Assessing Aircraft Noise Conditions Affecting Student Learning, Volume 2: Appendices (2014)

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 G. Detailed Analysis Results G.1. Single Decibel Noise Metrics The first analyses performed to determine the estimated effects of aircraft noise, as measured by the Leq metric, on average reading and math test scores in Grades 3, 4, and 5, are shown in Table G-1. The second and third columns in this table present the results of a two-predictor analysis (with two independent variables) relating change in test scores to aircraft noise and ambient noise. As would be expected, aircraft noise is negatively associated with school achievement scores; however, the effects are small, ranging from -0.0147 standard deviations per 10 dB for Grade 5 math to -0.0251 standard deviations per 10dB for Grade 4 math. TABLE G-1 Estimates of the Effects of Aircraft Noise (Leq) on School Test Scores. Achievement Test Estimated Noise Effect* 2-Predictor Analysis Aircraft + Ambient Noise Total, Ltot Aircraft Noise, Leq Ambient Noise, Leq Reading Grade 3 -0.0160 -0.0491 -0.0726 Reading Grade 4 -0.0242 -0.0883 -0.1078 Reading Grade 5 -0.0148 -0.0369 -0.0560 Math Grade 3 -0.0173 -0.0480 -0.0700 Math Grade 4 -0.0251 -0.0882 -0.1092 Math Grade 5 -0.0147 -0.0020** -0.0243 *Effect size estimates are in units of fractions of a standard deviation for a 10dB difference in noise level. ** Not statistically significantly different from zero The effects for total noise, Ltot, the (logarithmic) sum of aircraft and ambient noise, shown in the fourth column, are three to four times larger, ranging from -0.0560 to -0.1092 respectively, but still small. Assuming that these relationships covers the range of Ltot, this translates to a 3 to 4 percent reduction in percentile rank in the state for a 10 dB increase in level for Grade 4 math and reading. The analysis revealed that ambient noise also tends to negatively affect test scores (see the third column in Table G-1), more so than aircraft noise alone, although not as much as total noise. The effect for Grade 5 math, is not statistically significantly different from zero. In conclusion, the effects of aircraft noise on test scores is generally statistically significant but small for metrics Leq and Ltot. G-1

Two additional types of aircraft noise measure were considered, namely average SEL and Lmax of the school day aircraft events, with the averages taken over events exceeding 70 and 80 dB for SEL, and 65 and 75 dB for Lmax. In these comparisons, shown in Tables G-2 and G-3, data from schools not exposed to aircraft noise or only exposed to aircraft noise below the specified threshold are not included in the analysis. TABLE G-2 Estimates of the Effects of SEL on School Test Scores Taking Ambient Noise into Account. Test Estimated Noise Effect* SEL70 SEL80 Math Grade 3 -0.0256 -0.0920 Math Grade 4 -0.0262 -0.0027 Math Grade 5 -0.0183 -0.0220 Reading Grade 3 -0.0413 -0.0645 Reading Grade 4 -0.0037 0.0260 Reading Grade 5 -0.0144 -0.0509 *Effect size estimates are in units of fractions of a standard deviation for a 10dB difference in aircraft noise level. TABLE G-3 Estimates of the Effects of Lmax on School Test Scores Taking Ambient Noise into Account. Test Estimated Noise Effect* Lmax65 Lmax75 Math Grade 3 -0.0130 -0.0117 Math Grade 4 -0.0193 -0.0170 Math Grade 5 -0.0105 -0.0094 Reading Grade 3 -0.0121 -0.0106 Reading Grade 4 -0.0187 -0.0164 Reading Grade 5 -0.0111 -0.0098 *Effect size estimates are in units of fractions of a standard deviation for a 10dB difference in aircraft noise level. Generally, the results in Table G-2 yield no systematic patterns of effects on test scores, and in fact the estimates are not significantly different from zero. The estimated effects described with the Lmax metric, shown in Table G-3, are statistically significant, but small, with the maximum decrement on the order of 1 percentile reduction in state rank for a difference of 10 dB. Although the metrics SEL and Lmax describe the average maximum levels, they provide no information as to how many events occur. G-2

G.2. Non-Decibel Metrics Two non-decibel metrics were also analyzed. The effects of one set of these, the number of aircraft noise events above a selected threshold, NA(L), occurring in an average school day, are shown in Table G-4.5 There is a definite pattern of increasing effects as the threshold level is increased from 55 to 80 dB. For example, for Grade 4 reading, the effect of 10 aircraft noise events above 60 dB would be a decrement of 0.0086 standard deviations (SD’s) in average test scores, while the effect of 10 aircraft noise events above 70 dB would be 0.0185 standard deviations. Assuming that this relationship covers the range of NA, the decrement for 50 events greater than 70 dB is about 0.10 SD’s, which translates to a 4 percentile reduction in rank in the state for Grade 4 math and reading. In the Year 2008 this number of exceedances occurred at eighty of the target schools in this study. Similarly, for 100 events greater than 70 dB the decrement would be 0.185, which roughly translates to an 8 percentile reduction in rank in the state, say from the 50th to 42nd percentile. More than 100 events per school day were noted at twenty-five of the target schools in the Year 2008. TABLE G-4 Estimates of Effects of Number of Aircraft Noise Events On School Test Scores, Taking Ambient Noise into Account. *Effect size estimates are in units of fractions of a standard deviation for 10 noise events. ** Not statistically significantly different from zero The effects of the number of aircraft noise events greater than 80 dB are not reliable or statistically significantly different from zero, due to the small sample size of schools experiencing aircraft noise at this level. The small effect noted for the NA(L) metric is perhaps not surprising. Even though the number of disturbances from aircraft noise may be a factor in lowering of test scores, the metric contains no information on how high are the exceeding levels. For example, NA(65) = 10 may mean 10 events at 66 dB or 10 events at 80 dB. As a second alternative noise metric, the total duration of aircraft noise above a threshold level, TA(L), in minutes per day, was estimated for each school exposed to aircraft noise. The results are shown in Table G-5. The pattern is similar to the pattern for number of noise events 5 Results for the alternative metrics are all based on an analysis that assumes a linear combination of effects of ambient noise and of the particular aircraft noise metric. Test Estimated Noise Effect* NA55 NA60 NA65 NA70 NA75 NA80** Math Grade 3 -0.0061 -0.0076 -0.0096 -0.0153 -0.0184 -0.0221 Math Grade 4 -0.0075 -0.0099 -0.0131 -0.0191 -0.0209 -0.0190 Math Grade 5 -0.0064 -0.0085 -0.0109 -0.0149 -0.0182 -0.0228 Reading Grade 3 -0.0048 -0.0064 -0.0088 -0.0148 -0.0196 -0.0187 Reading Grade 4 -0.0063 -0.0086 -0.0117 -0.0185 -0.0225 -0.0210 Reading Grade 5 -0.0047 -0.0061 -0.0083 -0.0138 -0.0194 -0.0244 G-3

per day, a small but statistically significant negative effect that increases with the noise level as would be expected. For example, for Grade 4 reading, 10 minutes per school day of aircraft noise above 60dB is associated with a decrement of 0.0231 standard deviations of average test scores, and for 10 minutes above 70dB per day, the decrement is 0.0641 standard deviations. Because few schools were exposed to noises above 75 dB or 80 dB, the effects for durations at these levels were not statistically significant from zero. In Tables G-4 and G-5, as well as in Table G-1, other schools in the same districts were included for comparison with airport noise measures of zero. TABLE G-5 Estimates of Effects of Duration of Aircraft Noise Events on School Test Scores, Taking Ambient Noise into Account. Test Estimated Noise Effect* TA55 TA60 TA65 TA70 TA75** TA80** Math Grade 3 -0.0118 -0.0205 -0.0355 -0.0639 -0.0923 -0.1352 Math Grade 4 -0.0140 -0.0248 -0.0407 -0.0600 -0.0648 -0.0612 Math Grade 5 -0.0134 -0.0236 -0.0402 -0.0657 -0.0871 -0.1326 Reading Grade 3 -0.0100 -0.0183 -0.0333 -0.0661 -0.0928 -0.1020 Reading Grade 4 -0.0126 -0.0231 -0.0394 -0.0641 -0.0764 -0.0744 Reading Grade 5 -0.0103 -0.0188 -0.0343 -0.0676 -0.0998 -0.1500 *Effect size estimates are in units of fractions of a standard deviation for 10 minutes above the threshold level. ** Not statistically significantly different from zero The average decrement for a TA(70) of 10 minutes is 0.065 across all tests . Thus, the effect of exposure for 20 minutes a school day at a level greater than 70 dB is a 5 percentile change in state rank, say from the 50th to the 45th percentile. In the Year 2008, only twenty-two of the target schools in this study were exposed for this time above 70 dB. A similar reduction in state rank could be expected for 25 minutes in excess of 65 dB – a level exceeded by seventy-one target schools in 2008. The TA metric has the advantage over the NA metric in that it quantifies the time that aircraft noise can be a distraction, even though the absolute levels of the individual events are not known. In conclusion, the effects of aircraft noise on test scores are generally statistically significant but small for the TA and NA noise metrics, but larger than for Leq and Ltot. G.3. Aircraft Noise Increment In the evaluation of the Leq metric for describing the effect of aircraft noise on test scores (Table G-1) it was noted that ambient noise, by itself, tends to negatively affect test scores more so than aircraft noise alone, at least according to the Leq metric. The insensitivity of this metric in relating to student test scores can maybe be understood when it is realized that the aircraft Leq for at least one-half of the target schools is less than the ambient noise level at those schools – see Figure 5-3 of Chapter 5. Thus, even though the aircraft noise peaks will nearly always exceed the G-4

ambient levels, there will be a certain degree of masking by the ambient noise for a large portion of the target schools. Taking the approach that any decrement in test scores may be related to the amount by which the total (aircraft plus ambient) noise exceeds the ambient, the analysis was repeated with the noise increment, Ltot-Lamb, as the aircraft noise measure, where Ltot is the logarithmic addition of the aircraft Leq and the ambient Leq, and Lamb is the ambient Leq. The results shown in Table G-6 indicate aircraft noise effects based on the amount of noise, in decibels, added by the aircraft operations. The effects are more than seven times as large as for Leq alone, and up to two times as large for Ltot, ranging from 0.1122 standard deviations per 10 dB for Grade 5 math to 0.1429 standard deviations per 10 dB for Grade 4 math. This corresponds to a 6 percentile reduction in rank in a state, and about a 9 percentile reduction for a difference of 15 dB. In 2008 there were about 103 and 30 of the target schools respectively with noise increments equal to or greater than 10 and 15 dB. Furthermore, there were about ten schools with an increment of 20 dB or greater, which translates to a 12 percentile reduction in state ranking. The deficits reported in Table G-6 do not increase from Grade 3 to Grade 5. That is, they do not provide evidence for the hypothesis that the effects of aircraft noise might be cumulative over the elementary Grades. TABLE G-6 Estimates of the Effects of Aircraft Noise Increment on School Test Scores. Achievement Test Estimated Noise Effect* Aircraft Noise Increment Ambient Noise, Leq Reading Grade 3 -0.1420 -0.0610 Reading Grade 4 -0.1360 -0.1031 Reading Grade 5 -0.1202 -0.0463 Math Grade 3 -0.1350 -0.0597 Math Grade 4 -0.1429 -0.1035 Math Grade 5 -0.1122 -0.0113 * Effect size estimates are in units of fractions of a standard deviation for a 10dB difference in increment noise level. Comparison schools have 0 dB increment. This stronger relationship for the estimated aircraft noise effect raises the question of whether the incremental noise level is an appropriate metric for all values of ambient noise. Does an increment of 10 dB in aircraft noise level have the same effect on test scores at ambient levels of 50 dB as it does at ambient levels of 65 dB? In fact, when the results of the analysis are separated into categories of ambient noise, the estimated effects are very similar in the ambient ranges of G-5

50 to 55 dB, 55 to 60 dB, and greater than 65 dB. At ambient levels less than 50 dB, the sample size is too small to draw definite conclusions. The distribution of incremental noise levels as a function of ambient levels for the target schools is shown in Figure G-1 where it can be seen that, with few exceptions, higher incremental levels are associated with lower ambient levels. Figure G-1. Distribution of incremental noise levels as a function of ambient level. G.4. Disadvantaged Students Disadvantaged students are those whom family, social, or economic circumstances hinder their ability to learn at school. The term “disadvantaged” is a bureaucratic term stemming from the first large federal investment in public education, Title I of the Elementary and Secondary Education Act of 1965. That law provided for funds to go to schools based on the numbers of children in poverty enrolled in a school, to be used for the students in that school who were falling behind academically (often by hiring aides from the community). The staff in the school would determine which students were falling behind academically, often based on test scores. To determine the extent to which disadvantaged student categories are affected differently by aircraft noise exposure, the analytical method was applied separately to test scores for two subgroups of students in each school, disadvantaged and non-disadvantaged, instead of to test scores for all students in the school. The pre-adjustment for school-level demographic effects is much less for the analysis of the subgroups than it was for the analyses of the total set of students in the Grade at a school. In particular, the effect of the percentage of students in poverty (i.e., eligible for free or reduced price lunch) on the school average score is, logically, greater than its effect on the scores of disadvantaged students alone. In fact, one might expect that the effect on the average scores of disadvantaged students would be nil, that their numbers merely bring down the average of the total group. That is not quite the case – the concentration of poverty students in a school is correlated with somewhat lower scores for both disadvantaged and non-disadvantaged students in the school. Since the adjustment of scores for demographics is much less in these analyses than it is in the total group analyses, these subgroup comparison results cannot be directly compared with the total group results (i.e., to Table G-6). 0.0 5.0 10.0 15.0 20.0 25.0 30.0 30.0 40.0 50.0 60.0 70.0 80.0In cr em en t L to t- La m b, d B Lamb, dB G-6

There are two different designs for making the comparison: (1) Same Schools - comparing adjusted scores of students in the two groups in each school, then averaging the differences over all target schools; or (2) Separate Analysis - analyzing the adjusted scores of each group of students separately for all target schools, ignoring the scores of the other students. For each school the first design weights the school the same for both subgroups, whereas the second design gives more weight to the school in the disadvantaged student analysis if it has a higher number of disadvantaged students, and more weight in the non-disadvantaged student analysis if it has a higher number of non-disadvantaged students. For both designs, only average scores based on five or more students were included in the analysis. Note that for the “same school” design, this excluded all schools in which either one of the subgroups had fewer than five students. Table G-7 shows the effects of the aircraft noise increment on scores of disadvantaged and non-disadvantaged students. The results from the two sets of analyses are not the same. For example, the effects on disadvantaged student scores tend to be larger in the “Separate Analysis” than in the “Same Schools” comparison analysis. However, the “Same Schools” design probably addresses the comparison question better, in that it is a direct comparison between scores under identical noise and school conditions, i.e. teacher, teacher/pupil ratio, etc. TABLE G-7 Comparison of the Effects of Aircraft Noise on Disadvantaged and Non-Disadvantaged Students. Achievement Test Estimated Noise Effect1 Same Schools Separate Analysis Disadvantaged Non- Disadvantaged Disadvantaged Non- Disadvantaged Reading Grade 3 -0.11722 -0.1993 -0.2079 -0.2356 Reading Grade 4 -0.10942 -0.2160 -0.1895 -0.2058 Reading Grade 5 -0.13942 -0.2131 -0.1635 -0.2245 Math Grade 3 -0.14392 -0.1874 -0.2009 -0.1785 Math Grade 4 -0.19892 -0.2521 -0.2211 -0.1563 Math Grade 5 -0.11542 -0.2368 -0.1421 -0.2341 1 Effect size estimates are in units of fractions of a standard deviation for a 10dB difference in incremental level. 2 Not significantly different from zero. G-7

The results for the “Same Schools” analysis show that the effects of aircraft noise are significantly greater on non-disadvantaged than disadvantaged students, 71 percent more so for reading and 48 percent more for math. In fact, the effects for disadvantaged students in the “Same Schools” analysis, unlike the effects in the other three columns of Table G-7, are not statistically significantly different from zero. In terms of percentile change, the effect size of -0.2 to -0.25 noted in the table for non- disadvantaged students corresponds to a 10 percentile reduction in state rank, (say from a 50th to almost a 40th percentile school in the state ranking) for an increment of 10 dB (103 target schools in 2008) in incremental aircraft noise. An increment of 15 dB (30 target schools in 2008), corresponds to a 15 percentile reduction in state ranking, and 20 percent for a 20 dB increase (10 target schools in 2008). G.5. Effects of Sound Insulation on School Test Scores The most obvious method for determining the effectiveness of sound insulation is to compare test scores in the years before and after the insulation was implemented. The achievement scores both before and after the intervention have similar levels of random measurement error. Therefore, the measurement of change must treat them similarly, for example by creating a change measure by subtracting the earlier score from the later score. Of the target elementary schools analyzed in this study, there were twenty-nine insulated during the period between 2000 and 2008 that were both open for at least one year before and after being insulated and had test scores available. For these schools, it is possible to compare average test scores in years after sound insulation with scores in the same school before sound insulation. The results are shown in Table G-8. Based on changes in demographics of student bodies over the period, the demographically predicted changes in test scores in these schools were generally negative, and the apparent effects of insulation, as shown in the unadjusted results, were not sufficiently positive to overcome the demographic trend. None of the unadjusted changes were either noticeably or statistically significantly different from zero. After adjusting for the demographic trend, the effects were somewhat more positive, at Reading Grades 4 and 5, but not statistically significantly so. The sample size of schools undergoing sound insulation during the period of this study was insufficient to obtain a reliable estimate of the effects of that insulation. Estimating the benefits of sound insulation by comparing test scores before and after the insulation is introduced is fraught with uncertainties. Even in the absence of noise, test scores vary from year to year in a given grade, as well as from grade to grade, and these variations are superimposed on any changes resulting from insulation. A larger sample of schools with test scores recorded before and after sound insulation is needed for a definitive result. All of the analyses described in previous sections that were conducted to identify a relationship between test scores and aircraft noise (i.e., Tables G1- G7) had omitted records for insulated schools. In an attempt to overcome the limitations in the number of schools available for evaluating the effects of sound insulation, these previous regression analyses were repeated, but this time separately estimating aircraft noise effects on insulated and non-insulated schools. G-8

TABLE G-8 Average Changes in Test Scores Associated with School Sound Insulation. Math Grade 3 Math Grade 4 Math Grade 5 Reading Grade 3 Reading Grade 4 Reading Grade 5 Unadjusted Change -0.1635 0.0406 -0.0067 -0.1600 0.0198 0.0199 Students t -1.65 0.41 -0.07 -1.68 0.22 0.21 Demographically Predicted Change -0.1287 -0.0391 -0.1086 -0.0774 -0.1199 -0.1421 Students t -3.14 -0.89 -2.43 -2.07 -1.96 -3.06 Adjusted Change -0.0348 0.0797 0.1020 -0.0826 0.1397 0.1620 Students t -0.30 0.88 1.00 -0.75 1.56 1.59 Number of Schools 22 19 23 23 20 21 Note: Differences are in standard deviations of school average test scores The models posited three noise effects, namely (1) Ambient Noise, (2) Aircraft Noise Increment for insulated schools (Y), and (3) Aircraft Noise Increment for non-insulated schools (N). Between the years 2000 and 2008 the sample of open target insulated schools (those within the DNL 55 noise contour) ranged from 98 to 119. The results of the analysis are shown in Table G-9 where it is noticeable that the slope for the estimated noise effect is statistically significantly negative for non-insulated schools, similar to the results in Table G-6. For insulated schools, on the other hand, the aircraft noise effects did not differ significantly from zero. Thus, it would appear that the act of sound insulating a school exposed to aircraft noise is to return student test scores to what they would be if the aircraft noise were removed. G-9

TABLE G-9 Effect of Sound Insulation on Test Scores. Test Insulation Y/N Estimated Noise Effect* Adjusted t-Value** Reading Grade 3 N -0.134 -3.66 Y 0.048 0.53*** Reading Grade 4 N -0.128 -3.88 Y -0.093 -1.01 Reading Grade 5 N -0.111 -3.04 Y 0.033 0.36*** Math Grade 3 N -0.118 -2.86 Y 0.127 1.26 Math Grade 4 N -0.134 -3.74 Y -0.016 -0.17*** Math Grade 5 N -0.102 -2.50 Y 0.132 1.31 * Effect size estimates are in fractions of a standard deviation for a 10dB difference in incremental noise level. **Student’s t-value adjusted by the deign effect – see Appendix F.4. *** Not significantly different from zero. G-10