprcssi on. If you ’:d. flu: fooling that his answers did not always represent his true opinion, don't sit: t o to say so. "D" 1. -1 :

71lose items are roughly parallel to some of those v:c ask the respond­ ent on Survey 283. Again, just try to circle the one code that - ■•■vs closest, even though the phrase, nay not describe the situation exactly. ' '' i t :

"Sequence Dumber" refers to the succession of interviews in your quota. Dark, your first interview "1", your second interview "2", etc. If you ok wild Dave two quotas, number teem sonaratcly, not consecutively. This same soqui nee number, along with the respondent's address, should; be written on the env lope containing the respondent's pink form.

Ml CPSCIFICATICNS FOP. 317-7"/ 5:3''-2B3 - 10.

2 " . 73 20:

All pink forms which are returned to you, filled out by the respondent, ’ should be c i r c l e d cither Code 1 or Code 2, depending upon whether the respondent left the envelope sealed or unsealed, /my forms not returned will fit one of Codes 3, I* or $, dither the respondent insists on mailing it himself (Code 3) or he v;as unable to fill out the form because of illiteracy or some other reason (Code U), or he refused to fill out the form (Code 5>).

THINGS TO PEPS.TSR 07 T7IS ASSIGNMENT; Fill out your Interviewer’s Questionnaire and return it to the Pei.' Ycrk office right now. Keep a tally on your Quota Sheet of any persons who refuse to be interviewed. Put the Survey 2o3 forms in their envelopes before handing them out to respondents. Record tlie tine of day each interviev/ began and ended. Fake a separate estimate of the time you spend on Survey 283. Record the sequence number and respondent’s address on each Survey 2o3 envelope. Return all materials except these specifications, the card used on Q. 1U and any extra Survey 283 forms or envelopes. Return everything else.

■» PAUL B. SIT3ATSLEY Eastern Representative

SURVEY 232

16. (CONTITJED) //C. In what way do you think these charges have been a (pood) (bad) thing for the country?

o 4

;)'D. Can you think of any way they have been (pood)- (bad) for the country? (I? "YES") In what way?

36-

FACT’ JA'

1. RESPONDENT'S OCCUPATION f.Ti STATUS:

DATA ?, In politics today, do yon consider yourself a Democrat, or- a Icpv.biican, or do you favor some other party?

Job-:

Democrat. . . . . . . . IjG-l Republican............. 2 Other (seeelfy) . . . . 3 Independent . . . . . . h Don't know, Ncne. . . . ,5

Industry': 37(If respondent is not the main earner in the family, specify occupation of b re adwi nne r be lew. )

U.

-hat is pour arerodimaie ape?

Ll-

Job: p. dtBMT: (Ct.it for 'h m

respondents)

Industry: 38-

Contract rent (if renter)

__

Estimated rent (if ov.Tier)________

2. hat was- the name of the last school you attended?

HAND RESPONDENT ENVELOPE CONTAINING PI! K FORK

'hat was the last grade (or year) you completed in that school?

6 * SEX:

Completed college. . . . Some college . . . . . . Completed high school. , Some high school . . . . Completed grammar school Seme grammar school. * . No schooling . . . . . .

Hale . . . . • U2-1 Female . . . . 2

39-1 • • • • • •

2 3 U 5

6 7

7. ECONOMIC LEVEL: A .............. J'3-l .................’ B. •........... 2 C ............. 3 D ............. h

SURVEY 202

“ 6

1

9

16. Eon much did you yourself enjoy

8. RACE: V.'hite •

this interview?

Ui-i 2

Colored

Didn' t enjoy it at nil . . #. rryl Enjoyed it very little . . ■, 2 Enjoyed it somewhat..... 3 Enjoyed it very j.n.ch . . . . . !;

9. SIZE Of TC.JU AY l 'V', RESPONDENT LIVES: i:.D. over 1,000,000 . I.E. under 1,000,000. c ,l)G0 to .CO, 000 . . . Drier 2,000 (KHF) . .

. . . .

.U5-3'17. Circle whichever of vi.o-o t'.o statements cor.es d c s e s t :

1s.m

i Respondent and T t^lken ah ml. otb.er th.incs from timo to time durir.r the interview. . . . . . '"’.-I Respc ,;deni ar.d I didn't telle much at out anything except the survey...................... 2

10. ROSFOFrEIIT *S HOLE ADDRESS:

(If no re_ than ono f aj•>ily __lives it above address, describe unit or list apartment nunLer)

iI?,

Circle whichever of these two staoerento cores closes I: I 'V'r.t rirht into th? m e n t i o n s with out m e t e::plann tion. . , . 5&-1 T tool-: some lino to e cplain viie : purpose of the survey . . . . . 2

II. PLACE ATD E'TATE: h 712. T H E .OF DAY INTERVIEW ENDED :

: > r

119.

i,Li f:

CF III Li f.i1.:f■.VI:

U921. Circle one of these fiver

13. DATE CF INTERVIEW:

5o . was cooperative? Very cocoorative • . . Rather cooperative . . Ea tl er un 'oopera1.1ve . Very,- uncooperative , . Can't . . . . . . .

; *

















5l-l 2 -> • n* • • J

,



'» > ?

Re soondr nt ft 1 led o';t •ni nV r.irn and returned .it to rr^ uncr - Led. 57-1 Resoondert filled out ni'.i: ..Earn 2 a n 1 returned it to me seal Od‘. . fo;m kes ponder, t filled out c m sal i ho would rail i-1- . . . 3 Respondent unablv to f i n out ■f pink form ......... « • • . . . pes” onlent refused to fill out ** pink form . . . . . • * * • • •

Ilyl’EAvEL E R 'S CIGNAILRE: 15. Did you feel that th piil ': f c m : X J *

M3

CODES FOR PUNCHING DATA ON IB! CARDS Column Number 1-36 37-60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80

Questionnaire data Factual data Columns 61 through 74 contain data about interviewers Occupation Occupation Educational level Age Sax Economic level Numerical rating for past year's work Number of surveys used for basis of numerical rating NORC Rating Rank Group Number of surveys for NORC in which interviewer participated NORC Service Rank Group T^rpe of previous interviewing experience Amount of previous interviewing experience Refusal Rate Columns 75 through 80 contain respondents' answers to R. R. Scale R. R. Scale items 1 and 2 R. R. Scale items 3 and 4 R. R. Scale item 5 R. R. Scale items 6 and 7 R. R. Scale items 8 and 9 R. R. Scale item 10, and "Do you expect war...." question

Codes for measures of interviewer1s work Numerical rating for past year's work Rating Code number 5.00-4.01 5 4.00-3.89 4 3.88-3.50 3 3.49-3.00 2 2.99 or less 1 No answer 0

Col. 67.

Refusal Rate code Rate 0/33-0/15 0/14-0/1 1-8 10-14 15-19 20-24 25-34 35-44 45-59 60-91 100 or over No answer

Number of Surveys for NORC in which interviewer participated Number Code number 0 None 1 1 2 2 and 3 4 and 5 3 6-9 4 10-19 5 6 20-29 7 30-49 8 50-99 100 or over 9

Col. 74 Code number X Y 1 2 3 4 5 6 7 8 9 0

APPENDIX B CHI-SQUARE IABIES

The following tables show the obtained and expected frequency for each cell, the degrees of freedom (df), and the chi-square value (X^) for each table.

When the chi-square is significant at the .05 level

or better, a contingency coefficient (C) is also shown. In each cell the expected frequency is shown in parenthesis.

In

all chi-square tables in this report the expected frequencies are "forced" to total the exact obtained frequencies row total.

Normal

rounding procedures were followed except when the expected frequencies row total did not equal the obtained frequencies row total.

When the

expected row total was less than the obtained row total, the number which was nearest the rounding up point was raised (i.e., 1.649 raised to 1.7).

If the expected row total was more than the obtained row total,

the number nearest the point of not being raised was not rounded upward (i.e., 1.651 kept as 1.6). Whenever the chi-square value indicates a relationship significant at the .05 level or better, the table is also shown in the results chapter.

In the results chapter the table shows the cell square con­

tingency value for each cell instead of the obtained and expected frequencies.

The table number and page number for corresponding tables

in the results chapter are indicated in parentheses beneath each ap­ pendix table which has a significant chi-square value.

146

Table A Obtained and Expected Frequencies Between Interviewers' Numerical Rating? for Past Y ear and Respondents 1 A nglers to T en Items from. the R . R. Scale

R. R. Scale Item 1. Respondents' enjoymenl of interview

numerical Rating for Past Year 3.503.09- A. 01- Ho 3.So 4.00 5.00 rating Total

2.99 3.00or less 3.49

Enjoyed it very much

110 103 59 209 139 659 34 (53.2) (126.0) (121.4) (2 1 0 .2 ) (39.2) (109.0)

Enjoyed it somewhat

25 (30.5)

53 53 S3 129 25 378 (72.3) i (69.6) (120.6 ) (22 .5) (62.5)

T D i d n ’t enjoy it at all Hnjoyed it very little

Ho answer on R. R. Scale

Total

| 9 ! (12.0)

10 (7.3) 103

X2=22.12

31 19 49 ! 10 (28.5)! (27.5)j (47.5) 1 (S.9)

20 (17.2)

20 19 (16 .6 ): (23.7)

244

235

7 (5.3)

(14.9)j

76

211

00

14

"

Jl2 76

G=Unreliable

df=23

R. R. Scale Item 2. Respondents' indications of how long Interview 2.99 or less toolc

407

31 149 (24 .6 )

Numeric*3I Rating for Past Year

3 .00-

3.503.83

3.S9-

3.49

5.00

No rating

Hardly any time at all

13 (21 .1)

47 (50.1)

59 (43.3)

97 (99.2)

41 (43.3)

Very little time

36 (37.6)

96 (89.1)

76 (35.8)

194 (176.4)

(77.1)

A fair amount of time

31 (30.7)

73 (72.9)

74 (70.2)

135 (147, .2)

68 (63.0)

A great deal of time

5 (5.3)

9 (13.3)

6 (13.3)

29 (27.2)

23 (11.9)

No answer on R. R. Scale

13 (7.7)

19 (13.2)

20 (17.5)

28 (35.9)

15 (15.7)

247,

235

433

211

Total

262

64

466 331 ; |

?2

..

Total

103

X2»24.18

df=24

Grllnrcliable

95 1276

U 7

Table A (continued)

R. R. Scale Item 3. Respondents' indications of usefulness of opinion 2.99 3.00surveys Dr less 3.49

Numerical Rating for Past Year 3.503.88

3.894.00

4.01No 5.00 rating

Total

Very useful

56 114 133 " 235 (56.1) (132.9) (128.0) (221.7)

122 35 (41.4) (114.9)

695

Someviiat useful

28 (29.3)

27 55 (21.6) (60.0)

363

70 (69.4)

68 115 (66 .9 ) (115.8)

Not very useful Not useful at all

6 (9.1)

38 13 (21.4) (20.6)

No answer on R, R. Scale

13 (8.6)

22 (20.3)

Of 21 (19.5) (33.8)

244

235

Total _

103 __________

..

..

.

X 2 =30.15

R. R. Scale Item 4Respondents' indications of suitability of time for interview

_j

...

_____

df=23

33 (35.7)

16 (18.5)

112

8 IS (6.3) (17.5)

106

76

407 .

6 (6.7)

211

1276

______

C=Unreliable

Numerical Rating for Past Year 3.002.99 Dr less 3.49

3.503.88

4.01No 5.00 rating

3.894.00

Total

Very good time

146 57 140 239 (60 .4 ) (143.0) (137.8) (238.6)

35 131 (44-5) (123.7)

748

Fairly good time

28 (24.4)

302

Another time would have been somewhat better Another time would have been much better

No answer on R. R. Scale

52 (57.8)

55 (55.6)

99 (96.3)

21 (18.0)

47 (49.9)

32 (11 .0 ) (26.0)

14 = (25.0)

50 (43.4)

U (8.1)

18 (22.5)

136

20 (17.2)

20 (16.6)

19 (28.7)

6 (5.3)

15 (34.9)

90

244

235

407

8

10 (7.3) 103

Total

211 .] 76

X 2 =25.56

df=23

C=Unreliable

1

1276

14.8

Table A (continued) R. R. Scale Item 5. Respondents' indications of the way they felt during interview It was like:

Numerical Rating for Past Year

2.99 3.00or less 3.49

3.503.88

4.01No 5.00 rating

82 (72.4)

(13.5)

Total

I

3.894.00

oa H

w

Taking Intelligence test 14 49 Being on witness stand (13.4) (43.4) Having political argument

38 (37.5)

227

73 15 41 (70.5) (13 .2 ) (36.5)

221

Voting in an election

17 (17.3)

Having friendly discussior

96 102 40 159 (42.8) (101.3) (97.6) (169.1)

Answering questions on government form

No answer on P.. R. Scale

Total

10

90 34 (31 .6 ) (87.6)

10

6S (61.9)

(19.9)

23 (19.1)

25 (33.2)

(6 .2 ) (17.2)

244

235

407

76

47 (37.1)

13 (8.4)

21

(11.5)

6

26 (32.1)

16

211

530

194

10/, 1276

C=Unreliable

df=29

R. R. Scale Item 6 . Respondents' indications o f2 .99 or whether other things were discussed during interview less

44 (40.7)

33 (35.7)

(15.7)

103 X 2=23.36

31 (42.3)

11

Nuner ical Ra ting fo r Past Y’ear 3.003.49

3.503.38

3.39-

58 (46.1)

35 (79.7)

4.00

4.01-

Mo

5.00 rating 11

Total

Talked about other things from time to time

40 15 (20 .2 ) (47.3)

41 (41.3)

250

Didn't talk much about other things

I 84 73 153 59 153 294 ( 292 . 2 ) (54.6) (73.9) (175.1) ;i63.7) (151.5)

916

No answer on R. R. Scale

20 15 24 (3.9) (21 .0 ) (20.3) 103

Total

X2-15.80

df“17

244

235

(14.9)

6

23 (35.1)

(6.5)

17 (13.2)

407

76

211

C=Unreliable

110 1276

149

Table A (continued) R. R. Scale Item 7. Respondents' indications of whether interviewer took time to explain 3 .002.99 purpose of survey c>r less 3.4-9

Numerical Rating for Past Year 3.503.88

4 .01No 5.00 rating

3.89-

4.00

Total

Went right into questions

52 116 248 237 (57.6) (136.3) (131.3) (227.4)

45 115 (42.5) (117.9)

713

Took a lot of time to explain purpose

36 (35.5) . _

73 (84.1)

96 137 (SI.0 ) (140.4)

23 75 (26 .2 ) (72.S)

440

15 (9.9)

23 (23.5)

23 I 33 (22.7) (39.2);

No answer on R. R. Scale

8 21 (7.3): (20.4)

I

Total

103 X 2 =12.39

244

df=17

235 ___ L

1

407 _ 1

i

76

i 211

1

C=Unreliable

R. R. Scale Item S. Respondents' indications of whether interviewer 3 .00kept repeating questions 2.99 to get an answer c r less 3.49

Numerical Rating for Past Year 3.50-

3.89-

3.88

4.00

4 .01No 5.00 rating

Interviewer repeated questions

128 170 281 147 65 53 (68 .1 ) (161.4) (155.4) (269 .2 ) (50.3) (139.6)

Interviewer went right on to next question

23 (22.5)

(53.4)

56 (51.4)

87 (89.0)

No answer on R. R. Scale

15 (12.4)

32 28 (29 .2 ) (28.2)

39 (48.8)

103

244

Total

46

235

13 54 (16 .6 ) (46.1) ■

10 (9.1)

df=17

C=Unreliahle

29 (25.3)

211

407 76

X2 =10.01

123 . h'■— — 1276

Total 844

279

153

1276

150

Table A (continued)

R. it. Scale Item. 9. Respondents' indications 2.99 of whether they felt free or to give their first answer,less

Numerical Rating for Past Year 3.003.50- 3.89- 4.01No , 3.49 , 3.88 4 .00 5 .00 rating

Total

Pelt free to give first 45 ! 126 104 200 100 37 answer that came into mind (49.4)(117.0)(112.7) (195.2 ) (36.5) (101.2)

612

Felt that they should thinl: carefully

550

46 97 105 17s (44*4)(105.2)(101.3) (175.4)

No answer on R. R. Scale Total

12 (9.2) 103

..... .... ..

21 (21.8)

26 29 (21.0) (36.4)

i 244 235 ....._.i

407

OO

(90 .9 )

7 (6.8)

(18.8)

76

211

........ _ L

_

X'~=*6.38

df=17

R. R. Scale Item 10. 3.88 or Respondents ' estimations of interviewers1 opinion s less

n o

(32.8)

19

114 1276

.

■ .......J

C=Unreliable

Numerical Rating for Pc1st Year No 3.895.00 rating Total

Interviewers1 opinions similar to theirs

108 (107.2)

(89.0)

40 (38.8)

Interviewers1 opinions different than theirs

11 (11.4)

10 (9.5)

4 (4.1)

235

25

' 336 (325.9)

146 (142.4 )

861

84 (70.7)

21 (25.6)

155

m m

No answer on R. R. Scale

379 (392.7)

0 to

Interviewers didn't have opinions

1

— ....

582

Total

483

J X 2=5.62

211

1276

.j

; 1

df-11

CsUnreliable

4

Table B Obtained and Escpected Frequencies Between Humber of Surveys Tor 11PRC in which Interviewers Participated and Respondents 1 Answers to Ten Items from the R. R. Scale

R. R. Scale Item 1. Respondents’ enjoyrnent of interview

Number of Survevo

50