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.
7 Analysis of Response Rates Table 3 and Table 4 show the response rates obtained for each airport community for the mail and telephone surveys, respectively. The denominator used for each response rate is the number of addresses selected for the sample minus the postal non-deliverables (PNDs). The overall response rate for the mail survey is 35.1%, and the overall response rate for the telephone survey is 12.1%. These response rates are highly significantly different (Cochran-Mantel-Haenszel Ï2 = 455, p-value < 0.0001).8 Table 5 gives the number of completed interviews obtained, and Figure 11 displays the response rates for each DNL noise stratum by survey mode and airport. The response rate for the telephone survey was higher for the addresses with a matching telephone number (16.4%) than those without a matching telephone number (9.1%). For an address without a matching telephone number to be counted as a respondent, first the household had to respond to the request on the mail screener to provide a telephone number, and then the selected person had to respond to the telephone survey. Table 3 Response Rates for ACRP Mail Survey Mail Survey Number of Addresses Number of Addresses Less Postal Non- deliverables Number of Questionnaires Received Mail Response Rate Airport 1 1098 998 392 39.4% Airport 2 1100 1043 388 37.2% Airport 3 1095 1052 304 28.9% Overall mail 3293 3093 1084 35.1% 8 The Cochran-Mantel-Haenszel statistic is used to test association between two variables in stratified data (Simonoff, 2003, p. 317). In this analysis, it is used to test whether being a respondent (yes, no) is related to the survey mode (mail, telephone), after accounting for the airport and DNL noise stratum. The p-value, calculated from the value of the chi-square statistic, is the probability that you would obtain a value of the test statistic at least as large as the one observed from the data if the null hypothesis (that the response rates are the same for mail and telephone) is true. The smaller the p-value, the more evidence there is against the null hypothesis of no association between the two variables. Typically, a p-value less than 0.05 is considered to give a âstatistically significantâ result. 21
Table 4 Response Rates for ACRP Telephone Survey Telephone Survey Number Of Addresses Number of Addresses Less Postal Non- deliverables Number Of Households providing telephone numbers from screener Response rate for request for telephone number on Screener Number of Telephone Interviews Telephone Response Rate Airport 1, all 1098 997 144 14.4% Matched telephone number 369 365 79 21.6% Unmatched telephone number 729 632 140 22.2% 65 10.3% Long screener 365 318 90 28.3% 38 11.9% Short screener 364 314 50 15.9% 27 8.6% Airport 2, all 1100 1021 114 11.2% Matched telephone number 338 334 51 15.3% Unmatched telephone number 762 687 171 24.9% 63 9.2% Long screener 381 346 104 30.1% 27 7.8% Short screener 381 341 67 19.6% 36 10.6% Airport 3, all 1095 1046 114 10.9% Matched telephone number 594 588 81 13.8% Unmatched telephone number, long screener 501 458 99 21.6% 33 7.2% All airports, all 3293 3064 372 12.1% Matched telephone number 1301 1287 211 16.4% Unmatched telephone number 1992 1777 410 23.1% 161 9.1% Long screener 1247 1122 293 26.1% 98 8.7% Short screener 745 655 117 17.9% 63 9.6% NOTE: The matched telephone numbers in the table refer to those for which an initial match was found in the directory. Some of these numbers were later discovered to be invalid, and those addresses were subsequently mailed a screener requesting the telephone number. 22
Table 5 Number of Completed Interviews, by DNL noise stratum and Survey Mode DNL noise stratum Total 50-55 55-60 60-65 65-70 70+ Airport 1, all 170 185 181 536 Telephone 49 44 51 144 Mail 121 141 130 392 Airport 2, all 105 90 91 116 100 502 Telephone 25 19 23 22 25 114 Mail 80 71 68 94 75 388 Airport 3, all 63 64 92 199 418 Telephone 21 14 26 53 114 Mail 42 50 66 146 304 All airports 338 339 364 315 100 1456 Telephone 95 77 100 75 25 372 Mail 243 262 264 240 75 1084 Figure 11 Response Rates by Airport and Mode of Survey 23
7.1 Comparison of Sample with Population Demographic Quantities Demographic information was available from the 2010 Census for the census blocks containing the sampled households from each airport. The DNL noise strata used for sampling have irregular shapes because of the different noise exposures for households in the landing, takeoff, and sideline areas. The strata therefore do not match exactly with the census blocks; however, we can compare the demographic compositions of the set of census blocks containing the sampled regions surrounding each airport with the estimated demographic composition from the sample. Two sets of sample statistics are presented in Table 6and Table 7. The first set in Table 6 referred to as âunweightedâ estimates, subtracts the Census percentages from the percentages of respondents who fall in each demographic category. Thus, the percentage of respondents to the telephone survey at Airport 1 who are white non-Hispanic is 20.1 percentage points higher than the percentage of white non-Hispanics in the area from the 2010 Census. The second set of statistics in Table 7, the âweightedâ estimates, account for the disproportionate sampling done in each airport community, where a higher fraction of the population is sampled in areas with high noise exposures than in areas with low noise exposures. We present both sets of estimates because the Census figures are available only approximately for the sampled area as a whole, and not for the individual DNL noise strata. Both sets of estimates show the same pattern. Persons who are white non-Hispanic or are age 50+ are overrepresented among the respondents to the survey, while Hispanics are underrepresented. In the telephone survey, the percentage of persons age 50+ exceeded the true percentage in the area by 20 to 40 percentage points. Note, however, that although both mail and telephone surveys had deviations from the Census figures, the mail survey percentages are usually closer to the Census percentages. Table 6 Difference between percentage estimated from sample and percentage from 2010 census (calculated without survey weights) Airport 1 Airport 2 Airport 3 Telephone Mail Telephone Mail Telephone Mail Percent white non-Hispanic 20.1 (2.9) 22.1 (1.6) 25.2 (4.0) 19.6 (2.2) 30.2 (2.9) 24.7 (2.1) Percent male -5.2 (4.1) 1.9 (2.5) -8.5 (4.6) -1.6 (2.5) -7.9 (4.7) -6.3 (2.8) Percent age 50+ 39.4 (3.4) 28.1 (2.4) 26.1 (4.7) 12.4 (2.5) 30.7 (4.3) 22.5 (2.8) Percent Hispanic -13.2 (1.4) -11.3 (1.0) -14.2 (2.9) -7.8 (1.9) -20.5 (2.1) -18.5 (1.5) NOTE: Standard errors are in parentheses. Boldface values are statistically different from zero at the 0.05 significance level, indicating that the estimate from the survey is significantly different from the census value. Table 7 Difference between percentage estimated from sample and percentage from 2010 census, calculated using survey weights Airport 1 Airport 2 Airport 3 Telephone Mail Telephone Mail Telephone Mail Percent white non-Hispanic 21.5 (3.5) 22.1 (2.8) 25.2 (5.5) 19.6 (3.6) 21.5 (8.0) 20.0 (4.5) Percent male -5.2 (6.3) 1.9 (4.1) -8.5 (6.1) -1.6 (3.5) -1.3 (9.5) -8.4 (5.9) Percent age 50+ 39.4 (5.5) 28.1 (4.2) 26.1 (6.3) 12.4 (3.5) 19.8 (9.5) 17.3 (6.0) Percent Hispanic -13.2 (2.3) -11.3 (1.6) -14.2 (4.0) -7.8 (2.9) -20.8 (2.8) -20.1 (1.5) NOTE: Standard errors are in parentheses. Boldface values are statistically different from zero at the 0.05 significance level, indicating that the estimate from the survey is significantly different from the census value. Appendix F presents a response propensity analysis, predicting the probability that a household selected to be in the sample completed a survey. Logistic regression analyses were performed to predict the propensity to respond as a function of the characteristics known for the sample: airport, noise exposure, survey mode (mail or telephone), and demographic characteristics for census blocks from the 2010 24
Census. The results from the model are consistent with the analyses presented above: households sent the mail survey are much more likely to respond to the survey, after accounting for all other variables. In addition, households having a matching telephone number, living in a census block with a high percentage of persons aged 50 and over, or living in a census block with a low percentage of Hispanics are more likely to respond to the survey. Importantly, noise exposure, as measured by DNL, is not significantly related to the propensity to respond to the survey. 7.2 Selection of Within-household Respondent The Community Environmental Questionnaire asked the person opening the envelope to write down the number of adults (age 18 or older) in the household, and to write down the first name, nickname, or initials of the person with the next birthday. The person with the next birthday was to complete the questionnaire. This within- selection method was used to attempt to obtain a randomly selected adult within the household to complete the questionnaire: otherwise, there might be bias if the person within the household who is most (or least) annoyed by aircraft noise is consistently more likely to provide the responses to the survey. The last question on the survey, filled out by the person who completed the questionnaire, asked for the age and month of birth for each adult in the household. This question can be used to analyze the extent to which the instruction to have the person with the next birthday fill out the questionnaire was followed. Table 8 gives the results. In 856 households it appeared that the correct household member filled out the questionnaire. Among those households that had enough data to determine the birth month, this represents 86% where the correct respondent was selected. Table 8 Did the adult with the next birthday fill out the mail questionnaire? Airport 1 Airport 2 Airport 3 Total 1 adult in household 170 160 99 429 2+ adults, respondent had next birthday 146 143 138 427 2+ adults, other adult had next birthday 42 55 41 138 Insufficient information: missing # adults in HH 29 22 18 69 Insufficient information: missing birthday month 5 8 8 21 Total 392 388 304 1,084 25
