Free Response to Motion - District Court of Federal Claims - federal

ase 1:01-cv-00201-VJW Economics and Policy, Volume 38, Filed January 2004, pp. 1­28Page 1 of 2 Document 219-4 Journal of Transport Part 1, 10/05/2006

Meta-Analysis of Airport Noise and Hedonic Property Values
Problems and Prospects

Jon P. Nelson

Address for correspondence: Jon P. Nelson, D epartment of Economics, Pennsylvania State University, University Park, PA 16802 U SA.

Abstract M eta-analysis is applied to the negative relationship between airport noise exposure and residential property values. The effect size in the analysis is the percentage depreciation per decibel increase in airport noise, or the noise discount. Twenty hedonic property value studies are analysed, covering 33 estimates of the noise discount for 23 airports in Canada and the United States. About one-third of the estimates have not been previously reported in the literature or were not included in previous meta-analyses. A meta-regression analysis examines the variability in the noise discounts that might be due to country, year, sample size, model speci cation, mean property value, data aggregation, or accessibility to airport employment and travel opportunities. The analysis indicates that country and model speci cation have some effect on the measured noise discount, but the other variables have little systematic effect.

Date of receipt of ®nal manuscript: July 2003

1

ase 1:01-cv-00201-VJW Document 219-4 Journal of Transport Economics and Policy

Filed 10/05/2006 38, Part 1 2 of 2 Page Volume

1. Introduction
Aircraft noise continues to be an issue at many airports, especially where capacity expansions are under way or are being considered. In the U nited States, airport noise exposure has been reduced through operating requirements, quieter aircraft, and by soundproo ng or purchase of surrounding residential properties. As a result, the past 25 years have seen a dramatic decline in the number of persons exposed to signi cant noise levels of 65 decibels (dB) or more.1 H owever, from January 2000, the entire U S commercial eet had been converted to Stage 3 aircraft and further mitigat ion of noise at the source will require investments in improved technologies by airlines. The projected growth of air travel could also slow or reverse the decline in the exposed population, although currently there is considerable uncertainty in these forecasts. According to the F ederal Aviation Administration, 75 per cent of large hub airports and 47 per cent of transfer airports have proposed or begun building new runways (F AA, 2002). Because noise is the number one environmental concern at major airports, capacity expansion is often slowed by public concern with noise exposure (G AO, 2000a, 2000b).2 This concern highlights the importance of valuation of noise effects if expansion decisions are based on economic ef ciency criteria and bene t-cost analysis. M easurement of the economic value of quietn ess has traditionally focused on the effect of signi cant noise exposure on residential property values. U sing hedonic price analysis, a number of studies have measured this value empirically for airports located in Australia, Canada, N etherlands, U nited K ingdom, U nited States, and other countries. F irstgeneration studies for the U S employed aggregate census track and block data, while second-generation studies made use of sales data for individual houses. It can be anticipated that the next generation of hedonic studies will use geographical information systems (G IS), which has been
1

Since 1979, federal agencies have considered a Day-Night Average Sound Level (D NL) of 75 dB or greater as incompatible with all residential use, except transient lodging (F AA, 1985, 2000; F ICU N, 1980; F ICON , 1992a). U nder the Aviation Safety and Noise Abatement Act of 1979, the F AA adopted the DNL metric and 65 dB compatibility standard. Lands exposed to D NL 65­74 dB are regarded as ``normally'' incompatible with residential use, while lands exposed to a DN L of less than 65 dB are regarded as ``normally'' compatible with such use. F urther, at 65 dB and above, increases in exposure of 1.5 dB or more are regarded as a signi cant addition of noise, and require an environmental impact statement. 2 These problems also exist at smaller airports or where airport conversions have been proposed. See, for example, the controversy in Orange County, CA, surrounding the proposed conversion of the El Toro M arine Air Station to an international airport to compete with Los Angeles International Airport (http://www.eltoroairport.org).

2

ase 1:01-cv-00201-VJW Noise Meta-Analysis of Airport Document 219-4

Filed 10/05/2006

Page Nelson

3 of 2

applied to noise generated by road traf c (Bateman et al., 2002; Lake et al., 2000). Similarly, contingent valuation methods have been used to analyse airport noise valuation, but at present only one such study exists for Canada or the U S (F eitelso n et al., 1996; see also N avrud, 2002). In the light of these recent developments, a rst objective of this study is to provide a benchmark that can be compared against results from studies using alternative methodologies. H edonic price models have been used to estimate the effects of numerous amenities and disamenities on the value of residential housing.3 These models exploit the differentiation that exists in housing markets in terms of locational attributes or characteristics of a sample of properties. The model is especially useful if the attribute in questio n is ``localised'' so that only a small fraction of the properties within a metropolitan area is affected or exposed (Palmquist, 1992a). F or example, noise from a major highway or an airport affects adjacent and nearby properties, while leaving unaffected the general level of urban values. 4 In this case, the hedonic price of the attribute identi es a marginal bene t or willingnessto-pay schedule that is useful for the evaluation of public policies. Several analytical issues remain regarding: (1) the model speci cation used to separate the effect of noise from numerous other possib le in uences on property values; and (2) the transferability of noise abatement bene t estima tes across urban markets, regions, or countries. In particu lar, proximity to an airport also provides access to travel and employment opportunities, and a major airport might be expected to have both positive and negative effects on property values. Ignoring accessibility could result in a downward bias for the effect of noise alone. F urther, bene ts transfer issues arise with respect to noise abatement projects at speci c airports; ``full cost'' estimates for alternative modes of transport (G reene et al., 1997; Levinson et al., 1998; N AS, 2002); and bene t-cost analyses of major public policies (M orrison et al., 1999). G iven a large number of hedonic studies of airport noise, problems of model speci cation and bene t transfers can be addressed using metaanalysis techniques. This paper presents a meta-analysis of the effect of airport noise on property values in the vicinity of civilian airports in
3

There are a large number of narrative surveys of the hedonic price literature. F or surveys of the general model, see Palmquist (1992b) and Sheppard (1999). F or recen t surveys of externality issues, see Boyle and Kiel (2001), EPA (2000), F reeman (2003), and Palmquist and Smith (2002). F ollain and Jimenez (1985) contains a survey of empirical results from two-step estimation of structural demand functions for attributes. 4 M orrison et al. (1999, p.734) examined noise exposure maps for 35 major airports in the US. They concluded that airport noise typically affects less than 2 per cent of the total number of housing units in a metropolitan area.

3

ase 1:01-cv-00201-VJW Document 219-4 Journal of Transport Economics and Policy

Filed 10/05/2006 38, Part 1 4 of 2 Page Volume

Canada and the U S. Twenty different studies are included in the analysis, providing 33 estima tes of the effect of noise. About one-third of the estimates have not been previously reported in the property value litera ture or included in simila r analyses. Average values for the estima tes are presented and the variation in the estimates is analysed using ``metaregression'' analysis (Stanley, 2001; Stanley and Jarrell, 1989). Because two previous meta-analyses have considered subsets of the availab le estimates, the paper also comments on the problems that exist in these studies. The remainder of the paper is divided into four sectio ns. The next section provides a brief background on hedonic price studies of airport noise and discusses several measurement problems that can arise due to model speci cation, accessibility, and bene ts transfer. This is followed by a discussio n of meta-analysis of environmental effects and the average values for the sample of twenty studies. A summary of the shortcomings found in two previous meta-analyses of airport noise also is presented. In the third section, a meta-regressio n analysis is presented, which sorts out the possible sources of variation in the values. The conclusions from the study are found in the fourth section. Two appendices provide references for the empirical studies included in the analysis and those studies excluded and the reasons for the exclusions.

2. Airport Noise and Property Values
N oise is unwanted or unpleasant sound. At 65 dB and above, the most common human effects associated with aircraft noise are annoyance, speech and learning interference, and sleep distu rbance. In turn, these effects disrupt normal daily activities, such as conversation, televisio n viewing, school work, productivity, outdoor recreation and living, and family activities. Annoyance is the adverse psychological response to a given noise exposure, including the anxiety or apprehension that the noise may cause (F ICON , 1992b). At noise levels above 75 dB, the Environmental Protection Agency (EPA, 1982) cautions that more severe health effects may occur for some portion of the population, including temporary hearing loss. Those persons who are frequently outdoors are of greatest concern, including young children, retired people in warm climates, and people in certain outdoor occupations.

4

ase 1:01-cv-00201-VJW Noise Meta-Analysis of Airport Document 219-4

Filed 10/05/2006

Page Nelson

5 of 2

2.1. Hedonic model and noise annoyance Consider two residential properties that are identical in all respects, except that one house is located close to or under an aircraft ight path, and the other is not. A ``but for analysis'' establishes that the adverse environment for the rst house will result in a market value that is lower than the market value of the second house. This occurs because potential buyers reduce their demand for the rst house relative to the second house, re ecting the discounted present value of the costs of annoyance, loss of tranquillity, and possible health effects. A measure of the noise-induced damages is the difference between the market-determined value of the two houses. The analysis can be extended to analyse different levels of noise exposure because annoyance and other adverse effects of noise rise predictably with increased exposure levels (EPA, 1982; F AA, 1985; F ICON , 1992a, 1992b). H ence, while there is a missin g market for tranquillity, a complementary market exists wherein individuals register their willingness to pay to avoid different levels of aircraft noise exposure. Consumers thus reveal the implicit value that they place on quietness by the explicit choices that they make in the housing market. The willingness to pay for quietn ess and other amenities are part of the asset price of the ``housing bundle'', and econometric techniques are available that unbundle complex products and thereby reveal the implicit or hedonic price. As indicated above, a large empirical liter ature has developed using the hedonic method. It is rare that two residential properties will be identical in all respects, except for the pollutant in questio n. Consequently, in order to isolate a given hedonic price, it is necessary to control statistica lly for other in uences on property values, such as the size of house and lot, quality of constru ction, design of the house, merits of the neighbourhood, quality of local schools, crime rates, governmental services, and so on. 5 Some of these characteristics will vary little within a given data set, and separate measurement is not required to explain the observed variation in property values. In other cases, the excluded characteristics are uncorrelated or orthogonal to noise levels, and exclusion will not bias the resulting
5

A referee argued that the discount associated with an externality, such as noise, could be affected by general housing market conditions in an urban area, such as the presence of strong demand for housing in a ``hot'' market. Although the direction of the effect is uncertain, this is a general equilibrium issue for which no published empirical evidence seems to exist for aircraft noise. As such the issue is partly one of aggregation across areas that may not be homogeneous with respect to, say, the general level of housing price appreciation. However, as indicated above, properties that are in close proximity to an airport are generally a small proportion of the total housing market for an urban area. F or discussion of aggregation issues associated with hedonic property value models, see Bartik and Smith (1987), F reeman (2003), and Palmquist (1992a, 1992b).

5

ase 1:01-cv-00201-VJW Document 219-4 Journal of Transport Economics and Policy

Filed 10/05/2006 38, Part 1 6 of 2 Page Volume

estimate of the hedonic price. G iven differences in statist ical methods, samples, time periods, and urban locations, empirical studies have not produced a singular value for the effects of airport noise on property values. H owever, hedonic price studies have shown that airport noise has a negative impact on residential property values, and central tendencies can be determined based on the distribution of estimates. F urther, metaanalysis can establish the extent to which the variation is systematic. Each house and lot represents a unique combination of characteristics and locational attributes, which means that the decisio n to purchase a given property is complex. H owever, if the characteristics and attributes are provided in various combinations, it is possible to estimate an implicit price function that shows how these values vary conditionally on the distr ibution of a given characterist ic. F ormally, let V be a sample of observations on housing prices; S is a vector of stru ctural variables (house size, number of bedrooms, lot size, and so on); L is a vector of locational variables; T is a vector of local taxes; G is a vector of local government services; and E is a vector of localised environmental-quality attributes. The market-determined asset value of houses in the sample is given by V D V(S; L; T; G; E): The marginal implicit price for each characterist ic represents the increase in expenditures required to obtain one more unit of that characteristic, holding constant all other variables. The marginal price is the partial derivative of the V relationship with respect to the j-th characteristic, or @V=@Ej : A given marginal price could be a constant, but generally it will be a function of the data. M any empirical studies employ a non-linear function for property values, either a log­log or a semi-log relationship. The hedonic function for properties in the vicinity of an airport can be represented by V D b0 Zb1 Ab2 u1 ; where V is the property value; Z is all other physical and locational characteristics (that is, S; L; T; and G); A is annoyance due to aircraft noise; u1 is a stochastic error term; and b0 ; b1 ; and b2 are parameters. Annoyance can be approximated by the following semi-logarithmic relationship: A D c0 ec1(DNL) u2 ; where DNL is the D ay­N ight Sound Level in decibels; u2 is a stochastic error term; e is the natural log base; and c0 and c1 are parameters (N elson, 1980).6 Taking logs of both relationships, and substituting for ln A; yields the following relationship for property values: ln V D d0 C d1 (ln Z) C d2 DNL C u3 ; where d2 D b2 c1 ; and so on. The regression coef cient d2 £ 100 D (@V=@DNL ² 1=V) £ 100 represents the
6

The ``Schultz curve'' displays the percentage of exposed persons highly annoyed as a function of environmental noise (F inegold et al., 1994). The Schultz curve includes low levels of noise exposure (e.g., 40­50 dB), which is the non-linear left-hand tail of the curve. These noise exposures are typically not due to aircraft noise. The semi-log approximation of the Schultz curve is from Bishop (1966).

6

ase 1:01-cv-00201-VJW Noise Meta-Analysis of Airport Document 219-4

Filed 10/05/2006

Page Nelson

7 of 2

percentage decrease in a given property value due to a one dB increase in noise exposure on the DNL scale. 7 F or the log (or semi-log) functio n, the marginal implicit price is P D (@V=@DNL) D d2 V: In other words, the marginal price for noise increases as a function of the property value V: M eta-analysis requires a common effect size measure of damages due to airport noise. That is, the outcomes from the set of studies must be expressed in terms of a common metric and its standard error. With some adjustments, the ndings of empirical studies of airport noise can be summarised by means of a N oise D epreciation Index (N D I), which is the percentage rate of depreciation per dB (Walters, 1975). F or two properties that differ except for their level of noise exposure, the absolute amount of housing depreciation per decibel (the unit cost of noise) is given by D D (difference in the total noise discount) ¥ (difference in noise exposure in dB). D ividing D by the price of the given house (or an average house price), the percentage rate of depreciation is given by NDI D (D ¥ property value) £ 100 D (difference in total percentage depreciation) ¥ (difference in noise exposure in dB). This is the same result as d2 £ 100 above. 2.2. Functional form problems M any empirical studies use the logarithmic (or semi-log) model outlined above, and these studies directly estimate the N D I. The regression coef cient on D N L multiplied by 100 is the percentage decrease in a property value due to a one dB increase in noise exposure. F or example, if the N D I is 1 per cent, a property exposed to 70 dB is worth 10 per cent less than a property exposed to 60 dB, and so on. H owever, some studies use a linear functional form, where V D f0 C f1 Z C f2 (DNL) C u: H ence, an estima te of the discount is NDI D f2 (1=VM ) £ 100; where VM is, say, the mean value for the sample. As a result the N D I estimates for studies using a linear form might be systematically different from studies using a log speci cation. In the meta-analysis a dummy variable is included for those studies that used a linear functional form. 2.3. Threshold noise problems Because the ear's pattern of response is more nearly logarithmic in nature, decibels are measured on a log scale. The perception of noise doubles in
7

This functional form allows the marginal price to depend on the levels of other characteristics of the property. As pointed out by a referee, there is still a possibility that households might substitute among other activities in the face of increased noise exposure, such as reducing the amount of outdoor activity. Some evidence to the contrary is provided by H oyt and R osenthal (1997), who demonstrate that US households appear to sort ef ciently based on preferences for locational amenities.

7

ase 1:01-cv-00201-VJW Document 219-4 Journal of Transport Economics and Policy

Filed 10/05/2006 38, Part 1 8 of 2 Page Volume

loudness for every 10 dB increase in sound level. An 80 dB sound is perceived to be twice as loud as a 70 dB sound, four times louder than a 60 dB sound, and eight times louder than a 50 dB sound. H owever, the absence of aircraft noise is not associated with a zero value of the D ay­ N ight Sound Level. The D N L is the average sound level generated by all major environmental noise sources during a 24-hour period, with a 10 dB penalty for nighttime noise events. 8 N oise levels in the vicinity of airports range from about 65 to 80 dB, and noise contour data are generally available in 5 dB increments for this range. Typical background noise levels in urban areas are about 50­60 dB during daytime hours and 40 dB during nighttime. (A level of 50­60 dB corresponds to the noise from light traf c at 100 feet or an air conditioner at 100 feet.) A few studies incorrectly account for the threshold noise level. In some cases, they simply ignore the background level of urban noise and treat it implicit ly as zero, rather than 55 or 60 dB. This is especially true where the sample of property values covers a larger portion of an urban area, which leads to a relatively large sample size. A control variable for sample size is used to correct for this problem in the meta-analysis.

2.4. Accessibility speci®cation problems M ajor airports are commercial facilities that have the potential to create signi cant travel and employment opportunities. Employment opportunities exist at the airport site as well as at commercial facilities which develop in the vicinity of a major airport. F or individuals who might work at (or near) the airport or who use the airport for travel, the bene ts of proximity can be re ected in residential property values. Because it is possible for an airport to have negative and positive effects on property values, the net effect can be negative or positive. The empirical problem is the extent to which a particular empirical study has separated out the effect of noise from the effect of accessibilit y (if any). F ailure to allow for accessibility could lead to a downward bias in the hedonic price of airport noise. Previous studies have addressed the accessibility problem in a variety of ways. D eVany (1976) was the rst to investigate this issue, and he proposed a solution using a dummy variable speci cation. N elson (1979) suggested another solution based on the elongated shape of aircraft noise
8

Some empirical studies used an older noise metric, the N oise Exposure F orecast (NEF ) in A-weighted decibels. The two metrics are highly correlated. The relationship between D NL and NEF is given by DNL D NEF C 35 dB (EPA, 1982). The summaries below re ect this adjustment of noise metrics, and all noise levels are expressed using the DN L metric, regardless of the original metric used in the study.

8

ase 1:01-cv-00201-VJW Noise Meta-Analysis of Airport Document 219-4

Filed 10/05/2006

Page Nelson

9 of 2

contours and sampling for limit ed areas with more or less the same degree of accessibilit y. Li and Brown (1980) examined the general effects of disamenities and accessib ility on property values in the G reater Boston metropolitan area. Several studies of the M anchester Airport (U K ) have reached con icting conclusions about the importance of accessibility and noise. In particular, Tomkins et al. (1998) used straight-line dista nce to the airport as a measure of accessibilit y. The N D I was 0.78 per cent, but they found that the effect of accessib ility was greater for certain properties. H ence, for some properties, the net effect of the M anchester Airport on property values was positive. The purpose of the present analysis is to isolate the effect of noise on property values. Each of the twenty studies was examined for controls on accessibility to the airport. In general, three control methods have been used in past studies. F irst, many studies exploit the elongated nature of noise contours and select a sample area that holds accessibilit y constant and allows noise levels to vary importantly. This approach is illustrated by F igure 1, which shows elongated noise contours and circular rings for access to the airport. N ote that some properties can have a high degree of proximity, but low or moderate noise levels. As a result, the correlation between noise and accessibility is not necessarily high.9 Second, some studies use D N L noise contours and straight-line dista nce to the airport as a measure of accessib ility. This solution tends to ignore commercial opportunities that are not located at the airport. Third, dummy variable speci cations have been used to isolate the effects of aircraft noise. In the meta-analysis, a dummy variable is included for those studies that entirely ignored the accessibility problem. 2.5. Bene®t transfer problems Bene t transfer is the general problem of using damage estima tes based on primary data from a speci c study (or studies) to estimate the bene ts of abatement efforts at another settin g or location (Brookshire and N eill, 1992; D esvousges et al., 1998). Bene t transfers assume that the parameter estima te is applicab le in both areas, which requires that the parameter is drawn from a common population, or the estimate can be adjusted to be applicab le to the transfer area. Because bene t transfers reduce the costs
9

Several of the excluded studies used only the straight-line distance to the airport as a measure of the noise impact of the airport (see Appendix B). The dif culty with this approach is illustrated by F igure 1, which depicts a low correlation between noise and access. F or example, using a sample of 424 individual property sales from DeArau jo (1986), I calcu lated the correlation between aircraft noise exposure and straight-line distance to the Lambert St. Louis International Airport. The simple correlation between noise and distance was only ¡0:024:

9

se 1:01-cv-00201-VJW Document 219-4 Journal of Transport Economics and Policy

Filed 10/05/2006 38,Page 10 of Volume Part 1

Figure 1 Airport Noise Contours and Accessibility Rings

D NL 65 DNL 70

of decisio n making, the central issue is the acceptable degree of variability for a set of estima tes to be a reliable guide for public policy. In this context, using all of the available estimates of noise damages is probably unwarranted and undesirable. H ence, the present study examines: (1) hedonic price studies of airports in Canada and the U S; and (2) studies based on data from 1967 and later. These sample restrictio ns also facilita te the collection of data for the meta-analysis.

3. Meta-Analysis of Twenty Studies: Basic Results
M eta-analysis was developed to address the general issue of research synthesis. 10 A meta-analysis combines the ndings of studies to assess the

10

Meta-an alysis has seen fairly wide applications in economics, especially for non-market valuation studies in the environmental, natural resource, and transportation areas (Smith and Pattanayak, 2002; van den Bergh et al., 1997). R epresen tative applications include Smith and Huang (1995) on air pollution; Espey et al. (1997) and Dalhuisen (2003) on residential water deman d; Rosenberger and Loomis (2001) on outdoor recreatio n; and Waters (1996) on value of time savings. Standard reference for technical aspects of meta-analysis are Hedges and Olkin (1985), Hedges (1992), and Lipsey and Wilson (2001).

10

se 1:01-cv-00201-VJW Noise Meta-Analysis of Airport Document 219-4

Filed 10/05/2006

Page Nelson

11 of

magnitude and signi cance of a measure of effect size. Among the hallmarks of a high quality meta-analysis are comparability, completeness, and transparency. A meta-analysis compares studies to identify possible moderators of effect size, such as the sample characteristics, time period, model speci cation, or location. M eta-analysis also can be used to assess the consistency of research ndings by determining the extent to which variation in ndings is systematic or due to random factors. H owever, caution is required regarding the set of estimates that form the basis for the analysis. F or example, Johnson and Button (1997) examined a sample of noise damage estimates that included airports in Australia, Canada, U K , and U S. They also included studies using data from the early 1960s as well as more recent data sets. N ot surprisingly, they concluded that the range of damage estimates is wide. The position in the present study is that a more complete and controlled applicat ion of meta-analysis is required before the existing research in this area is cast aside or labelled inadequate for bene t transfers and public policy analysis. Additional comments on previous meta-analyses are presented after the summary of the twenty studies.

3.1. Summary of the twenty studies Table 1 displays information from twenty studies of property values in the vicinity airports in Canada and the U S. A number of new studies were uncovered by searches using web-based resources, especially Dissertation Abstracts (http://wwwlib.umi.com/dissertations). These studies provide 33 estima tes of the N D I for 23 different airports. Except for R eno and R ochester, the airports are major facilit ies for all regions of each country. Some studies use the same data set or cover different aspects of the same study (for example, the two studies of the Vancouver Airport). In order to avoid problems of statistica l dependence, only one estima te is used from these joint cases. U sing independent samples, several airports have been studied more than once, including Atlan ta, D allas, R eno, St Louis, San F rancisco, and Washington, D C. Table 1 also presents relevant information for each study that is used in the meta-regression analysis. This information is discussed below. Assume N independent studies, each yielding an estimate, Di ; of the noise depreciation index Ii ; where i D 1; 2; . . . ; N: In meta-analysis, this outcome is labelled the ``effect size'' estimate. A ``no stru cture'' model or xed-effects analysis assumes that I1 D I2 D . . . D IN : That is, each study is estima ting an identical, but unknown, true population value of I: F ollowing H edges and Olkin (1985), the asymptotically ef cient estimator of I is a weighted-mean of the individual effects Di ; with the optimal

11

se 1:01-cv-00201-VJW Document 219-4 Journal of Transport Economics and Policy

12
Airport (& Area) ind. prop. ind. prop. ind. prop. ind. prop. census blocks census blocks census tracts census tracts ind. prop. census tracts ind. prop. ind. prop. ind. prop. Study Period (Smpl Size) Data Type Mean Property Value (2000 US $) NDI %: Abs Value (std err) Dep. Variable (R -sq) Access Adjust? Yes (1) Yes (1) Yes (1) Yes (1) Yes (2) Yes (3) Yes (2) Yes (2) Yes (1) Yes (1) Yes (2) No No 1990 (30) 1991 (24) 1993 (30) 1993 (30) 1970 (4,264) 1970 (1,270) 1970 (82) 1970 (98) 1967 (222) 1970 (28) 1991­95 (1,596) 1985­86 (1,635) 1969­70 (6,553) $123,698 ($170,703) $351,062 ($449,359) $422,500 ($523,900) $222,534 ($275,942) $25,000 ($136,250) $22,000 ($119,900) $27,600 ($150,420) $21,000 ($114,450) $19,683 ($132,270) $30,068 ($163,871) $110,970 ($137,603) $72,316 CN $ ($70,104) $15,015 ($81,832) 1.07 (0.823) 1.26 (0.788) 1.20 (na) 0.67 (na) 0.99 (0.330)¤ 0.80 (0.267)¤ 0.50 (0.250)¤ 0.70 (0.422)¤ 0.58 (0.366) 1.49 (0.753)¤ 0.28 (0.183) 1.30 (0.342)¤ 0.56 (0.240)¤ linear (0.91) linear (0.83) linear (0.77) linear (0.57) linear (0.82) linear (0.82) log (0.66) log (0.67) log (0.80) log (0.75) log (0.85) log (0.80) log (0.67)

Table 1 Meta-Analysis of Airport Noise and Hedonic Property Values (NDI D % depreciation per decibel)

Study (Publication Date & Page no.)

BAH-F AA (1994, p18)

Baltimore

BAH-F AA (1994, p22)

Los Angeles

BAH-F AA (1994, p27)

BAH-F AA (1994, p27)

Blaylock (1977, p79)

New York (JF K) New York (La Guardia) Dallas

D eVany (1976, p213); N AS (1977, p139) D ygert (1973, p105)

Dallas

D ygert (1973, p113)

San F rancisco (San M ateo) San Jose

Emerson (1969, p68; 1972, p271)

M inneapolis

F romme (1978, p100)

K aufman (1996, p33); Espey and Lopez (2000) Levesque (1994, p207)

Wash. DC (National) R eno

Winnipeg

Filed 10/05/2006 38,Page 12 of Volume Part 1

M ark (1980, p112)

St. Louis

Table 1 continued
Airport (& Area) ind. prop. ind. prop. ind. prop. ind. prop. ind. prop. ind. prop. census tracts census blocks census blocks census blocks census blocks census blocks census blocks census blocks Study Period (Smpl Size) Data Type Mean Property Value (2000 US $) NDI %: Abs Value (std err) Dep. Variable (R-sq) Access Adjust? No No No Yes(1) Yes(1) No Yes(1) Yes(1) Yes(1) Yes(1) Yes(1) Yes(1) Yes(1) Yes (2)

Study (Publication Date & Page no.)

M aser et al. (1977, p130); Quinlan (1970)

M aser et al. (1977, p130); Quinlan (1970)

M cM illan et al. (1980, p319); M cM illan (1979)

Rochester (urban) Rochester (suburban) Edmonton

M ieszkowski and Saper (1978, p430)

Toronto (M ississauga)

M ieszkowski and Saper (1978, p430)

M yles (1997, p21)

Toronto (Etobicoke) Reno

se 1:01-cv-00201-VJW Noise Meta-Analysis of Airport Document 219-4

Nelson (1978, p98)

Nelson (1979, p325; 1980, p45)

Wash. DC (National) Buffalo

Nelson (1979, p325; 1980, p45)

Cleveland

Nelson (1979, p325; 1980, p45)

New Orleans

Nelson (1979, p325; 1980, p45)

St. Louis

Filed 10/05/2006

Nelson (1979, p325; 1980, p45)

San Diego

Nelson (1979, p325; 1980, p45)

San F rancisco

Nelson (1979, p327; 1980, p69; 1981)

Six airports

1971 (398) 1971 (990) 1975­76 (352) 1969­73 (509) 1969­73 (611) 1991 (4,332) 1970 (52) 1970 (126) 1970 (185) 1970 (143) 1970 (113) 1970 (125) 1970 (153) 1970 (845)

$19,100 ($99,893) $21,800 ($114,014) $51,933 CN$ ($108,730) $31,450 CN$ ($89,982) $37,770 CN$ ($108,063) $135,000 ($178,200) $27,455 ($149,630) $20,656 ($112,575) $20,898 ($113,894) $21,975 ($119,763) $16,411 ($89,440) $32,241 ($175,713) $29,686 ($161,789) $23,713 ($129,236)

0.86 (0.319)¤ 0.68 (0.279)¤ 0.51 (0.224)¤ 0.87 (0.212)¤ 0.95 (0.187)¤ 0.37 (0.111)¤ 1.06 (0.714) 0.52 (0.200)¤ 0.29 (0.128)¤ 0.40 (0.195)¤ 0.51 (0.267)¤ 0.74 (0.233)¤ 0.58 (0.184)¤ 0.55 (0.200)¤

linear (0.62) linear (0.84) log (0.71) log (0.90) log (0.92) log (0.74) log (0.86) log (0.61) log (0.89) log (0.75) log (0.74) log (0.76) log (0.71) log (0.84)

Page Nelson

13

13 of

se 1:01-cv-00201-VJW Document 219-4 Journal of Transport Economics and Policy

14
Airport (& Area) census blocks ind. prop. census tracts ind. prop. ind. prop. ind. condos Study Period (Smpl Size) Data Type Mean Property Value (2000 US $) NDI %: Abs Value (std err) Dep. Variable (R -sq) Access Adjust? No Yes(1) No No Yes (2) Yes (2) 1970 (248) 1979­80 (96) 1970 (270) 1989­90 (427) 1987­88 (645) 1987­88 (907) $18,964 ($103,354) $28,889 ($81,178) $103 per month (na) $148,525 CN$ ($118,985) $139,100 CN$ ($124,076) (na) log (0.74) log (0.71) linear (0.50) linear (0.64) log (0.64) log (0.79) 0.64 (0.200)¤ 0.67 (0.300)¤ 0.81 (0.238)¤ 0.65 (0.325)¤ 0.65 (0.164)¤ 0.90 (0.323)¤ 0.75 (0.295)¤ 0.67 0.58 (0.041)¤ 31.868 0.0033 0.59 (0.043)¤

Table 1 continued

Study (Publication Date & Page no.)

O'Byrne et al. (1985, p175)

O'Byrne et al. (1985, p173)

Price (1974, p40 & 59)

Tarassoff (1993, p83)

Atlanta (blocks) Atlanta (houses) Boston (rentals) M ontreal

U yeno et al. (1993, p9); Biggs (1990, p136) U yeno et al. (1993, p11)

Vancouver (houses) Vancouver (condos)

M ean N DI (sd) ­ unwt.

M edian NDI ­ unwt. Wt. mean NDI (sd)

Q statistic R andom effect variance R andom effect mean

Filed 10/05/2006 38,Page 14 of Volume Part 1

Notes: All individual housing values are actual sales or list prices, except for Baltimore where professional appraisal values are used; the census values are based on self-appraisals by owner-occupants. Asterisks indicate that the t-statistic > 2.0. The mean property values were obtained by in ating by the consumer price index (CPI-U) and, if necessary, converting Canadian dollars to U.S. using the exchange rate in 2000. All other data and information are obtained from the individual studies as listed in the table. See text for explanation of accessibility adjustments and for explanation of the calculation of the means. The critical value for the Q statistic at the 5% con dence level is 19.01.

se 1:01-cv-00201-VJW Noise Meta-Analysis of Airport Document 219-4

Filed 10/05/2006

Page Nelson

15 of

weights given by the inverse of the variance of each estima te. The xedeffects model assumes that this variance is due to sampling error only. The inverse variance weight is wi D (1=s2 ); where s2 is the estimated conditional i i variance of Di for any i: The weighted-mean and variance are given by DD
N X iD1

wi D i =

N X iD1

wi ;

V(D) D 1=

N X iD1

wi ;

(1)

where D is the cumulative effect size and V(D) is its variance. As indicated, these estima tes are obtained without imposing any structure on the data. In order to test for equivalence of effect sizes among the studies, H edges and Olkin (1985, p.123) suggest the following homogeneity test statistic: QD
N X iD1

wi (Di ¡ D)2 ;

(2)

which has a chi-square distr ibution with N ¡ 1 degrees of freedom. If Q is large, the null hypothesis is rejected and the xed-effects model is inappropriate; that is, each study may not be estima ting the same population value and a single cumulative estima te is inappropriate. A signi cant Q indicates that the variance among the effect sizes is greater than expected due to sampling error alone. If homogeneity of the estimates is rejected, several alternatives to the xed-effects model can be considered for a given data set. A metaregression analysis seeks to explain categorical within-study variabilit y, while a ``random-effects'' analysis incorporates both the within-stu dy variabilit y and sampling variation that may be due to an underlying population of effect sizes, namely, the between-study variability. In a random-effects model, the random component of the effect-size variation is calculated and incorporated into the summary statist ics. A brief discussio n of these alternative models follows. In a meta-regressio n, the studies are grouped together according to one or more differentiating characteristics or predicto rs (for example, time period, country, linear vs log speci cation, and so on). The true effect size is assumed to a function of the predictors, that is, each estima te Di is an unbiased estimate of Ii : The general form of the meta-regression model is Di D b0 C b1 Xi1 C b2 Xi2 C ¢ ¢ ¢ C bP XiP C ei ; i D 1; . . . ; N; (3)

where (Xi1 ; Xi2 ; . . . ; XiP ) is a vector of predicto r variables, which are typically dummies; (b0 ; b1 ; . . . ; bP ) are unknown parameters; and Ei is a random error term with zero mean and constant variance. The estima tes

15

se 1:01-cv-00201-VJW Document 219-4 Journal of Transport Economics and Policy

Filed 10/05/2006 38,Page 16 of Volume Part 1

of the parameters are obtained using a weighted least-squares model. Tests or comparisons can be made to determine if the effect sizes are heterogeneous after removing the variability associated with the predictor variables. If the effect sizes are homogeneous conditionally on the predictor variables, the regression constant term b0 is an estima te of the ``true'' effect size. The results of a meta-regression analysis are reported in the next section. A second alternative is a random-effects analysis, which allows for random variation in the true effect size from one study to another. Each estimate Di is a draw from the same statist ical distribution for I: The true effect sizes are not assumed to be identical, but rather to be observations of the same statistica l distrib ution. Application of random-effects analysis requires an estimate of the unknown variance due to sampling, and the total variance, v¤ ; is the sum of the random or between-stu dies variance i and the within-study variance. The inverse variance weight is given by w¤ D (1=v¤ ): H ence, random-effects analyses involve greater variance i i around the average estimate, and weaker results generally. Subjective statements about the results also are weaker, such as ``on average, the effect size is D;'' or ``in most cases, the effect size is D:'' H edges and Olkin (1985) recommend the following method-ofmoments estimate of the random-effects variance component: ni D P Q ¡ (N ¡ 1) ¡P 2 P ¢ ; wi ¡ wi = w i (4)

where Q is the value of the homogeneity test statistic in (2); N is the number of effect size estimates; and wi is the inverse variance-weight de ned above. Previous meta-analyses of airport noise have ignored random-effects models. The bottom row of Table 1 summarises the results for the xed-effects and random-effects models. The unweighted mean N D I (standard deviation) is 0.75 (0.295) and the median is 0.67. The xed-effects weighted-mean N D I is 0.58 (0.041). The Q statistic is 31.868, and the null hypothesis of homogeneity is rejected at the 5 per cent con dence level. The random-effects variance component is only 0.0033, which is added to the variance of each effect size to obtain the total variance. U sing this new set of weights, the random-effects weighed mean N D I is 0.59 (0.043). H ence, application of the random-effects model does not alter the summary statistics in a substantial manner. The ``no stru cture'' model suggests an N D I of about 0.60 per cent per dB, which is a cumulative sum of 31 estimates in Table 1. The xed-effect 95 per cent con dence interval is 0:58 § 0:080; which yields a fairly tight range of estimates from 0.50 to
16

se 1:01-cv-00201-VJW Noise Meta-Analysis of Airport Document 219-4

Filed 10/05/2006

Page Nelson

17 of

0.64 per cent per dB. H ence, further analysis of the variabilit y in Table 1 requires a meta-regression analysis.

3.2. Comparisons with previous meta-analyses The results in Table 1 can be compared to those reported in three previous studies. N elson (1980) summarised N D I estimates from thirteen studies covering 18 airports, including Australia (2 estimates), Canada (2), U K (1), and U S (13). Only unweighted means were calculated, although pooled results were presented for six U S airports. The study concluded that `` . . . the noise discount is commonly 0.5­0.6% , although a higher value may occur in some high-income areas'' (N elson, 1980, p.46). This result compares favourably with the estima tes derived in Table 1. H owever, a second study by Johnson and Button (1997) reached a different conclusion. Their study covered 18 estimates of the N D I for airports located in Australia (1 estimate), Canada (5), U K (3), and U S (9). They presented a xed-effects regression with three predictor variables, but none of the predictors were statistica lly signi cant and the observations were not weighted. 11 A random effects analysis was not included. D espite the limited nature of the analysis, the study concluded that `` . . . the results provide litt le by way of overall explanation for variabilit y in results'' (Johnson and Button, 1997, p.228). A third study provided a more ambitio us applicat ion of meta-analysis. Schipper et al. (1998) examined nineteen studies that provide 30 N D I estima tes for airports located in Australia (2 estimates), Canada (5), U K (2) and U S (21). Standard errors for the N D I estimates were not reported. The simple mean N D I was 0.83 (0.72), but this result is affected by one very large estimate for London and by several estimates from the time period of the early 1960s. The median N D I was 0.61. The Q test rejected homogeneity, but the authors did not report the weighted mean. Weighted least-squares were used to obtain the effect of four predictor variables on the variation of the N D Is, but the constant term is inexplicably negative. The most important predicto r variable was the ``relative mean sample house price,'' which is the mean sample house price divided by per capita income. This variable is interpreted as a measure of buying power.
11

The simple mean NDI in Johnson and Button (1997) is 0.68, but standard errors of the NDIs are not reported. F urther, results in several studies are misreported (e.g. Nelson, 1979; Paik, 1972) and other relevant studies are omitted (e.g. N elson, 1978). In related work, Button and N ijkamp (1997) pooled estimates of the NDI for airport noise with estimates for road traf c noise. They failed to report standard errors for the regression analysis of this sample. The subjective nature of these studies is inconsistent with the objectives of meta-an alysis, which stresses comparability, completeness, and transparency.

17

se 1:01-cv-00201-VJW Document 219-4 Journal of Transport Economics and Policy

Filed 10/05/2006 38,Page 18 of Volume Part 1

H owever, there are two reasons why the variable is misspeci ed: (1) the value of a house is itself a measure of household permanent income; and (2) the authors fail to explain how they estima te per capita income for studies that use disaggregate data for individual houses. D ividing the mean value of a house by the average level of per capita income in the entire urban area is clearly a missp eci cation. A predictor variable for loglinear and semi-log functional forms had a signi cantly negative sign and the year of publication had a signi cantly positive sign. A dummy variable for studies that use data for 1960 (such as Paik, 1972) was positive and weakly signi cant. The study concluded that `` . . . the noise depreciation is larger for locations or samples that have a larger average house price'' (Schipper et al., 1998, p.121). This conclusion must be questio ned due to the problems associated with the speci cation of the relative house price variable and the signi cantly negative intercept term. In summary, two previous meta-analyses covered a variety of countries and dates. N either study considered the effect of accessibility on the estimates of the N D Is. Statist ical results are incompletely reported. The present study restricts the sample by country and years to better ensure a homogeneous sample. The remainder of the paper seeks to improve the regression analysis by expanding the sample to include several new studies, and by reconsidering the effects of average house price, accessibility, and other in uences on the range of estimates.

4. Meta-Regression Analysis
This section presents a xed-effects regression analysis of the studies in Table 1, which covers 33 N D I estimates for airports in Canada (7 estimates) and the U S (26 estimates). All estimates are for the year 1967 and later. About one-third of the N D I estima tes have not been previously reported in the litera ture. Table 1 presents the following information set that is relevant for applicat ion of meta-regression analysis: Sample Characteristics Airport and country (area if applicable) Sample time period Sample size Census data or individual sales M ean property value (2000 U S dollars)

18

se 1:01-cv-00201-VJW Noise Meta-Analysis of Airport Document 219-4
Econometric Speci®cation and Results

Filed 10/05/2006

Page Nelson

19 of

N D I estimate (absolute value) and standard error (page no. for estima tes) Logarithmic vs linear functio nal form Coef cient of determination (R-square) Speci cation for airport accessib ility (``no'' means explicit adjustment is absent) G iven this information set, a regression analysis was conducted for a sample of 29 to 31 observations. (Two estima tes are deleted due to lack of mean property values and two estima tes are omitted in some regressions due to lack of standard error estimates.) R esults are reported for three different corrections for heteroscedasticity: (1) White's heteroscedasticconsistent standard errors; (2) weighted least-sq uares using inverse variance weights; and (3) weighted least-squares using inverse standard error weights. The resid uals are assessed using the Jacques-Bera statistic for normality. The overall t of the regression model is assessed by a standard F-test, which is equivalent to a partitio ning of Q into explained and unexplained portions (Lipsey and Wilson, 2001). I also report two additional diagnostics for the quality of the regressions, which are White's heteroscedasticity test and R amsey's R ESET test for speci cation error bias. Table 2 shows the results of six regressions, which illustrate different treatments of heteroscedasticity and model speci cation. R egressio ns (1) and (2) do not use weighted least-squares, but standard errors are computed by White's heteroscedastic-consisten t estimator. The quality of these two regressions is poor as judged by the high p-values for the F-test and the low p-values for the Jacques-Bera normality test. Only the constant term is signi cant in regression (1), while the constant and average real property values are signi cant in regression (2). The two dummy variables for aggregate census data and 1967­1970 data performed poorly in regression (2), and have been omitted from the other regressio ns. R egressions (3) and (4) are estimated by weighted least-squares using inverse variance weights. The summary and diagnostic statistics indicate high-quality regressions on all counts. The only quali cation is an outlier among the residuals, which is due to the N D I estimate by F romme (1978) for Washington, D C. R egression (3) has a signi cant constant of 0.53, and the coef cients for linear models and Canada are both signi cantly positive. There is some overlap between sample size and the access dummy variable. When the latter variable is dropped from the model, the constant term declines slightly to 0.51 in regressio n (4). The coef cient magnitudes

19

se 1:01-cv-00201-VJW Document 219-4 Journal of Transport Economics and Policy

Filed 10/05/2006 38,Page 20 of Volume Part 1

Table 2 Meta-Regression Results for the Noise Depreciation Index (NDI)
Variable (1) (2) (3) (4) (5) (6)

0.8316 0.7020 0.5332 0.5069 0.6466 0.6651 Constant (0.3062)¤ (0.2969)¤ (0.1893)¤ (0.1425)¤ (0.2254)¤ (0.2043)¤ M ean real property value 0.0006 0.0008 ¡0.0001 ¡0.0001 ¡0.0002 ¡0.0002 (£0.001) (0.0004) (0.0004)¤ (0.0013) (0.0013) (0.0012) (0.0011) Accessib ility dummy ¡0.0106 0.0100 0.0196 ¡0.0208 (no adj D 1) (0.1207) (0.1363) (0.0900) -- (0.0959) -- Sample size ¡0.0504 ¡0.0475 ¡0.0186 ¡0.0140 ¡0.0231 ¡0.0272 (log) (0.0469) (0.0447) (0.0342) (0.0261) (0.0336) (0.0274) Linear model dummy 0.1862 0.2021 0.3320 0.3340 0.3035 0.3004 (linear D 1) (0.1180) (0.1175) (0.1579)¤ (0.1544)¤ (0.1225)¤ (0.1192)¤ Country dummy 0.2236 0.2708 0.3389 0.3357 0.2797 0.2807 (Canada D 1) (0.1354) (0.1600) (0.0834)¤ (0.0805)¤ (0.0939)¤ (0.0919)¤ Census data dummy 0.0502 (census D 1) -- (0.1406) -- -- -- -- Year dummy 0.0553 (1967­70 D 1) -- (0.1653) -- -- -- -- R-sq 0.309 0.322 0.774 0.773 0.234 0.233 F-test (p value) 0.083 0.196 0.002 0.001 0.023 0.010 RESET test (p) 0.322 0.470 0.129 0.202 0.088 0.062 White test (p) 0.171 0.123 0.500 0.420 0.509 0.397 J­B test (p) 0.041 0.042 0.141 0.146 0.530 0.502 Weights used White se White se inv. var. inv. var. inv. se inv. se Sample size 31 31 29 29 29 29 M ean dep. var. 0.739 0.739 0.568 0.568 0.632 0.632 (sd) (0.303) (0.303) (0.370) (0.370) (0.223) (0.223) Notes: D ependent variable is the absolute value of the noise depreciation index (NDI) from Table 1. Asterisks indicate that the t-statistic > 2.0. Sample size D 31 omits two studies without a mean property value estimate and samp le size D 29 omits additionally two studies without a standard error estimate. F-statistic for the test of joint signi cance of all regressors; the null hypothesis is jointly insigni cant. RESET test for speci cat ion error bias; the null hypothesis is no bias. White's test statistic for heteroscedastic residuals; the null hypothesis is no heterosced asticity. Jacques Bera (J­B) test statistic for normally distributed residuals; the null hypothesis is normality.

for linear models and Canada are not affected. R egressions (5) and (6) show the results using inverse standard error weights. Although a strong case can be made for this weighting scheme (Saxonhouse, 1976), the overall results are somewhat poorer as judged by the F-test and R ESET test. The constants in regressions (5) and (6) are 0.65 and 0.67, respectively. The main result in Table 2 is a signi cant and positive constant term that lies between 0.51 and 0.67. U sing regression (4), the 95 per cent

20

se 1:01-cv-00201-VJW Noise Meta-Analysis of Airport Document 219-4

Filed 10/05/2006

Page Nelson

21 of

con dence interval for the constant term is 0.28 to 0.79, which includes 22 of the 33 estimates in Table 1. Three of the estimates that lie outsid e this range are based on a linear functional form, including Los Angeles and N ew York City, and four of the outlier s involve Canadian airports. The N D Is from linear models required an estima te of the effect size based on the mean property value. The N D I estima tes for Canada may re ect special or unique features of Canadian real estate markets, climate, or operating conditio ns (curfews, frequency). The other outlier s are two estima tes for Washington, D C; one estimate for R ochester; and one estima te for D allas. H ence, there are the only four estimates for three airports that are not explained by the meta-regression model. U sing the constant terms in regressions (4) and (6), the effect of airport noise on U S property values is 0.51 and 0.67 per cent per dB, respectively, which compare favourably to the weighted-mean of 0.58 per cent per dB in Table 1.

5. Conclusions
The results in the present study are consisten t with an earlier contribution by the author (N elson, 1980), which concluded that the noise discount was about 0.50 to 0.60 per cent per dB. The present study expands the sample of estimates from 18 to 33, including a doubling of the number of estima tes for U S airports. Although a number of estima tes in Table 1 employ data for the 1970s, there does not seem to a measurable effect of time on the N D Is. H ence, a given property located at 55 dB would sell for about 10 to 12 per cent less if it was located at 75 dB, all other things held constant. Stated differently, under these same circumstances, a $200,000 house would sell for $20,000 to $24,000 less, which yields a hedonic price of $1000 to $1200 per dB. The noise discount in Canada appears to be greater, 0.80 to 0.90 per cent per dB, and may re ect differences in legal rules as well as other economic differences. It remains to be seen whether the results in this paper are robust in the face of other analytical methods, such as G IS studies, contingent valuation methods, and new hedonic studies that consider spatial autocorrelation of housing prices (Salvi, 2003). F urther, empirical estimation of structural demand models has not been applied to noise avoidance, although additional progress has been made on the theoretical front (Sheppard, 1999). Lastly, caution should be exercised in applying the estimates in this paper to residential areas near airports that are affected by noise in excess

21

se 1:01-cv-00201-VJW Document 219-4 Journal of Transport Economics and Policy

Filed 10/05/2006 38,Page 22 of Volume Part 1

of 75 dB. Survey studies by F eitelso n et al. (1996) and F rankel (1991) suggests that the noise discount per dB could be substantially higher where the level of noise exposure makes land virtually unsuitable for any residential use.

Appendix A Hedonic Studies Included in Table 1
Biggs, A. J. G . (1990): The Impact of Airport Noise: A Case Study of Vancouver International Airport, U npublished M .S. thesis, U niversity of British Columbia. Booz-Allen and H amilton, Inc. (1994): The Effects of Airport Noise on Housing Values: A Summary Report, PB95­212627, BAH and F ederal Aviation Administration, Washington, D C. Blaylock, J. E. (1977): Airport Noise and Housing Values: An Investigation into the Hedonic Theory of Housing and the Value of Quiet, U npublished Ph.D . dissertation, Texas A&M U niversity. D eVany, A. S. (1976): ``An Economic M odel of Airport N oise Pollution in an Urban Environment,'' in Theory and Measurement of Economic Externalities, S.A.Y. Lin, ed., Academic Press: N ew York, 205­14. D ygert, P. K . (1973): Estimation of the Cost of Aircraft Noise to Residential Activities, U npublished Ph.D . dissertation, U niversity of M ichigan. Emerson, F . C. (1969): The Determinates of Residential Value with Special Reference to the Effects of Aircraft Nuisance and Other Environmental Features, U npublished Ph.D . dissertation, University of M innesota. Emerson, F . C. (1972): ``Valuation of R esidential Amenities: An Econometric Approach,'' Appraisal Journal, 40, 268­78. Espey, M . and H . Lopez (2000): ``The Impact of Airport N oise and Proximity on R esidential Property Values,'' Growth and Change, 31, 408­19. F romme, W. R . (1978): Conceptual Framework for Trade-Off Analysis of Multiple Airport Operations: Case Study of the Metropolitan Washington Airports, U npublished Ph.D . dissertation, University of M aryland. K aufman, H. F . (1996): No Plane, Big Gain: Airport Noise and Residential Property Values in the Reno-Sparks Area, U npublished M .S. thesis, U niversity of N evada, Reno. Levesque, T. J. (1994): ``M odelling the Effects of Airport Noise on Residential H ousing M arkets: A Case Study of Winnipeg International Airport,'' Journal of Transport Economics and Policy, 28, 199­210. M ark, J. H . (1980): ``A Preference Approach to M easuring the Impact of Environmental Externalities,'' Land Economics, 56, 103­16. M aser, S. M., W. H. R iker, and R. N . R osett (1977): ``The Effects of Zoning and Externalities on the Price of Land: An Empirical Analysis of M onroe County, N ew York,'' Journal of Law and Economics, 20, 111­32. M cMillan, M . L. (1979): ``Estimates of H ouseholds' Preferences for Environmental Quality and Other H ousing Characteristics from a System of D emand Equations,'' Scandinavian Journal of Economics, 81, 174­87.

22

se 1:01-cv-00201-VJW Noise Meta-Analysis of Airport Document 219-4

Filed 10/05/2006

Page Nelson

23 of

M cMillan, M. L., et al. (1980): ``An Extension of the H edonic Approach for Estimating the Value of Quiet,'' Land Economics, 56, 315­28. M ieszkowski, P. and A. M . Saper (1978): ``An Estimate of the Effects of Airport N oise on Property Values,'' Journal of Urban Economics, 3, 425­40. M yles, C. L. (1997): Sound Effects: Measuring the Impact of Aircraft Noise on Residential Property Values Around McCarran International Airport, U npublished M .A. thesis, University of N evada, Las Vegas. National Academy of Sciences (1977): Noise Abatement: Policy Alternatives for Transportation, N AS, Washington, D .C. Nelson, J. P. (1978): Economic Analysis of Transportation Noise Abatement, Ballinger, Cambridge, M A. Nelson, J. P. (1979): ``Airport N oise, Location R ent, and the M arket for R esidential Amenities,'' Journal of Environmental Economics and Management, 6, 320­31. Nelson, J. P. (1980): ``Airports and Property Values: A Survey of R ecent Evidence,'' Journal of Transport Economics and Policy, 14, 37­52. Nelson, J. P. (1981): ``M easuring Bene ts of Environmental Improvements: Aircraft Noise and H edonic Prices,'' In Advances in Applied Microeconomics, V.K . Smith, ed., JAI Press, G reenwich, CN , 51­75. O'Byrne, P. H., J. P. Nelson, and J. J. Seneca (1985): ``H ousing Values, Census Estimates, Disequilibrium and the Environmental Cost of Airport N oise: A Case Study of Atlanta,'' Journal of Environmental Economics and Management, 12, 169­78. Price, I. (1974): The Social Cost of Airport Noise as Measured by Rental Changes: The Case of Logan Airport, Unpublished Ph.D. dissertation, Boston U niversity. Quinlan, D . A. (1970): Property Devaluation due to Jet Aircraft Densities, U npublished M .S. thesis, U niversity of R ochester. Tarassoff, P. S. (1993): A Hedonic Model of the Impact of Localized Aircraft Noise on Housing Values, Unpublished M .S. thesis, M cG ill University. Uyeno, D., S. W. Hamilton, and A. J. G . Biggs (1993): ``D ensity of Land U se and the Impact of Airport Noise,'' Journal of Transport Economics and Policy, 27, 3­18.

Appendix B Hedonic Studies Excluded from the Meta-Analysis
Al-K habbaz, A. A. (1987): Modeling Aviation Facilities Impact on Residential Property Values, Unpublished Ph.D . dissertation, U niversity of Arizona. { Only distance measures used to obtain noise impacts.} Clark, D . E. and W. E. Herrin (2000): ``The Impact of Public School Attributes on H ome Sale Prices in California,'' Growth and Change, 31, 385­407. { Only distance measures used to obtain noise impacts.} Cockerill, L. W. (2000): Airport Proximity and Single-Family Home Price in Southern California: A Hedonic Housing Value Approach, U npublished M .A. thesis, California State U niversity, F ullerton. { Only distance measures used to obtain noise impacts.} Crowley, R . W. (1973): ``A Case Study of the Effects of an Airport on Land Values,'' Journal of Transport Economics and Policy, 6, 144­52. { Could not compute an N DI.}

23

se 1:01-cv-00201-VJW Document 219-4 Journal of Transport Economics and Policy

Filed 10/05/2006 38,Page 24 of Volume Part 1

D e Araujo, L. A. (1986): The Impact of Airport Noise on Residential Values, U npublished M .S. thesis, Washington U niversity (St. Louis). { Could not compute an N D I.} D WG R esearch Associates (1990): The Management of Airport Noise, Transportation D evelopment Center, M ontreal. { Could not compute an N DI.} K nickerbocker, N . N . (1991): Aircraft Noise and Property Values, U npublished Ph.D . dissertation, U niversity of M aryland. { repeat sales analysis; could not compute an A N DI.} K oda, L. S. (2003): ``A Comparison of M ethodologies to M easure Effects of Airport Siting D ecisions,'' U npublished paper, Texas A&M U niversity. { Case study of an airport closure; could not compute an N D I.} Lane, T. (1998): ``The Impact of Airport Operations on Land Values: A Case Study of Seattle Tacoma International Airport,'' Paper presented at the 32nd Annual Paci c N orthwest R egional Economic Conference, Olympia, WA. { Only distance measures used to obtain noise impacts.} Li, M . M . and H. J. Brown (1980): ``M icro-Neighborhood Externalities and H edonic H ousing Prices,'' Land Economics, 56, 125­41. { oise estimates are apparently due to N traf c and general noise.} M cClure, P. T. (1969): Some Projected Effects of Jet Noise on Residential Property Near Los Angeles International Airport by 1970, P-4083, R AND Corp., Santa M onica, CA. { Could not compute an N D I.} M cD ougall, G. S. (1976): ``An Inquiry into the Demand for Aircraft Noise Abatement,'' Review of Regional Studies, 6, 59­69. { Could not compute an N DI.} Paik, I. K . (1972): Measurement of Environmental Externality in Particular Reference to Noise, U npublished Ph.D . dissertation, G eorgetown University. { ses census data for U 1960.} West, R . J. (1988): ``Statistical Inference: An Avigation Easement Analysis,'' Real Estate Issues, 13, 35­39. { Could not compute an N D I.}

References
Bartik, T. J. and V. K. Smith (1987): ``U rban Amenities and Public Policy,'' In Handbook of Regional and Urban Economics, E.S. M ills, ed., Elsevier: Amsterdam, 1207­54. Bateman, I. J., et al. (2002): ``Applying G eographical Information Systems (GIS) to Environmental and R esource Economics,'' Environmental and Resource Economics, 22, 219­69. Bishop, D. E. (1966): ``Judgements of the R elative and Absolute Acceptability of Aircraft N oise,'' Journal of the Acoustical Society of America, 40, 108­22. Boyle, M . A. and K . A. Kiel (2001): ``A Survey of H ouse Price H edonic Studies of the Impact of Environmental Externalities,'' Journal of Real Estate Literature, 9, 117­44. Brookshire, D. S. and H . R. N eill (1992): ``Bene t Transfers: Conceptual and Empirical Issues,'' Water Resources Research, 28, 651­55. Button, K . and P. Nijkamp (1997): ``Environmental Policy Assessment and the U sefulness of M eta-Analysis,'' Socio-Economic Planning Sciences, 31, 231­40. D alhuisen, J. M ., et al. (2003): ``Price and Income Elasticities of Residential Water D emand: A M eta-Analysis,'' Land Economics, 79, 292­308.

24

se 1:01-cv-00201-VJW Noise Meta-Analysis of Airport Document 219-4

Filed 10/05/2006

Page Nelson

25 of

De Araujo, L. A. (1986): The Impact of Airport Noise on Residential Values, U npublished M .S. thesis, Washington U niversity (St. Louis). DeVany, A. S. (1976): ``An Economic M odel of Airport N oise Pollution in an Urban Environment,'' In Theory and Measurement of Economic Externalities, S.A.Y. Lin, ed., Academic Press: N ew York, 205­14. Desvousges, W. H., F . R . Johnson, and H . S. Banzhaf (1998): Environmental Policy Analysis with Limited Information: Principles and Applications of the Transfer Method, Elgar: Cheltenham, U K. Espey, M ., et al. (1997): ``Price Elasticity of R esidential Demand for Water: A M etaAnalysis,'' Water Resources Research, 33, 1369­74. F eitelson, E. I, R . E. H urd, and R . R . M udge (1996): ``The Impact of Airport N oise on Willingness to Pay for Residences,'' Transportation Research, 1D , 1­14. F inegold, L. S, C. S. H arris, and H . E. von G ierke (1994): ``Community Annoyance and Sleep D isturbance: U pdated Criteria for Assessing the Impacts of G eneral Transportation N oise on People,'' Noise Control Engineering Journal, 42, 25­30. F ollain, J. R. and E. Jimenez (1985): ``Estimating the D emand for H ousing Characteristics: A Survey and Critique,'' Regional Science and Urban Economics, 15, 77­107. F rankel, M . (1991): ``Aircraft N