Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
APPENDIX B SAMPLING METHODOLOGY This section includes a detailed account of the various deci- sions involved and the actual steps taken in selecting the sample and in assigning respondents to interviewers. A representative sample of the Port Jervis population was required in order to get an accurate picture of how city residents reacted to the flood and threat of disas- ter, as well as to test the various hypotheses which are included in the main report. A. Possible Sources for Sample Selection In the course of the preliminary survey, a number of possible sources of sampling information were discovered. These were (~) a complete listing of all Port Jervis residences using the municipal water supply, (2) the list of all registered voters in the last municipal ele ction, ~ 3) a map of the city which contained contemporary informa- tion on zoning regulations and dwelling-units, and (4) a recent issue of the city directory. Despite our interest in the non-residents (tourists) to the false report, city officials, and various respondents, assured us that all tourists had departed during the early phases of Me emergency. Because of this, no attempt was made to include non-residents in our sample . The municipal water records were seriously contemplated as a reasonably accurate and up-to-date source of resident sample data. The chief reason for their abandonment was that each residence was listed under the owner's name. In instances in which a number of in- dividuals were living in a residence, the interviewer would have had difficulty getting the correct respondent. This would not have been a severe handicap except that the arrangement of the records did not permit a simple discrimination between single and multiple dwelling- units. The people in the Water Department suggested some rules-of- thumb for deciding whether or not the residence was a multiple dwell- ing -unit, but the e stimation p roc e s s would have tripled or quadrupled the time involved in selecting respondents. 91
The second source we considered, the roster of registered voters, was discarded because of its typical lack of representative- ness. The third source, the city ~nap, was found to be completely inadequate for our purposes. Although the zones and zoning regula- tions seemed to be indicated clearly, there was no means of deter- mining the number of dwelling-units in any given area of the city. The map s we re the refor e used only as a means of indicating the lo ca- ti on of a r e spondent' s dwelling -unit onc e he had be en s ele cte d for the sample. The fourth source, the city directory, was finally selected for sampling information. According to officials in the City Clerk's Office and in the municipal water company office, the publisher of the directory had canvassed the city during the early part of 1955. The dir e cto ry was publi she d about Mar ch of 1 9 5 5. Information from the aforementioned offices indicated that minimum of individuals moved in and out of the city. This, coupled with the fact that 1 95 0 census data indicated an unusually large propor - tion of older people in the city, gave the investigators some assurance that they were working in a community with a relatively stable popula- tion. However, the use of this directory involved a calculated risk, since it was at least eight months out of date. Despite this, it seemed the most accurate source of information available and the sample was selected from the listings it contained. B. Selection of the City-Wide Sample The city directory had the usual lists: an alphabetical list of adults, and a street-by-street list of residences by address. Both of these lists were used in the selection of the sample. There were only approximately 5, 300 listings in the alpha- betical list because wives whose husbands were living were noted parenthetically after the listing for the husband, and because children were not listed at all. It was decided that a minimum of a 1 per cent sample of the total population of about 9, 700 residents was needed. Accordingly, 110 random numbers under 5, 300 were selected from a table of random numbers. 92
The numbers were arrange] in order of magnitude and mem- bers of the sample were selected from the alphabetical list by count- ing through until a given random number was reached. For example, suppose that the first random number is twenty-five. The investigator counted through the alphabetical list from its beginning, omitting all business establishments and names of individuals living outside the city limits, until he reached the twenty-fifth resident's name. This name and address was then copied down as the first member of the s ample . If the s e cond r ando ~ ~ ~ numb er was twenty - eight, the inve sti - gator would count down three more names, and copy this name for the second member of the sample. The procedure was continued until all 110 members of the city-wi(Le sample had been selected. As a proportional representation by sex was desired, the s ex of the r e spondent in a given dwelling -unit was also determined when the name was drawn. If the nth listing included the name of only one man or woman, that name was included in the sample. If, on the other hand, both a man and his wife were listed, the wife was selected if the random number was even, the husband if the number was odd. We anticipated that a number of residents would be unavail- able for the period of the study. Therefore an additional group of fifty random numbers was chosen so that replacements for the sample . . ~ . ~ . ~ ~ ~ could be selected at random. These numbers were kept in the order in which they were drawn from the table of random numbers. Thus, an individual assigned a very high random number might be the first one selected as a replacement for the sample. C . Selection of the Saturation Sample Only a small proportion of the city-wide sample lived in the flooded areas. Because some of the hypotheses in the research pro- posal involved a comparison between individuals who had evacuated as a function of Friday's flood and those who had not, it was necessary to increase the number of respondents in the flooded] portions of the city. Accordingly, an additional sample was drawn in those areas of the city which had been flooded. _ This was done by s electing forty new random number s, which were then arranged in order from the lowest to the highest. It was not possible to use the alphabetical list which was used for the city 93 -
wide sample, because this list included individuals living anywhere in the city. Instead, we first determined which residences were actually in the flooded areas. Those residences which were on streets which hac] been flooded were marked in the street section of the directory. The investigator then counted from the beginning of each street to its end (or to the end of the flooded area) until he came to the random number assigned to a particular respondent. This process was con- tinued until forty residents had been selected. An additional twenty numbers were selected from the table and were arranged in the order in which they were selected. These were kept on reserve for what- ever replacements might be needed. As in the city-wide sample, the interviewer asked for the woman of the house if the random number was even, the man of the house if the number was odd. D. Assignment of Respondents to Interviewers When the interviewer was ready for an assignment, he re- ported to the individual in charge of sampling, who gave him a street- map of Port Je rvi s containing hi s as signment for that day. The name and sex of each respondent and the approximate location of his dwell- ing-unit were indicated on the map. An assignment usually included about a half-dozen respondents, all located in a restricted geograph- ical area to minimize travel cost and time between interviews. How- ever, the use of this method raised the possibility of-introducing some bias. If interviewers worked only in a particular area, differ- ence s betwe en are as could po s sibly be attributable to inte rviewe r differences, rather than to the respondents' socio-economic status, experiences with the flood, interpretation of the false report, etc. In order to circumvent this problem partially, interviewers were rotated from one portion of the city to another whenever possible. Having received his assignment, the interviewer attempted to locate and interview each of his respondents. Upon his return to the office, he reported on which interviews had been completed, which re- spondents were unavailable but would be contacted later, and which were completely inaccessible for the period set for the study. Where respondents were inaccessible, replacements were assigned to the interviewer. A check on all call-backs was made periodically to in- sure that the interviewers were doing their best to obtain the respondent indicated. 94
Additional checks on the performance of interviewers were made to make sure that no one was falsifying data. Periodically, two supervisors would contact several of the respondents who were re- ported to have been interviewed. If this contact was made by telephone, the supervisor would introduce himself and ask for an appointment for an interview. If contact was made in person (at the respondent's home) the interview was begun. When the supervisor was informed that an individual from the Institute hac] already completed the inter- view, contact was quickly terminated with an apology, and an explana- tion to the effe ct that two indivi dual s we re customarily as signed the same list and that the other person had obviously already reached the respondent. art no case clid the supervisors final that an interviewer hac] falsely reported a completed interview. E. Replacements All interviewers were instructed to make at least three call- backs before asking for a replacement. In some cases as many as six call-backs were made before a replacement was used. Specific- ally, the interviewer s were given the following instructions: 111. Your respondents will be assigned to you by name, address and sex. Go to that address and try to find the person designated. (Make sure that the complete name is correct as there may be a number of people in the house with the same last name. ~ If the respondent is not at home, try to make an appoint- ment for later on. If no one is at home, call back at a later time. In any event, make at least three call-backs, spaced far enough apart so that the respondent has a chance to return. If unsuccessful then, call Dr. Thayer. If the respondent has moved out of town, take the person (preferably of the same sex) who is now living in his old quarters. If the respondent has moved, but to some other determinable place in town, go to that address and interview him. (If this is at some distance from your area, get your other interviews in the area first. 95
4. If the respondent is clearly inaccessible for the dura- tion of the study, take the respondent in that dwelling- unit who is of Me same sex as the clesignatecl person. If the re is no one the re of that sex, s e e Dr . Thayer about a replacement. " In summary, the interviewers were told to take replacements only if the re spondent de signaled was clearly inacce s sible. Under this condition, he was to take a replacement of the original respondent's sex, who lived in the designated dwelling-unit. If no such person lived there, he was to get a replacement from the list of random num- bers referred to above. The proportion of r eplacements needed for the city sample ran higher (21 per cent) than for the saturation sample (S per cent). Althoughitis impossible to accountfor this difference with the data available, a few hypotheses may be considered: (1) The people in the saturation sample were more likely to be involved in repairing dam- age to their home s and therefore were tempo rarity le s s mobile . (2) The people in this area were less well-off financially both as a nor- mal condition and as a function of the flood losses and therefore were le s s mobile . Other po s sibilitie s, of cour s e, exi st. An analysis of the reasons for respondents' lack of availa- bility reveals no consistent trends. In a few cases, the respondent was unknown, either because he had moved since the directory had been prepared and was therefore unknown to the new tenant, or the directory was in error. A few had been drafted, or had recently mar- ried and were no longer living in Port Jervis. One was a mental case and another did not speak English. A few railroad men were away on their jobs. A few people were out of town on vacation or business. Finally, some were not available after as many as six call-backs for unknown reasons. F. Additional Factors Considered . During the planning phase of this study, the investigators seriously considered the po s sibility of contacting re spondents by tele - phone before the interview. It was hoped that making an appointment would simplify the task of the interviewer, cut down the number of call-backs, etc. This plan was abandoned because many of the 96
respondents did not have telephones. Those who were telephone sub- scribers could, of course, have been called. However, if all individu- als having telephones had been contacted or forewarned, while those not having telephones had received no notification, we might have introduced a systematic error into our data-collection procedures. The extent of this error would have been impossible to deter- mine. When cloaking socio-economic comparisons, we might have found marked differences in the responses of upper versus lower- clas s re spondents . Lower - clas s individuals are generally tho s e who do not have telephones, so that differences observed in the responses might have been attributable to the forewarning rather than to socio- e cono colic factor s. A previous study by Nuckols (15) involving the forewarning of respondents indicated that this had no effect upon the responses of the members of a sample. However, this survey dealt with materi- als which involved little or no defensiveness or ego-involvement. In addition, the forewarning was clone through the mail. The investi- gators felt that it would have been extremely unwise to attempt to generalize from this one quite different study, ant] so clecidecl not to forewarn respondents. G. Empirical Checks Although our sampling technique s are de signed to maximize the probability of representativeness, it is still incumbent upon the investigator to offer evidence to support his contention that the sample is representative. In the usual market or opinion study, comparisons of the demographic characteristics of the sample and census data are frequently made. However, following a disaster, the mortality, hospitalization, or movement (to endamaged or healthier areas) of inhabitants may leave a study population whose demographic character- istics are no longer congruent with those originally surveyed] in the Census. Moreover, Census data becomes rapidly outdated in our mobile society. It seems, therefore, that investigators should run checks on the internal consistency of the data as a measure of reliability, and should also check the validity of respondent reports by comparing 97
them with known facts concerning the times, places, and occurrences of various events during the disaster. Some major examples of em- pirical checks on the respondent data in this study are given below. The reader will note other such checks in the body of the report. The sample contains too large a proportion of women (59. 8 per cent) because of difficulties encountered by interviewers in trying to contact male respondents. In most other respects, the sample seems to be representative of the Port Jervis population. (See also page 23. ~ When the sample is divided into two groups according to geo- graphical location within the city, we find that the discrepancy between the groups in educational level and in socio-economic status is con- sistent with what we know about neighborhoods in the community. The behavior of the sample conforms to that which would be predicted from the descriptions of officials who were at the scene of the disaster. The official report of the Police Chief estimated that between 2, 500 and 3, 000 persons evacuated as a result of the false report. When the number evacuating is estimated from the sample data, the range is the same. Information from the key interviews leacis us to believe that the fire-truck which disseminates} the rumor was broadcasting for a few minutes around 11:15 that night. Eleven-thirty p. m. is the median time indUcated by respondents for receipt of the false report. Resi- dents in the flooded area, in which that fire-truck was broadcasting, are Dose who report hearing the threat from an official source. We also know from the reports of Police, Fire Department, and the local radio station, WDLC, the time and contents of the denial me s sage s which the se agencie s broadcast. The re spondent data are accurate with respect to this information in at least the following ways: At approximately ~ 1: 3 0 - 1 1: 45, official center s were attempt- ing to verify Me report. At this time, the fire -truck loudspeaker s gave a short public message which, in effect, told people that an at- tempt to verify was in progress, and that they thought the report was false. Only a few minutes later, denial was received from officials at Wallenpaupack Dam, ant] a le s s ambiguous `~'e s sage was s ent out. 98
Whenever respondents reported hearing this ambiguous message, it was invariably reported as the first denial message they heard. 11:40 and 12:30 The fire-truck loudspeakers broadcast the denial between Fire-truck messages tend to be reported as first and second messages received. On the other hand, the radio station did not go back on the air until 12:10 that night. Radio messages tend to be reported as second and third messages. Despite the fact that the total number of messages decreases between first and second denials heard, the total number of radio messages increases. Our empirical checks, therefore, indicate that the sample is representative with respect to: education, socio-economic status, flight, rumor dissemination, and denial dissemination. Only with regard to sex distribution did we find evidence of non-representative- ness. 99
KEY RESPONDENTS 1. Police Department a. Chief b. Radio dispatcher 2. Civil Defense a. Director b. Relief Director c. Radio op erato r 3. Sparrowbush Fire Department a. Chief b. Captain c. Radio operator d. Volunteer fireman involved in spread of rumor 4. Port Jervis Fire Department a. Chief b. Chief radio operator Assistant radio operator 5. WDLC - local radio broadcasting company a. Program Director b. Two announcers 6. Newspaper - Union Gazette a. Editor b. City Editor Publi she r 7 . Other Po rt Je rvi s Official s ~ . . ~ . _ a. Mayor b. Alderman 8. Others in Communication Centers a. Manager of the telephone company b. A local ham radio operator c. Erie Railroad communication center 9. Other individuals in important community positions a. Director of a light and power company b. Priest in a Port Jervis church Di r e c for of a P art Je rvi s ho spital d. A ho spital doctor and nur s e s e. YMCA night clerk f. Manager of a Port Jervis the atre Bartencler at a Port Jervis hotel 100 3 operators
