We've put this document up before completion of all the links, plus there are a couple of tables not finished yet, but we're sure you'll find the document full of useful information about noise and its effects on people and the environment. You can search it using the search feature of your particular browser.

 We will be updating it over the next week or so, so please keep coming back.

 Thank you for your interest.

 Noise Pollution Clearinghouse, April 25th, 1997.

Return to top of document

Return to NPC Library
Return to NPC Home Page

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

The author is indebted to colleagues at Bolt Beranek and Newman for discussions, suggestions, and assistance in the analysis of the data of the National Urban Noise Survey. In particular, Drs. William Galloway and David Green suggested statistical analyses; Dr. Theodore Schultz's work was the basis of comparisons with other survey data; Dr. Glenn Jones examined the attitudinal data for inflection points; Mr. Myles Simpson prepared much of the information on noise sources; Mr. Richard Horonjeff analyzed the relationship between noise levels and annoyance as a function of time of day; and Messrs. Karl Pearsons and Dwight Bishop commented on draft material. Mr. Suyeo Tomooka assisted extensively in data reduction.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

A major responsibility of the Environmental Protection Agency, Office of Noise Abatement and Control (EPA/ONAC), is to protect public health and welfare from the deleterious effects of noise by coordinating research activities, promulgating Federal noise emission standards, and providing information to the public regarding the effects and control of noise. Such activities must be based as firmly as possible upon scientific understanding of the effects of noise on people. EPA/ONAC has thus far relied extensively upon the information contained in the "Levels Document" (EPA, 1974) for information about the extent and severity of various impacts of noise.

The research from which these public health and welfare criteria were derived, however, was quite specialized and narrow. In particular, the great bulk of the data on community response to noise exposure (principally annoyance) concerned aircraft and airport related noise only. Since only a small proportion of the American population is exposed to such noise, a nationwide Urban Noise Survey (UNS) was undertaken in the Spring of 1974.

UNS differed from previous studies of noise pollution in several important ways:

(1)      The survey was national rather than local in scope. Prior social surveys had generally been restricted to a small number of geographically related sites.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

(2)      UNS did not place emphasis upon the evaluation of any single noise source. Almost all previous study of community reaction to noise exposure had been limited to transportation noise.

(3)      UNS was specifically intended to investigate community reaction over broad ranges of noise exposure conditions and lifestyles.

(4)      UNS was designed to take advantage of systematic a priori noise exposure information. The interviewing sites were selected from one hundred sites nationwide at which very detailed noise measurements had been made.

Thus, the data of UNS offer the most comprehensive sampling of public reaction to noise exposure yet available. The data cover virtually the entire range of noise exposure and population density conditions in non-rural America. Data were collected at twenty four sites in seven cities across the nation at which previous detailed noise measurements had been made for other purposes (Galloway et al., 1974). These sites, although exposed to occasional aircraft overflights, were intentionally selected to avoid significant airport and highway noise exposure. Human exposure to surface street traffic noise was nonetheless comparable in level to highway noise at some sites.

More than two thousand interviews of randomly selected respondents were conducted at these sites, with a comprehensive yet brief questionnaire that contained questions about all major effects of noise on people and all pre dominant sources of community noise. One unique feature of

Return to top of document

Return to NPC Library
Return to NPC Home Page 

this survey was that a continuous set of 24 hour noise measurements was taken at the sites at the same time that interviewing was in progress. Another important difference in design was direct measurement of annoyance, as discussed by Rylander et al. (1972) inter alia. The prevalence of annoyance was not inferred from constructed statistical indices; it was determined from respondents' answers to specific questions.

This report presents the overall analysis of the data of the national Urban Noise Survey. Like the experimental design, the analysis departs from some prior analyses of social survey data. In particular, greater emphasis is placed on prevalence of noise effects in groups of people instead of individual attitudinal variables. Thus, little effort is made to "explain" individual attitudes by comparing their intensities. Rather, attention is concentrated on predicting population proportions affected in various ways by noise exposure.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

The following summary of procedures, excerpted in part from Simpson et al. (1974), is intended only as a brief summary. The reader is referred to Simpson et al. (l974) for more detailed information and a discussion of the rationale of the survey.

Four disjunctive criteria were employed for site selection.

The questionnaire (contained in Appendix A) was designed to gather information about the respondents' attitudes toward their environment, with the greatest emphasis on noise. Simple random sampling without replacement was elected as the sampling procedure. The sample frame most appropriate to the available resources was the reverse telephone directory. The target population of the survey was the adult American urban population habitually exposed to community noise not predominantly of aircraft or highway origin.

Return to top of document

Return to NPC Library
Return to NPC Home Page

Interview data were keypunched on tabulating cards and processed by computer. Numerous tabulations of these data may be found in Appendix B. They are of interest primarily to those who wish to make uses of the data beyond those reported here. This section proceeds from the general to the specific, through successively finer analyses of findings. Few readers will be equally interested in all sub-sections. Those satisfied with a descriptive account of "what happened" need not read beyond the preliminary sections for a narrative account of findings. For readers more interested in statistical analyses, the introductory sections may be tedious. Such readers may wish to proceed to Sections III-17 et seq. after reading Section III-1.

Section III-1 presents an overall view of the findings as a context within which other analyses may be understood. Sections III-2 and III-3 describe major effects associated with the two independent variables of UNS (noise exposure and population density). Sections III-4 through III-6 pre sent demographic differences associated with age, sex, and socioeconomic level. Sections III-7 through III-16 contrast response patterns associated with answers to key questions.

For the sake of clarity and brevity, most of these introductory sections contain contrasts between extreme sub samples; e.g., high vs. low noise exposure, high vs. low

Return to top of document

Return to NPC Library
Return to NPC Home Page 

socioeconomic level, young vs. old respondents, etc. Differences not specifically mentioned are of small size or little relevance. Furthermore, percentages are reported rounded off to the nearest integer. The reader is also cautioned against drawing causal inferences about the simple relationships discussed in the first sixteen sub sections, since virtually all of these first order relation ships have strong higher order interactions.

Sections III-17 et seq. are given to statistical inference rather than simple description. Section III-17 presents findings pertaining to noise sources. Section III-18 summarizes regression analyses for key variables. Section III- 19 details a search for critical noise levels. Section III- 20 explores the relationship between noise exposure and annoyance as a function of time of day. Section III-21 addresses a methodological issue, the mode of interviewing. Section III-22 is concerned with another methodological issue, sampling bias.

III-1      0verview of Data

A total of 2037 Persons (762 men, 1275 women) was interviewed, of whom 670 men and 1164 women were interviewed by telephone. The other respondents were interviewed in person. Table III-1 summarizes the number of interviews conducted at each site, as well as the noise level and population density of each site.

Nationwide, 69% of all respondents rated their neighborhoods as good or excellent places to live, with only 23% seriously thinking of moving within the next year. Of these people, only 1% cited noise as a reason for moving. Sixty two percent of all respondents regarded their neighborhoods

Return to top of document

Return to NPC Library
Return to NPC Home Page 

*Sites at which personal (face-to-face) interviews were conducted.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

as quiet, but 46% claimed to have been "bothered or annoyed" by noise in their neighborhoods.*
Thirty-one percent of the ever-annoyed people were "highly annoyed" (self-rated "very" or "extremely" on an adjective scale that also included the terms "not at all", "slightly", and "moderately") by noise in their neighborhoods. Neighborhood noise was thought to be equally annoying at all times of day by 22% of the ever annoyed; another 22% of these people found neighborhood noise more annoying in the evening than at other times of day; and 27% found such noise more annoying at night.

Over half of the ever-annoyed found noise more bothersome when inside the house than when outside; the others either found noise more bothersome outdoors or felt there was no difference outside or inside the house. The major findings with regard to time and place of annoyance are summarized in Figure III-1.

Table III-2 rank orders the frequency with which ever annoyed people reported hearing various noise sources. The table also indicates the average annoyance on an arbitrary 5 Point adjective scale (where 1 corresponds to "not at all annoyed" and 5 corresponds to "extremely annoyed") associated with each source. As the table shows, motor vehicle noise was the most pervasive noise source

*These latter respondents are referred to henceforth as the "ever-annoyed". Because the structure of the questionnaire concentrated attention on the ever-annoyed, most of the findings reported below concern this group of people. Figures based on the total sample are referred to as "percentages of all respondents".

Return to top of document

Return to NPC Library
Return to NPC Home Page 

FIGURE III-1. SUMMARY OF FINDINGS WITH REGARD TO TIME AND PLACE OF ANNOYANCE

Return to top of document

Return to NPC Library
Return to NPC Home Page 

heard nationwide (reported by 865 of these respondents), and also the most annoying. People and pets were the next most often noticed sources, followed by aircraft, construction, power garden equipment and electronic sources (radios, TVs, etc.).

Table III-3 rank orders the frequency with which people who had ever been annoyed by noise in their neighborhoods experienced various effects of noise. Sleep disturbance, the most common effect of noise exposure (reported by 60% of these respondents) was also the most annoying (with a mean value of 3.6). Startle and speech interference were somewhat less pervasive effects, and of lesser annoyance.

Nineteen percent of the ever-annoyed people (9% of all respondents) claimed to have complained to officials about noise in their neighborhoods. Twenty-four percent of all respondents felt themselves to be more sensitive to noise than most people, while only 6% of all respondents felt that noise exposure had affected their health.

III-2      Differences Associated with Noise Exposure

The numerous effects associated with noise exposure are most simply presented by comparing data from two extreme subsamples: one of six heavily exposed sites (mean L_dn = 70.0 dB) and one of seven lightly exposed sites (mean L_dn = 54.6 dB). All comparisons in this section are of averaged data from the high noise exposure subsample with respect to the low noise exposure subsample.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

¹These figures must be multiplied by .46 if extrapolated to the entire sample. For example, the 60% of the ever-annoyed people who reported sleep disturbance constitute 20% of the entire sample.

²Mean annoyance on an arbitrary five point scale.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Neighborhood satisfaction was considerably lower in the high exposure subsample. Thirty four percent fewer people in the high exposure subsample described their neighborhoods as an excellent place to live, and 24% more people described their neighborhoods as only a fair place to live (Q. 4)*. Fifteen percent more people at the high exposure sites intended to move out of the neighborhood in the next year (Q. 9). Thirty eight percent fewer people regarded their neighborhoods as quiet (Q. 11). Seventeen percent more people had been annoyed by noise (Q. 13) at the high exposure sites; and twenty seven per cent more people were annoyed in their homes (Q. 18). Figure III-2 is a plot of the percentage of respondents at each of the 24 sites who were highly annoyed by noise exposure (i.e., rated themselves as "very" or "extremely" annoyed). The correlation coefficient between L_dn and annoyance, .70, is extremely unlikely to have arisen by chance alone from a sample of size 24. Its fiduciary limits (for a 95% confidence interval) are from 0.45 to 0.86.

Emphasis placed upon the annoyance of various noise sources differed considerably between the two subsamples, with smaller numbers of respondents in the high exposure sub sample reporting annoyance from pets (21% fewer), helicopters (33% fewer), power garden equipment (475 fewer), sports cars (11% fewer), and motorcycles (9% fewer). On the other hand, more respondents in the high noise exposure subsample reported annoyance from construction noise (9% more), people's voices (24% more), radio and TV sets (11%

*This number refers to questionnaire item 4, found in Appendix A.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

more), motor vehicle noise (13% more), large trucks (15% more), buses (19% more), and constant traffic (38% more).

Similarly, more respondents in the high exposure sub sample reported activity interference such as listening (25% more), speaking (20% more), and sleeping (8% more). Seven percent more respondents in the high exposure sub sample claimed to have registered complaints about noise with officials. Figure III-3 plots complaint rates as a function of noise exposure at the 24 sites. The correlation coefficient, .23, is likely to have arisen by chance alone. In general, the direction of differences between responses in the two subsamples were consistent with the position that noise exposure degrades the quality of life.

III-3      Differences Associated with Population Density

Effects of population density on response patterns were analyzed in the same fashion as in Section III-2, through comparisons between extreme subsamples. Data from five high population density sites (mean density = 44,920 people per square mile) are compared with data from five low population density sites (mean density = 1600 people per square mile). Comparisons in this section are of averaged data from the high density subsample with respect to the low density sample.

Response patterns in the extreme population density sub samples closely paralleled (within a few percent) those

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

associated with noise exposure. While neighborhood satisfaction was lower in the high density subsample, the incidence of noise induced annoyance was higher. The proportions of respondents reporting annoyance from various sources differed very little from those reported in Section III-1. Those sources more prevalent in highly urbanized areas were more often mentioned than was the case in high noise areas; e.g., people's voices, air planes, radio and TV sets, and automobiles.

Twenty percent more people in the high density subsample reported interference with listening, 9% more reported interference with talking, and 9% more reported sleep disturbance. These figures. hardly differ from those noted in Section III-2.

III-4      Differences Associated with Age

To assess differences in opinions associated with age, respondents were divided approximately into thirds on the basis of age, as estimated from the year in which formal schooling was completed. This section contrasts the opinions of the 30% of the respondents aged 30 years or younger with the 34% of the respondents aged 45 or older.

Differences in neighborhood satisfaction associated with noise between the two groups were negligible. Older respondents had lived longer in their neighborhoods, while younger respondents were more ready to move within the year. Nonetheless, differences in percentages of the two groups citing noise as a cause for discontent were mostly less than 5%.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Six percent fewer of the younger group thought their neighborhoods were quiet, but 15% fewer reported being annoyed by noise in their neighborhoods. Eleven percent more of the younger respondents could not distinguish seasonal differences in annoyance, but 13% more of the older respondents thought neighborhood noise was more annoying in the summer. Greater percentages of the older respondents thought neighborhood noise was more annoying weekdays (13% more) and inside the house (15% more).

Fourteen percent fewer of the older respondents were annoyed by construction noise, but greater percentages of the older respondents reported annoyance from airplanes (11% more), helicopters (11% more), power garden equipment (15% more), sports cars (11% more), and motorcycles (11% more). Nonetheless, uniformly greater percentages of the younger respondents reported speech or listening interference (18% and 6% more, respectively), startle or fright (15% more), or sleep interference (19% more). Nine percent more of the older respondents felt they were more sensitive than most to noise. A gross relationship between age and complaint rates may be seen in Figure III-4.

III-5      Differences Associated with Sex

Differences between male and female respondents were small both in number and magnitude. For example, the largest difference between men and women among the neighborhood satisfaction questions was less than 6%. More men intended to move within the next year than women, but only about 1% of either sex respondents gave noise as a reason for moving.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Differences between the sexes with regard to assessment of noisiness and annoyance associated with neighborhood noise exposure were also trivial. The largest difference of opinion was an 8% difference on the issue of season of the year of greatest annoyance - more men than women felt that summer was the most annoying time.

Differences between men and women in ratings of noise sources were also inconsequential, rarely exceeding two or three percent. A sole exception was that 10% more women reported hearing construction noise in their neighborhoods. No differences on other substantive matters (such as activity interference, complaint rates, sensitivity to noise, or health effects) exceeded 5%, and most were on the order of one or two percent.

Perhaps the most notable difference between the sexes on the entire questionnaire was in time spent at home. Women reported spending about 3-1/2 more hours at home on both weekdays and weekends than men.

III-6      Differences Associated with Socioeconomic Level and Income

A.      Socioeconomic Level

The subsamples contrasted in this subsection are respondents in the upper and lower halves of the Duncan scale of socioeconomic level. The observed differences tend to support the hypothesis that high socioeconomic level respondents suffer less from noise pollution than do low socioeconomic level respondents.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

For example, neighborhood satisfaction was higher among the high socioeconomic level respondents (18% more rated their neighborhoods as excellent places to live); 6% fewer of the high socioeconomic level respondents were thinking of moving within the year; 9% more of the high SEL respondents considered their neighborhoods quiet; and 19% more of the high socioeconomic level respondents were unable to distinguish differences in annoyance with neighborhood noise among the seasons.

Differences in rates of identification of various noise sources were relatively small but consistent; 8% fewer high socioeconomic level respondents reported hearing people's voices, 4% fewer reported airplanes, 6% fewer reported automobiles, and 7% fewer reported traffic. On the other hand, 6% more high socioeconomic level respondents reported hearing pets, 7% more reported power garden equipment, and 8% more reported sports cars.

Similarly, 9% fewer high socioeconomic level respondents reported interference with listening, and 6% fewer reported fear or startle. Seven percent more of the high socioeconomic level respondents reported complaining about neighborhood noise. High socioeconomic level respondents spent an average of an hour and a quarter more at home on weekdays, and two and a half hours more at home on weekends.

B.      Income

The pattern of differences associated with extreme income groups was predictably similar to those associated with extreme socioeconomic groups. The magnitudes of the differences tended to be greater, however. The two income subsamples

Return to top of document

Return to NPC Library
Return to NPC Home Page 

contrasted here are those respondents reporting annual household incomes below $10,000 and those respondents reporting annual household incomes above $20,000.

Forty two percent more high income respondents rated their neighborhoods as excellent places to live; 15% fewer high income respondents were thinking of moving within the year; and 20% more of the high income respondents thought their neighborhoods were quiet. A relationship between income and exposure levels is seen in Figure III-5.

Differences in identification of noise sources were also similar to those associated with high socioeconomic levels. Twenty one percent more high income respondents reported power garden equipment, 18% reported more sports ears, and 12% reported more motorcycles. On the other hand, 11% fewer reported constant traffic noise.

Likewise, 16% fewer high income respondents reported that noise interfered with listening, and 9% fewer were startled or frightened by neighborhood noises. Nonetheless, 7% more high income respondents reported sleep disturbance. Seven percent more high income respondents also reported complaining about neighborhood noise. The high income respondents spent about an hour and forty minutes more time at home on weekdays than did the low income respondents, and an additional hour and a half on weekends.

III-7      Differences Associated with Neighborhood Satisfaction
          (Q. 4)

The 69% of all respondents who rated their neighborhoods as good or excellent places to live (the "highly satisfied") differed from the 31% of all respondents who rated their neighborhoods as

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Thirty six percent more of the respondents who thought their neighborhoods were good or excellent viewed their neighborhoods as quiet. Further, nineteen percent more of those highly satisfied with their neighborhoods had never been bothered or annoyed by noise in their neighborhoods. The highly satisfied who had been bothered or annoyed were not as aware of differences in annoyance as a function of time of day or season of the year. The highly annoyed identified fewer neighborhood noise sources as annoying, and were generally less annoyed by them.

Not surprisingly, the respondents who were highly satisfied with their neighborhoods were of a higher socioeconomic level than those who were not (by about one and a half deciles on the Duncan scale), and reported average annual household incomes twice as great as respondents less satisfied with their neighborhoods ($11,500 VS. $5,700).

III-8      Differences Associated with Rated Noisiness of
          Neighborhoods (Q. 11)

Sixty two percent of all respondents described their neighborhoods as quiet when asked to characterize them as quiet, noisy, or

Return to top of document

Return to NPC Library
Return to NPC Home Page 

neither quiet nor noisy during the preceding year. The responses of these respondents are contrasted with those of the 31% of all respondents who characterized their neighborhoods as noisy.

Thirty four percent more of those who thought their neighborhoods were quiet also rated them as excellent or good places to live. Sixteen percent more spontaneously mentioned the absence of noise as the most favored aspect of living in their neighborhoods. Fourteen percent more of those characterizing their neighborhoods as noisy were thinking of moving during the next year. Forty nine percent more of the respondents who thought their neighborhoods were quiet had never been annoyed by noise in their neighborhoods. Thirty seven percent more of the respondents who thought their neighborhoods were quiet reported that annoyance was only minimal (not at all or slightly), whereas 38% more of those respondents who thought their neighborhoods were noisy found their annoyance considerable (moderately, very, or extremely).

Twenty one percent more of the respondents who thought their neighborhoods were noisy thought that noise was more annoying in the evening or at night, while 15% more of the same respondents were more annoyed indoors than outdoors.

The predominant noise sources heard by people who thought they lived in quiet neighborhoods were peoples' voices (16% more than in noisy neighborhoods) and constant traffic (18 more than in noisy neighborhoods). Conversely, greater percentages of respondents who thought they lived in noisy neighborhoods reported hearing power garden equipment (17% more), helicopters (15% more), and motorcycles and sports cars (8% more each).

It is quite clear that people who thought they lived in quiet neighborhoods suffered fewer effects of noise exposure, since 25%

Return to top of document

Return to NPC Library
Return to NPC Home Page 

fewer reported interference with listening, 18% fewer reported interference with speaking, 20% fewer reported sleep interference, 15% fewer reported startle or fear, and 22% fewer reported shutting windows because of intrusive noise. Slightly fewer (3%) of the respondents who thought they lived in quiet areas reported complaining about noise, while 11% more thought noise had not affected their health.

The mean annual household income was somewhat greater for those who thought they lived in quiet neighborhoods ($10,650 vs. $8,250).

III-9      Differences Associated with Annoyance from.
          Neighborhood Noise (Q. 13)

The major breakpoint in the interview was at Question 13, "Have you ever been bothered or annoyed by noise in your neighborhood?" If answered negatively (as 53% of all respondents did), the interview concluded quickly without questioning about noise sources or effects. This section contrasts the responses of the "never-annoyed" with those of the "ever-annoyed".

Seventeen percent more of the never-annoyed respondents thought their neighborhoods were good or excellent places to live. Ten percent more of the never-annoyed specifically mentioned a noise-related aspect of their neighborhoods (e.g., "peace and quiet", "no noise from ....", etc.) as the "first most liked thing" (Q. 5). Thirteen percent fewer of the never-annoyed specifically mentioned a noise related aspect of their neighborhoods as the "least liked thing" (Q. 7). Eleven percent fewer of the never annoyed were thinking of moving within the year.

Forty one percent more of the never-annoyed respondents described their neighborhoods as quiet places to live, and 11% fewer of them thought neighborhood noise had affected their health.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

III-10      Differences Associated with Intensity of Annoyance
          (Q. 14)

Fourteen percent of all respondents described noise in their neighborhoods as either very or extremely annoying over the past year. The relationship between annoyance so measured and average income at the 24 sites is seen in Figure III-6. Among the most notable differences between these highly annoyed respondents and the others were their self reports of the effects of noise exposure.

Twenty one percent more of the highly annoyed respondents judged their health to have been affected by neighborhood noise; specifically, in the form of hearing damage. Eleven percent more of the highly annoyed respondents thought themselves more sensitive to noise than most people. Twenty four percent more of the highly annoyed reported sleep interference, 20% more reported interference with listening, 21% more reported interference with speaking, 21% more reported shutting windows to keep out noise, and 12% more reported startle from noise. In general, greater numbers of highly annoyed respondents identified the various noise sources as annoying, and were consistently more greatly annoyed by each noise source then were the respondents who were not highly annoyed.

Fifty one percent more of the highly annoyed described their neighborhoods as noisy places to live, 26% fewer rated their neighborhoods as good or excellent places to live, 16% more spontaneously mentioned noise as the least liked aspect of their neighborhoods, and 14% more were thinking of moving.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

III-11      Differences Associated with Startle

Thirteen percent of all respondents reported considerable annoyance from startle or fear produced by neighborhood noises. Their responses are contrasted in this section with those of a subsample of ever-annoyed respondents composed of those who were minimally annoyed by startle or fear (those who reported they were not at all or slightly annoyed) and those who reported no startle or fear at all.

The opinions of respondents who were considerably annoyed by startle and fear differed from those who were not in many ways. Twenty percent more of them thought their neighborhoods were noisy, and 13% more were thinking of moving. Those experiencing considerable annoyance with startle or fear also suffered more from other noise effects: 16% more were highly annoyed by neighborhood noises, 21% more experienced interference with listening, 31% more experienced interference with speaking, 30% more reported sleep disturbance, and 21% more kept their windows shut because of noise. Nineteen percent more felt that noise had affected their health, and 8% more felt that they were more sensitive to noise than most people. In short, these 13% of all respondents represent an extreme subsample both in terms of effects of noise and opinions about exposure.

III-12      Differences Associated with Sleep Disturbance

Twenty one percent of all respondents expressed considerable annoyance from sleep disturbance caused by neighborhood noises. The opinions of these people are contrasted with those who experienced no annoyance or only slight annoyance from sleep disturbance, or whose sleep was not disturbed by noise.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Twenty one percent more of the people considerably annoyed by sleep disturbance considered their neighborhoods noisy, and 17% of them considered their neighborhoods more noisy at night than at other times of day. Eighteen percent more reported that they were more annoyed inside their homes. More of the people who were considerably annoyed by sleep disturbance also heard construction noise (17% more), people's voices (16%more), pets (l4% more), and radio and TV sounds (12% more). Fifteen percent more of these people experienced interference with listening, while 20% more experienced interference with speaking. Twenty three percent more reported startle or fear, and 27% more shut their windows to keep out noise. Although 16% more of these considerably annoyed people felt that noise had affected their health, only 1% more had complained to officials. Six percent more of these people felt themselves to be more sensitive than most to noise exposure.

III-13      Differences Associated with Complaints

Nationwide, 9% of all respondents (13% of the ever-annoyed) said they had complained about noise in their neighborhoods. The views of these people are contrasted with those of respondents who had not complained about noise in this section.

Twenty percent fewer of the complainers thought their neighborhoods were good or excellent places to live, and 13% more of them spontaneously mentioned noise as the least liked aspect of their neighborhoods. Thirty two percent more of the complainers rated their neighborhoods as noisy during the previous year, while 57% more were annoyed by neighborhood noise. The intensity of their annoyance was greater as well; 22% more of the complainers were highly annoyed. Nine percent more of the complainers found neighborhood noise more annoying on weekdays than weekends.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Greater percentages of the complainers identified every neighborhood noise source (except for light trucks) as sources of annoyance. These differences, however, were generally on the order of 5%. Similarly, greater percentages of the complainers reported every noise effect: speech interference (8%), listening interference (75), startle or fear (10%), and sleep disturbance (23%). Eleven percent more of the complainers kept their windows shut because of neighborhood noise.

On average, complainers spent an additional 25 minutes at home weekdays, but 13 minutes fewer on weekends. Twelve percent more of the complainers described themselves as more sensitive than most to noise, while 17% more felt that noise had affected their health. Complainers averaged about ½ decile higher on the Duncan Scale of socioeconomic level, and enjoyed about $1000 more annual household income.

III-14      Differences Associated with Sensitivity

Twenty four percent of all respondents judged themselves to be more sensitive to noise than most other people. This section contrasts their opinions with those of the respondents who judged themselves to be about as sensitive or less sensitive than most.

In demographic terms, the respondents who judged themselves more sensitive than most included 7% more women, had an average annual income $1250 higher, and averaged half a decile higher in socioeconomic level than other respondents. Differences in neighborhood satisfaction between the two groups of respondents were minimal. Although only two percent more of the more

Return to top of document

Return to NPC Library
Return to NPC Home Page 

sensitive respondents judged their neighborhoods to be noisy, it more had been annoyed by neighborhood noise. Eight percent more of the more sensitive respondents found neighborhood noise more annoying on weekdays than on weekends. Nine percent more of the more sensitive respondents were unable to distinguish whether neighborhood noise was more bothersome inside or outside the house. More of the more sensitive respondents identified all neighborhood noise sources (except automobiles and small trucks) as annoying. These differences, however, were relatively small (on the order of 5%).

Perhaps the greatest differences observed were in susceptibility to noise effects. Eleven percent more of the more sensitive respondents reported listening interference, 18% more reported startle or fear, 6% more reported sleep disturbance, 8% more reported speech interference, and 11% more reported keeping windows shut because of neighborhood noise. Seven percent more of the more sensitive respondents reported complaining about noise.

III-15      Differences Associated with Self Rated Health Effects

Five percent of all respondents thought that noise in their neighborhoods had affected their health in some way. This section contrasts their opinions with the 95% of the respondents who did not think noise had affected their health.

It is clear that the health-affected respondents are an extreme group: 28% more of them experienced interference with listening, 31% more suffered sleep disturbance, 36% more experienced speech interference, and 29% more shut their windows to keep out neighborhood noise. Twenty percent more had complained

Return to top of document

Return to NPC Library
Return to NPC Home Page 

about noise, and 22% more felt more sensitive to noise than most people.

Larger percentages of the health-affected respondents reported various neighborhood noise sources as annoying: these included construction noise (7% more), people's voices (11% more), radios or TV sets (16% more), sports ears (13% more), small trucks (10% more), large trucks (l4% more), constant traffic (l4% more), and so forth.

Although 23% fewer of the health-affected viewed their neighborhoods as good or excellent places to live, 19% fewer were considering moving within the year. Forty two percent more of the health-affected thought their neighborhoods were noisy, and 48% more had been bothered-or annoyed by noises in their neighborhoods. The health-affected respondents had no clear consensus on the time of day or season of the year when noise was more annoying, nor on whether noise was. more annoying indoors or outdoors.

III-16      Differences Associated with Duration of Exposure
               to Neighborhood Noise

This section examines differences observed as a function of duration of exposure to neighborhood noise. In Part I, comment is made on differences associated with short vs. long daily exposure. In Part 2, comment is made on differences associated with short vs. long duration of residence.

  1.      Daily Exposure

All respondents were divided into two groups: those who spent 20 or more hours at home daily, and those who spent

Return to top of document

Return to NPC Library
Return to NPC Home Page 

l4 or fewer hours at home daily. This division corresponded to approximately + or - .50 from the grand mean for all respondents (17 hours) spent at home daily.

Four times as many women as men spent more time at home (81% vs. 19%). Five to ten percent more of the respondents who spent more time at home experienced all of the noise effects (speech and sleep interference and fear or startle). Greater percentages of these respondents (about 6% more on average) also reported hearing most of the noise sources. The respondents who spent more time at home tended to be of slightly lower socioeconomic level (about half a decile, on average).

Understandably, more of the respondents who spent less time at home found noise in the mornings and evenings to be more annoying than at other times of day, and noise inside the house to be more annoying than outside the house. Most of the above differences in extensity were relatively small (on the order of 10% or less). Differences in intensity of opinions were even smaller, rarely exceeding 0.3 of a response category.

  2.      Duration of Residence

The overall distribution of respondents' duration of residence is seen in Figure III-7; an exponential fit to the distribution is remarkably good. Ail respondents were divided into two groups: those who had lived in their neighborhoods for six months or less, and those who had lived in their neighborhoods for five years or more. Because only, 2% of the sample fell into the former category, the reliability of comparisons between the two categories of respondents is poor.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Thus, it is not surprising that there is no clear trend of differences apparent in comparisons of response patterns for the two groups. New residents of a neighborhood tended to differ from long term residents considerably in which noise sources they heard and found annoying; they reported more construction noise, pets, airplanes, helicopters, power garden equipment, specific vehicles, and miscellaneous sources, but fewer peoples' voices, radios and TVs, and less motor vehicle noise in general.

No generalization about sensitization vs. habituation to neighborhood noise sources seems to be supportable on the basis of differences observed between the two groups. Although differences in intensity of annoyance produced by various sources were relatively large (approaching a full category on a five point scale in some eases), they were inconsistent in direction. Similarly, there were sizable differences but no consistent trends in reported effects of noise. For example, twenty three percent more of the newcomers reported having been annoyed by noise in their neighborhoods, but nine percent fewer reported interference with listening to TV and radio.

III-17      Noise Sources

A.      Prevalence of Sources

Questions 19-34 posed the question "Over the past year have you heard ... in your neighborhood?" for major community noise sources. Respondents who had heard one of these sources were asked to rate how annoying the source had been over the year on a five-point adjective scale. Table III-4 rank orders these noise sources within population density strata. Table III-5 is an overall ranking of noise sources that affect the

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

urban population, calculated by weighting the responses by the percentage of the total population living in each population density stratum. Table III-6 rank orders other outdoor noise sources mentioned in response to Question 34 by number of occurrences.

B.      Relationship Between Source Identification and Level

At each of 23 sites the outdoor noise environment was estimated by making an 8-minute long analog recording once an hour for a full day*. These recordings were processed to yield a time-history plot of the A-weighted noise level. During playback the sources of discrete noise events were identified by listening. Each noise event with peak level 5 dB or more above the total hourly equivalent level for that site for that hour (as determined from digital noise data) was considered to be a noise "intrusion". All intrusions were tabulated by level and source type, with peak levels classified into 5 dB increments and sources categorized as automobiles, trucks, buses, motorcycles, aircraft, sirens or horns, people, animals, mechanical equipment, telephones, radios or stereos, door slams, thunder, or rain. According to this definition, automobile and truck intrusions were observed at all sites; aircraft were observed at twenty two sites; and motorcycles were observed at seventeen sites.

Several physical indices of these noise intrusions were developed from the tabulated data at each site. The two basic indices were the daily number of intrusions by a specific source and the maximum level of the source at any time during the day (i.e., the peak level of the greatest noise intrusion). An energy-averaged peak level (determined by logarithmic addition of

*No such recordings were available at Site 1001.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

all the peak levels occurring during the day less 10 times the logarithm of the number of these intrusions) was also developed.

A fourth index, partial day-night sound level for noise sources, L_dn , was computed as well. The notation L_dnp was used to distinguish the partial L_dn values for each source from the total L_dn at a site*. The absolute value of the partial day-night level for different sources is relatively unimportant. It suffices for current purposes that the relative magnitude of L_dnp be reasonably accurate across sites for each source, so that relationships between response data and L_dnp values remain consistent.

For each of the four major intruding sources (aircraft, automobiles, motorcycles and large trucks), a linear regression was
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

(1)

where SEL_di and SEL_ni are the sound exposure levels of individual events during daytime (7 a.m. to 10 p.m.) and nighttime (10 p.m. to 7 a.m.) periods respectively. These SEL values depend on the duration of the noise intrusion and its distance (level). For point sources traveling in a straight line with a velocity v (in ft/sec) and distance r (in feet) from an observer, the sound exposure level can be approximated by:

(2)

The ratio of r to v can be assumed constant for sources at all sites. A factor of two error will result in a difference of only 3 dB. Since the peak level itself is only known to within + or - 2.5 dB, such an error is acceptable. Equations (1) and (2) can be combined to:

(3)

Return to top of document

Return to NPC Library
Return to NPC Home Page 

performed with the percentage of annoyed respondents as the dependent variable and the four indices of intrusion as independent variables: the number of noise intrusions (N)*, the maximum peak level during the day (max L_p), the average peak level during the day (L_p ), and the partial day-night level (L_dnp). Correlations were calculated for both the percentage of respondents highly annoyed by each source, and the percentage annoyed to any degree by each source. Table III-7 contains only those correlations unlikely to have arisen by chance alone (p < .05, n = approximately 20, r_c > approximately 0.4).

As maybe seen in Table III-7, the day-night average sound level has a correlation coefficient comparable to or better than that of most other noise measures. Considering the degree of uncertainty associated with the individual L_dnp values, a correlation coefficient of the order of 0.5 between annoyance responses and the day-night average level for each source is a useful finding. It suggests that annoyance associated with intrusive noise sources can be related to measurable noise exposure from such sources in the community, even when the magnitude of noise exposure from an intrusive source is below the total L_dn for a measurement site.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

A ratio of r to v of 1:1 was assumed for present purposes. For typical values for r and v, the partial day-night levels, based only on the peak levels of noise intrusions, are well below the total L_dn at each site.

*N is the number of noise intrusions measured during the 24-hour 8-minute samples. The total number of noise intrusions that might have occurred during a full 24-hour day could be approximated by multiplying N by 7.5.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

CORRELATIONS BETWEEN THE PERCENTAGE OF RESPONDENTS EITHER ANNOYED OR HIGHLY ANNOYED AND VARIOUS NOISE INTRUSION MEASURES FOR INTRUSIVE NOISE SOURCES

 Return to top of document
Return to NPC Library
Return to NPC Home Page 

III-18     Correlational and Regression Analyses

A.     Correlation Matrices

1.     Individual Data

The simple (linear) correlations among all respondents' answers to all major questionnaire items were computed as a first step. An alphabetized and cross indexed listing prepared from the correlation matrix is included in Appendix C, only those coefficients greater than 0.2 in absolute size appear in the listing. All of the coefficients are statistically significant (in the sense that they are extremely unlikely to have arisen by chance alone), primarily because of the very large sample size.

Perusal of this list of correlations yields few surprises: the composition of clusters of related variables (noise sources, attitudes, effects, etc.) are all similar to those predictable from the relationships observed in comparisons of extreme sub-samples. Among the demographic variables, for example, Population density and income correlated -.30, and age and duration of residence correlated .29. Among the situational variables, noise level correlated .36 with traffic as an identifiable noise source, but -.29 with power garden equipment. Among the attitudinal variables, responses to the ever-bothered question (Q, 13) correlated .50 with responses to the neighborhood noisiness judgment question (Q, 12), while responses to the latter question correlated .42 with the degree of annoyance question (Q, l4).

By themselves, the simple correlations are of little predictive value, since they are all confounded by their large numbers of

Return to top of document

Return to NPC Library
Return to NPC Home Page 

significant partial correlations with one another. For example, the observed correlation of .27 in the individual data between responses to the questions "has noise made you keep your windows shut" (Q. 39) and "has noise affected your health" (Q. 45) does not imply any causal relationship. It is not clear from the simple relationship whether the attitude (noise affects health) produces the behavior (keeping windows shut), whether the behavior (keeping windows shut) reinforces the attitude (noise affects health), or whether the degree of association between answers to the two questions is attributable to common associations with one or more other attitudes, behaviors, and/or noise effects.

Policy making agencies are more properly concerned with how their decisions will affect proportions of populations than with the prediction of interrelationships among individual attitudes. Thus, no further efforts were made to interpret the simple correlations among individual intensive variables.

 2.      Grouped Data

A second set of correlation matrices was therefore computed by grouping respondents within sites. This treatment of the social survey data concentrates on extensity of attitudes and behaviors. The variables of interest in the analyses reported below are therefore percentages of respondents holding common views, rather than the fervor of individual beliefs.

Table III-8 shows the simple correlations among the two major independent variables of this study (noise exposure and population density), three demographic variables (mean age, duration of residence, and annual household income), and three related measures of annoyance, computed site by site for all respondents

Return to top of document

Return to NPC Library
Return to NPC Home Page 

*The probability that a correlation in this table differs from 0 is greater than 95 if its absolute value is greater than 0.4.

**"Highly annoyed" is a linear combination of "very" and "extremely annoyed

Return to top of document

Return to NPC Library
Return to NPC Home Page 

at each site. Other demographic variables (such as socioeconomic level, time spent at home, and sex ratio) were excluded from this table for a number of reasons: poor correlation with annoyance, high correlation with other demographic variables, or difficulty in assessment. Table III-9 shows the simple correlations among the same independent variables and measures of annoyance, three effects of exposure (speech interference, sleep interference, and startle), and two attitudinal variables (sensitivity to noise and self-rated health effects).

The greater magnitude of the correlations in Tables III-8 and III-9 compared with those in the listing of individual data reflects the truism that groups of people behave more predictably than individuals. The absolute size of the sitewise correlations provides a reasonable basis for the regression analyses discussed below.

B.      Multiple Regression Analyses

An aspect of community response to noise exposure of significant interest is the percentage of people highly. annoyed by different sources of noise. The present data were analyzed to determine 1) the relationships among a number of attitudinal and situational variables known to be related to annoyance, and 2) the limits of their utility in predicting annoyance. The first two analyses presented are restricted to demographic and situational variables. A third analysis is restricted to noise effects and attitudes., A final analysis mixes the various types of predictor variables.

This analysis was accomplished by stepwise regressions conducted on the data shown in Tables III-8 and III-9. Table III-10

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

displays a stepwise regression of noise level, mean age, mean household income, and mean duration of residence on the percentage of highly annoyed respondents at the twenty four sites. The preponderance of variance accounted for b- the regression is due to the first term, noise exposure. In fact, the multiple correlation accounts for only 3% more variance than the simple correlation between noise exposure and annoyance. Thus, for most practical purposes, the percentage of respondents at a site highly annoyed by noise exposure can be predicted from exposure information simply by the relationship seen in Figure III-2.

Table III-11 displays a stepwise regression of population density, mean household income, mean duration of residence, and mean age on the percentage of highly annoyed respondents at the twenty four sites. Once again, the first variable contributes the major portion of variance accounted for, although the additional variables do account for an additional 11% of the total variance.

Table III-12 displays a stepwise regression of percentages of respondents at each site reporting speech interference, sleep interference, and startle on the percentage of respondents highly annoyed at each site. Speech interference, with the highest simple correlation with annoyance, was the first term in the regression, and accounted for virtually all the variance not attributable to error. Thus, subsequent terms are absent, since they would have contributed only trivial additional predictive power.

Similar analyses were also undertaken to predict the proportion of a community describing itself as "very" and "extremely" annoyed. The results closely paralleled those reported here, but accounted for slightly less variance. Restricting the

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

A final stepwise regression is seen in Table III-13, in which all predictor variables were permitted. The first three variables (percentage of respondents reporting speech interference, population density, and percentage of respondents believing that noise had damaged their health), all surrogates for noise exposure per se, have a multiple r of 0.95 with the percentage of respondents highly annoyed. These three variables thus account for fully ninety. percent of the variance in the annoyance data.

III-19      Critical Level Analysis

A recurring problem in a comprehensive noise abatement program is the definition of a level of community noise that represents a serious disamenity for neighborhood residents. Efforts to determine whether such "critical levels" are identifiable are reported in this section. The underlying strategy in the following analyses is to search for systematic trends in response data arranged along a continuum of increasing noise exposure.

The first step in the search for critical levels that may be inherent in the data was to tabulate noise-reaction data along the continuum of exposure as is done in Table III-l4. This informal exploration showed that while respondents at the noisiest sites generally exhibited more extensive and intensive reactions to noise than those at the quietest sites, the progression along the noise continuum was not smooth. This implied that critical levels (underlying discontinuities) would be difficult to detect visually in curves plotted from these data.

It was also observed, however, that respondents at the three noisiest sites exhibited markedly more numerous and vigorous

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

reactions than did respondents at the other 21 sites. Furthermore, it appeared that reactions at these three sites resembled one another more consistently than did reactions at other sites. It was therefore hypothesized that these three sites were distinguishable from all of the others. If this hypothesis were true, and if enough information about reactions to noise exposure greater than 70 dB were available to permit stable estimates, then a distinct upturn in most noise reaction curves might be evident at L_dn values in excess of 70 dB.

To test the hypothesis that 70 dB(A) on the L_dn scale represents a critical level at which the relation between sound levels and noise related reactions are intensified, a number of statistical tests based on the binomial sampling distribution were devised. Three groups of sites were formed: Group 1, with a mean L_dn of 71.5 dB; Group 2, with a mean L_dn of 68.4 dB; and Group 3, with a mean L_dn of 63.6 dB.

As an initial test, seventeen measures of noise reactions were considered. These measures, derived from the social survey data on a site-by-site basis, appear in the leftmost column of Table III-15. Part A of Table III-15 contains scores for ten measures of the extent of noise reactions (e.g., percentage of respondents at a site whose sleep was ever disturbed by neighborhood noise during the previous year). Part B of Table III-15 contains seven measures of the intensity of noise reactions (e.g., the percentage of respondents who were highly annoyed at having their sleep disturbed.)

The three sites in each of Groups 1 and 2 allowed nine intercomparisons for each of the ten extensity measures and seven

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

intensity measures, for a total of 153 intercomparisons. The number of times the values of these measures for each of the three noisiest sites (Group 1) exceeded the corresponding values for the three sites immediately below 70 dB(A) (Group 2) is tabulated in the middle column of Table III-15. For each of the 17 measures, Group 1 values exceed Group 2 values in six or more of the nine possible pairings. This finding establishes that reactions to noise are stronger in Group 1 than in Group 2.

A second test was devised to establish a discontinuity or critical level: a third group was created to serve as a comparison for Group 2. *

The same scoring conventions were then applied. The resulting scores are shown in the rightmost column of Table III-15. On measures of extent, Group 2 exceeds Group 3 only three times of a possible ten, and on measures of intensity, three times of a possible seven. Although the average Ldn is 3.1 dB(A)higher in Group 2, noise related reactions seem weaker than in Group 3.

If 70 dB(A) constitutes a critical level, one could predict that Group 1/Group 2 scores would be higher than Group 2/Group 3 scores. In actuality, for extensity measures Group 1/Group 2 scores exceed Group 2/Group 3 scores nine out of ten times and for intensity measures six out of seven times. This result -- 15 of 17 confirmed predictions -- would happen by chance alone
_____________
*When the seventh, eighth, and ninth noisiest sites were examined, tie was found for ninth place. The tie was resolved by averaging the two values of the pair.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

only once in a thousand times, according to the binomial sampling distribution. The greater disparity between Groups 1 and 2 than between 2 and 3 is also indicated by the finding that there were four ties between the first two pairs and 12 between the last two. Ties may be considered symptomatic of ambiguous relations.

It therefore appears that reactions to noise exposure in excess of L_dn values of 70 dB differ qualitatively from reactions to lesser exposures. In other words, the pervasiveness and strength of people's reactions to noise may grow more rapidly at exposure levels in excess of L_dn values of 70 dB than they do at slightly lower levels. Although the evidence for a discontinuity in reactions to exposure at this point is not as strong as might be desired, it seems worthy of serious consideration. It is unlikely that the present data (which include few sites with L_dn values greater than 70) would support more intensive analyses of this sort, however.

III-20      Relationship Between Noise Levels, Annoyance, and
Time of Day

This section explores the relationship between annoyance and noise exposure as a function of time of day. The annoyance information considered was the distribution of respondents who indicated noise was more annoying at one time of day (Q. 15, "Is noise in your neighborhood more annoying at one time of day than another?"). The noise exposure information was derived from continuous digital records of exposure divided into "morning", "afternoon", "evening", and "night" periods in accordance with common practice (0800-1200, 1300-1900, 2000-2200, and 2300-0700, respectively).

Return to top of document

Return to NPC Library
Return to NPC Home Page 

 a.      Social Survey Data

Seventy one percent of the ever-annoyed respondents (about one-third of all respondents) indicated that noise was in fact more annoying at one time of day than another: l4%, 205, 315, and 35% thought noise was more annoying in the morning, afternoon, evening, and night, respectively. The distribution of numbers of respondents more annoyed at the various times of day differed significantly from a chance distribution (x² 3df= 64.69, p < .01) because morning and afternoon periods were under-represented with respect to evening and night.

It would appear from this observation alone that neighborhood noise during the evening and night annoys people more than it does during the day. This observation is hardly conclusive, however, since it ignores the distribution of people at home at different times of day. As observed earlier, there are marked demographic differences (primarily number, age and sex) in neighborhood populations during the day and night that could be equally responsible for the differences in annoyance at different times of day.

  b.      Noise Exposure Data

Mean values of four measures of noise exposure (L₁, L_eq, L₉₉ and ) at all 24 sites are shown in Figure III-8 as a function of time of day. There are no meaningful differences among any of these mean values between morning and afternoon periods. All four measures dropped uniformly at night, however: the peak (L₁) by 8.6 dB, the energy mean (L_eq) by 7.8 dB, the minimum (L₉₉) by 4.5 dB, and the standard deviation ( ) by 1.1 dB, relative to their daytime values.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

  c.      Relationship Between Social and Physical Measures

Two major observations may be made about the relationship between the two types of information. First, it must be noted that the differences in percentages of respondents expressing greater morning or afternoon annoyance is not reflected in any gross physical measure of exposure. If these respondents are therefore combined into a "daytime" annoyance category, the resulting distribution of numbers of respondents in the three categories "daytime", "evening", and "night" is highly likely to have arisen by chance alone (x² 2df = l.2, p = .5); i.e., equal numbers are more annoyed in each time period.

Second, it should also be noted that these equal numbers were more annoyed during time periods of unequal duration, and despite the fact that exposure levels at night were appreciably lower than at other times of day. If annoyance per unit time is considered, the evening period (only three hours long) produces the greatest excess of annoyance. If. annoyance per decibel of exposure is considered, the night period (with levels about 7 or 8 dB lower than the day) produces the greatest excess of annoyance. The current data provide few grounds for preferring one of these viewpoints to the other.

III-21      Differences Associated with Mode of Interviewing

Both personal (face-to-face) and telephone interviews were conducted at four sites. This section examines some differences in response patterns observed in the two types of data.

One way to compare the response patterns is to correlate mean intensity scale values obtained by the two procedures. Figure III-9 is a geometric interpretation of such a comparison.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Points on the plot are for questions 19-39 at the four sites. The two dimensional space in which they are plotted represents mean intensity scale values obtained by telephone and personal interviews. If the two methods yielded identical data, all of the points would fall along the positive diagonal.

The overall mean difference for all data was 0.19, or one fifth of a response category on a five point scale. The overall product moment correlation was .73, which accounts for over half of the variance. Although these comparisons were possible for only 200 personal interviews and 300 telephone interviews, it is unlikely that data from larger numbers of personal interviews would substantially change this relationship, beyond increasing precision.

No overall pattern of differences between the two sorts of data was apparent for substantive questions. The telephone interview data showed higher percentages of respondents reporting certain attitudes, effects and noise sources, while the personal interview data showed higher percentages of respondents reporting other attitudes, effects, and noise sources. To place this difference in perspective, it should be noted that this variation was of no greater magnitude than that observed between the telephone data sites of similar population density and noise exposure.
III-22      Discussion of Sampling Bias

Cross-indexed ("reverse") telephone directories were used as sampling frames at all twenty four sites. The degree to which samples obtained from such directories are representative of the communities for which they are prepared, and the degree to

Return to top of document

Return to NPC Library
Return to NPC Home Page 

For current purposes, the fundamental issue in assessing sampling bias is whether the assumption of equal probability of inclusion in the sample of all members of the target population (English speaking adults) has been seriously violated. The most likely way for this to have occurred is by systematic exclusion of certain groups of people; i.e., those whose names fail to appear in the cross-indexed directories for various reasons. In order of estimated size, these generally include the following groups:

  1)      Telephone subscribers with unlisted numbers.

The proportion of subscribers with unlisted numbers probably varies widely with population density (i.e., lifestyle). In some urban areas, estimates of the ratio of unlisted to listed numbers are as high as 1:3 (Trendex, 1976).

 2)      Telephone subscribers with listed numbers too
                    recent for inclusion in directories.

The proportion of subscribers in this group varies with the size and stability of a community. Newcomers to neighborhoods (people establishing households since publication of the latest directory) and transients (people who move often) are the most obvious members of this group. In fact, only 35 respondents in the current survey had lived at their current addresses for six months or less.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

 3)      Non-subscribers.

Telephone subscription in urban America appears to be near universal except among persons of extremely low income. These include most notably non-English speaking ethnic minorities and the elderly.

In addition, a fourth class of people who may be under or over represented must be considered, for reasons not directly related to particular sampling frames. These are people who by virtue of time spent at home or social custom are differentially available for interviewing.

 4)      The often or rarely at home.

Differences in time spent at home are most strongly related to sex, employment, and age: housewives and the elderly are demonstrably more available for interviewing than young and working people. Systematic inclusion or exclusion of people in these four groups may produce sampling biases, whose effects on inferences drawn from interview data are usually assessed under worst-case assumptions. Thus, it is commonly assumed that all members of a misrepresented class share a common view opposite to that expressed by the population actually sampled. If observed differences, after adjustment for assumed biases, are still of significant magnitude, it is concluded that bias attributable to sampling would not affect conclusions.

Since the above four groups of people were identified in advance of data collection as likely to be under-or over represented in the current sample, measures were taken to minimize biases resulting from such deviations from representativeness.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

To minimize sampling bias due to misrepresentation by age and sex, a counterbalanced schedule for identifying potential respondents within households was used. Had it not been used, the sample would have included even fewer men than it did.

In most eases, the magnitudes of potential worst case biases attributable to sampling appear quite small. For example, although women are disproportionately numerous in the present sample (there are about 10% more women in the sample than in the adult American population), the opinions of women as a group differed very little from those of men as a group. Thus, if the sample had contained 10% more men, all of whose opinions were similar to those expressed by the 762 male respondents, the net change in mean values for substantive questions would have been negligible.

Furthermore, it is not clear that over-representation of women in the present sample should be regarded as a bias. As the data show, women spend an average of 3.5 hours more time at home than men on weekdays and weekends both. Since their exposure and knowledge about neighborhood noise is greater, their

Return to top of document

Return to NPC Library
Return to NPC Home Page 

opinions about it may be correspondingly more accurate. Similarly, it would be difficult to construe over-representation of long term residents as a serious bias for present purposes, since such people would clearly be more familiar with the local. noise environment than newly-arrived residents.

Another potential source of bias is non-response. This form of bias is usually assessed in terms of the percentage of respondents in the sample that contributed data to a survey. Response rates in excess of about 80% are usually regarded as good or excellent, while rates below 60% are usually viewed as suspiciously poor.

The percentage of respondents in the current sample that completed the interview varied from site to site. Table III-17 contains completion rates averaged over sites within cities. The overall completion rate (weighted by numbers of respondents at each site) was 70%. The bulk of the non-completions were attributable to failures to contact potential respondents, rather than refusals to answer questions.

Difficulties had been anticipated in contacting a mobile urban population; however, available resources permitted only four callbacks at different times of day. Thus, although a higher response rate would have been desirable (and probably achievable with greater resources), the 70% overall rate was adequate for present purposes.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Return to top of document

Return to NPC Library
Return to NPC Home Page 

IV-1      0n the Validity and Reliability of the Data

The overwhelming impression gained from a detailed examination of the interview data is of consistency. Sizable correlations in the directions dictated by common sense are found among all a the major variables (noise exposure, population density, annoyance, speech and sleep interference, etc.). People who described their neighborhoods as quiet suffered fewer noise effects and identified fewer sources; people who had never been annoyed by noise clearly valued the quiet nature of their neighborhoods; filers of noise complaints thought they lived in less pleasant neighborhoods; people who thought they were more sensitive to noise or spent more time in their neighborhoods suffered more from noise effects and were more alert to noise sources; and so forth. Significant counterexamples are absent.

It is also apparent that respondents gave serious consideration to the questions asked them by the interviewers. Apart from the coherence and interpretability of the answers, this can be seen most clearly in responses to the dichotomous (yes/no, quiet/noisy, noise/no-noise) questions. Proportions of respondents answering these questions in the two available response categories are compared in all data tabulations with proportions that would be expected by chance alone. If respondents had answered these questions frivolously or randomly, equal numbers of respondents in each category might have been expected. In fact, enormous departures from chance responding are uniformly found in all cross-tabulations. These observations strongly suggest that meaningful inferences may be drawn from the present data.

IV-2      0n the Predictability of Annoyance

A.      Magnitude of Correlation

The strength of the relationships between annoyance and noise exposure, population density, and speech interference are among the most striking findings of the current study. The correlation coefficients reported in section III-18, and particularly the multiple correlations. of Table III-13, are so much higher than those reported by other researchers (e,g., Tracor, 1971) that they demand closer scrutiny. In particular, a number of potential explanations for the disparity in size of correlations deserve discussion.

Perhaps the most fundamental difference between UNS and prior noise surveys (e,g., Borsky, 1965; Grandjean, 1973; NIL Research, Ltd., 1971; etc.) that could account for the improved correlation is the nature of the noise exposure under study. Most earlier research concentrated on aircraft noise, while UNS specifically avoided such exposure. It may be that aircraft noise exposure is a special case in which annoyance is either "saturated", or not strongly related to level alone (cf. Rylander et al., 1972). In the more general case of urban noise, the relationship may simply be far stronger.

Other differences between UNS and prior work readily come to mind as well. The current correlations were developed over a 20 dB range of noise exposures and a wide range of population densities, whereas most prior studies were greatly restricted in this regard. Furthermore, most prior studies were directed to discrete (transportation) noise sources rather than the entire community noise environment.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

One indication of the quality of the noise measurements of UNS is their stability over time. The product-moment correlation coefficient between the two sets of L_dn measurements made approximately one year apart at the 24 sites. was 0.88. No significant differences were found between measurements at the same site by t-test (t_{23 df} = .ll). Indeed, the mean difference in L_dn values between the two sets of measurements was only 1 dB!

Another possible source of the disparity in magnitude of correlation between UNS and earlier work is the nature of the annoyance measure. Rather than constructing an indirect index of annoyance, inferred from responses to noise-effects questions by factor analytic techniques, the current study sought to measure annoyance through direct questioning. Thus, the regression equation that uses only three variables to account for over 90% of the variance in annoyance in the present data predicts respondents' self rated annoyance, not a complex structure of assumed attitudes.

It should also be noted that the large observed correlations are for measures of the extent of annoyance in a community, not for the degree of an individual's beliefs. This measure of

Return to top of document

Return to NPC Library
Return to NPC Home Page 

The least likely explanation for the high correlation between annoyance and noise reported here is sampling error. Although it is true that the lower bound of the 95% confidence limit for a correlation of .70 calculated from 24 cases is only .45, there is little reason to believe that the observed correlation is spuriously high. Too many other relationships in the UNS data are also very strong and regular to dismiss this one correlation as a statistical fluke.

B.      Prediction Equations

Selection of variables to be used in predicting annoyance is more a pragmatic than a statistical matter. In the current situation, in which there are a number of potential predictor variables strongly related to the predicted variable, the selection may best be guided by administrative convenience. Thus, if only demographic or situational information is available, the equations in Tables III-10 or III-11 may be used. If attitudinal and noise effects information is also available, the equations of Tables III-12 or III-13 may be used.

Simply because a predictor variable does not appear in one of these equations does not imply that it is poorly related to annoyance. Quite the opposite may be the case, even though no additional variance may be accounted for by including variables

Return to top of document

Return to NPC Library
Return to NPC Home Page 

After each predictor variable is put in a regression, the step wise procedure recomputes the correlation matrix, extracting the covariance associated with the partial correlations between the remaining predictor variables and the predicted variable. Thus, the variance accounted for by each variable in the regression equation so produced is strongly influenced by its position in the equation. If the later predictor variables are related to the predicted variable in the same manner as earlier variables, they will appear to account for little additional variance, even though they may have high simple correlations with the predicted variable. There are no statistical guidelines for "best" or "unique" regression equations under these conditions; instead, one selects the variables of greatest interest for the initial positions in the equation.

IV-3      0n Noise Sources

It is interesting to note that all of the highly annoying noise sources on a national basis (Table III-5) are amenable to level-oriented regulation. Mechanical sources, rather than barking dogs and people talking in the streets, are the major noise problems in urban America.

Additionally, it appears from the analysis of Section III-17 that noise sources that do not make a major contribution to the total day-night sound level of a community nonetheless can be significant sources of annoyance. Manner of use of a noise source (e.g., motorcycles), the perceived appropriateness of a noise source,

Return to top of document

Return to NPC Library
Return to NPC Home Page 

IV-4      0n Complaints

The incidence of complaints in the UNS data is notably low -
less than a tenth of all respondents had ever complained about noise sources in their neighborhoods, even though much higher fractions of the population experienced speech and sleep interference and annoyance. Furthermore, no linear combination of major demographic, situational, and attitudinal variables was capable of reliably discriminating complainers from non-complainers. A discriminant function analysis (a statistical procedure analogous to a factor analysis in which loadings are calculated for arbitrarily specified dimensions) was able to correctly identify only 62% of all cases as complainers or non-complainers, a result that does not differ significantly from chance.

Perhaps the single most important factor that may account for both the small number and unpredictability of complaints is the lack of opportunity for complaining characteristic of the urban noise exposure situation. In airport neighborhoods, there are abundant opportunities for complaints about aircraft noise; indeed, special agencies often exist for the purpose of collecting complaints and taking action on them. To whom is an urban dweller to complain about a passing motorcycle, a noisy automobile on the next street, or a bus? What good would it do to complain to the police about occasional sleep interference from a police helicopter? Given these constraints on the utility of complaint behavior, it would seem wise to avoid

Return to top of document

Return to NPC Library
Return to NPC Home Page 

over-reliance on complaint rates as an index of noise impact in urban areas.

IV-5      0n the Relation Between Annoyance and
          Demographic Variables

The major demographic variable that is strongly related to annoyance is population density. Care should be taken not to interpret this relationship as a causal one, since it is well known that population density correlates highly with noise exposure. In the current sample of 24 sites selected in a manner that would tend to minimize the association between population density and noise exposure (a wide range of population density sites was purposely chosen for each noise exposure level), the correlation was 0.55.

Other demographic variables, such as age, sex, and duration of residence in a neighborhood, contribute little predictability to the relationship between annoyance and either noise exposure or population density. Income and socioeconomic level are somewhat more closely related to annoyance than other demographic variables, but not in a causative manner. Income and socioeconomic level are highly related to one another, and both are inversely related to neighborhood noise levels (as may be seen in Figure III-5).

Thus, noise exposure, like other forms of environmental pollution, does not affect all segments of society equally. It is not that the ears of the high socioeconomic level respondents are more or less sensitive than those of other segments of society; they simply can afford to live in quieter neighborhoods. The fact that neighborhood satisfaction is inversely related to noise exposure but directly related to income and socioeconomic level suggests that quiet is a valuable attribute of neighborhoods. As may be seen in Section III-10,

Return to top of document

Return to NPC Library
Return to NPC Home Page 

those who are most highly annoyed are not at all confused about this issue; more of the highly annoyed found their neighborhoods noisy and not especially pleasant to live in, were thinking of moving, and spontaneously mentioned noise as the least liked aspect of their neighborhoods.

IV-6      0n the Relationship of Current Findings to
          Prior Findings

Figure IV-1 plots the relationship (regression equation) between annoyance and noise exposure derived from the current data on the same axis as plots derived from two other sources. The upper plot compares the UNS data with the relationship found in EPA's "Levels Document". The latter relationship was derived principally from two aircraft noise surveys. It is apparent that the aircraft noise data greatly overestimate the annoyance found in general urban noise environments.

The lower plot of Figure IV-I compares the UNS data with a synthesis of all major social survey data prepared by Schultz et al. (1976). The curve of Schultz et al. (1976) resembles the UNS data far more closely than the aircraft noise curve of the Levels Document. Indeed, disparities between the two curves in the lower plot are readily attributable to errors of prediction in the two regression equations. The reader is referred to Schultz et al. (1976) for a fuller discussion of the methods whereby prior social survey data were manipulated to derive a composite function.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

The following are among the major conclusions that may be supported by inferences drawn from the data of the national Urban Noise Survey.

1.      Exposure to noise levels typical of many urban (non-aircraft, non-highway) environments produces widespread annoyance, speech interference, and sleep interference in the American public.

2.      The relationship between exposure level and the proportion of a community highly annoyed by noise is strong and predictively useful.

3.      The prevalence of speech interference in a community is an especially good predictor of the prevalence of annoyance.

4.      Population density is another important correlate of noise exposure that may be useful as a surrogate for physical exposure in predicting the prevalence of annoyance.

5.      The proportion of the population exposed to urban (non-aircraft, non-highway) noise that complains about the exposure is a poor predictor of the prevalence of annoyance.

6.      Demographic factors alone (age, sex, income, socioeconomic level, duration of neighborhood residence, etc.) are relatively poor predictors of annoyance.

7.      The public is aware that noise exposure degrades the quality of urban living, inasmuch as freedom from exposure is a component of neighborhood satisfaction, and quiet is highly valued.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

8.      Noises associated with automotive sources are the most pervasive sources of annoying noise exposure in urban America.

9.      Annoyance associated with intrusive noise sources may be related to measurable noise exposure from such sources, even when their magnitudes are not as great as the overall exposure levels in a community.

10.      There is some evidence that human response to noise exposure at L_dn values in excess of 70 dB is more acute than at lower exposure levels.

11.      Although annoyance due to noise exposure is more prevalent during the evening and night periods than during the day, the current data do not support any clear inferences about the magnitude of a nighttime noise exposure penalty.

12.      People of high socioeconomic level suffer less noise exposure and are more satisfied with their neighborhood environments than people of lower socioeconomic level.

Return to top of document

Return to NPC Library
Return to NPC Home Page 

Borsky, P. N., "Community Reactions to Sonic Booms in the Oklahoma City Area, Vol. III: Questionnaires - Appendix to Vol. II," Report No. AMRL TR 65-37, Wright Patterson Air Force Base, Ohio, March, 1966.

Environmental Protection Agency, "Information on Levels of Environmental Noise Requisite to Protect Public Health and Welfare with an Adequate Margin of Safety", Report No. 550/ 9-74-004, March, 1974.

Galloway, W., Eldred, K., and Simpson, M., "Population Distribution of the United States as a Function of Outdoor Noise Level", EPA Report 550/9-774-009, June, 1974.

Grandjean, E., Graf. P., Lauber, A., Meier, H. P. and Muller, R. A., "A Survey of Aircraft Noise in Switzerland". Proceedings of the International Congress on Noise as a Public Health Problem, Dubrovnik, 1973. Washington, D. C.: U. S. Environmental Protection Agency Report No. 550/9-73-008, 645-659.

Jones, G., and Galloway, W., "Motor Vehicle Noise: Identification and Analysis of Situations Contributing to Annoyance", BBN Report 2082, June, 1971.

MIL Research, Ltd., "Second Survey of Aircraft Noise Annoyance Around London (Heathrow) Airport", Report No. SS 394, Her Majesty's Stationery Office, London, 1971.

Rylander, R., Sorensen, S., and Kajland, A., "Annoyance Reactions from Aircraft Noise Exposure", J. Sound and Vibration, 24, 4, 419-444, 1972.

Schultz, T., Galloway, W., Beland, D., and Hirtle, P., "Recommendations for Changes in HUD's Noise Policy and Standards", BBN Report 3319R, November, 1976.

Simpson, M., Pearsons, K., Fidell, S., and Muehlenbeck, R., "Social Survey and Noise Measurement Program to Assess the Effects of Noise on the Urban Environment: Data Acquisition and Presentation", BBN Report 2753, July, 1974.

Tracor, Inc., "Community Reaction to Airport Noise", NASA CR-1761, July, 1971.

Trendex, Inc., "A Comparison of Phone Book Samples and Random Digit Dialing Samples", presented at 22nd Annual Conference of the Advertising Research Foundation, New York, October, 1976.

Return to top of document