ML12088A308
| ML12088A308 | |
| Person / Time | |
|---|---|
| Site: | Indian Point  |
| Issue date: | 03/28/2012 |
| From: | Boardman A, Greenberg D, Vining A, Weimer D Prentice-Hall, Simon Fraser Univ, Univ of British Columbia, Vancouver, Canada, Univ of Maryland - Baltimore, Univ of Wisconsin - Madison |
| To: | Atomic Safety and Licensing Board Panel |
| SECY RAS | |
| References | |
| RAS 22098, 50-247-LR, 50-286-LR, ASLBP 07-858-03-LR-BD01 | |
| Download: ML12088A308 (7) | |
Text
ENT000154 Submitted: March 28, 2012 THIRD EDITION COST-BENEFIT ANALYSIS Concepts and Practice Anthony E. Boardman University of British Columbia David H. Greenberg University of Maryland Baltimore County Aidan R.Vining Simon Fraser University David L. Weimer University of Wisconsin-Madison PEARSON Prentice Hall Upper Saddle River, New Jersey 07458
r ,llInll'), oj' (:Hngrcss Cataloging-ill-Publicati on Data Cost-benefit an alysis: concepts (lnd pr:1 cli,,;c / AnthollY E. Boardman ... fe t aJ.J. - 3rd cd.
- p. em.
Incllides bibliographical n: fc rc IICC!> ;1 1111 ilHlcx.
ISBN 0 1435?D-J (t';lsehlHlnd : ,d l..:. pnp!.!r)
- 1. Cos t d'fc c(ivl.'Jl(,.~s. l. lIomLi Ill<l 11 , Anthony E, HD47,lI.CM() :Wl)(l 6,'iK IS'5:1 Ik :~.~
2005023256 t\(' qlli~ilh IU ~ 1':( li[(I I': .1011 Axelrod 1\ VI'I I ~ .\l.' l.' Il[ iv!.' I :di[or: David Alexander VI'II ;dil'll'i"i I )ircc(or: Jeff Shelstad 1'14)( 111\ '1 1k"dopmcnt Manager: Pamela Hersperger I'lI ljl 'cl Manager: Franccsc~, Cn logcro hlill H'ial Assistant Michael D illamo AV I'/I~xCCl!livc Marke ti ng Manager: Sharon Koch 1'I'!;lrket ing Assis tantTin a Panagiotoll
- )"'l1ior MHn aging Editor (Production)
- Cynthia Regan Production Editor: D enise Culhane Perm issions Coordinator: Cha rles Morr is P roduct io n M,lIlager: A rno ld Vila Manufacturing Buyer: Michelle Klein Cover Design: Bruce Kensclllar Cover ft!u stration/Photo: Brian St(lblyk/Photographers Ch oice/Gctty Images, Inc.
Manager, Multimedia Production: Christy Mahon Composition: Integra Soft ware Services Full-Se rvice Project M<lnagcme nt: BookMasters, Inc.
PrintcrlD indcr: H amilton Printi ng Typeface: 10112 Times Ten Copyright © 2006, 2001 by PeHrsoli EducatioJl, Inc., Upper Saddlc Rivcl',
New Jcrscy 07458.
Pearson Prcntit.'c Hall. All righ ts reserved. Prin ted in the United Sta tes o f America.
Th is publica tion is pro tected by Copyrigh t and permission shou ld be o btained from the publisher prior to any pro hibited reproduction, storage in a retrieval system , or transmission in any form or by a ny means, electronic, mechani cal, photocopying, recording, o r likewise. Pm' info rmation regarding perm ission(s) , write to: Rights and Pe rmissions D epartment.
l'earsoll ))rc lltice HaUl'll is a trademark of Pearson Education, Inc.
Pearson is a registered tradema rk of Pearson p Ic Prentice H111I is a registe red trademark of PCRrson Education , J nco P earsol1 Educ(ltion LTD. Pearso n Educat ion Austf(l iia PTY, Lim ited P earson Education Singapore, PIe. LuJ Pearson Education NOl'l h Asia Ltd Pea rsall Educ.atio n, Canada , Ltd Pea rson Educaci6 n de Mexico, S.A. de C.V.
Pearso n Education**-Japa n Pearson Education Ml.Ilflysia, Pte. Ltd PEARSON 10 98765 ISBN 0-13-143583 -3
348 PART III Vaillation o(lmpacts announcemen t of a new program or policy. T he main advantage of using stock prices is tha t new information concerning policy changes is qu ickly and efficientl y capitaliz,;"
into stock prices. Changes in stock prices provide an unbiased estimate of the va lue of a policy change to shareholders. A lso, stock price data are readily accessible ill computer-readable fo rm.
In an event st udy, researchers estimate the abnormal return lO a security, which i ~
the differcnce between the retu rn to a security in t he presence of an event and 11\,'
return to the security in the absence of the event. Usually, researchers estimate daily abnormal returns during an event window, that is, for th e period during which the eveli!
is assllmed to affect stock prices - often a few days. Because the return to the securily in the absence of the event is unobservable, it is inferred from changes in the price, of other stocks in the market, such as the D ow Jon es Index or the FTSE lOO.J1 '1'111; estimated daily abnormal returns during the event window can be aggregated to obtaiII the cumulative abnormal return, which measures the total return to shareholders thai can be attributcd to the evcnt. Cumulative abnormal rcturns provide an estimate 01' Ih o change in producer surplus due to some ncw policy.
The va luation methods discussed earli er in this chapter have several potential limita -
tions, many of which were discussed earlier. This section focuses on the mnitted v(/ri ,
able problem and self-selection bias.
The Omitted Variable Problem All of the methods discussed thus far in this chapter implicitly aSSlllllC that all other explanatory va riables arc held constant, but this* is unlike ly in practice. Consider, for example, using the intermediate good method to val ue irrigation. Ideall y, ana lysts wou ld compare the incomes of farmers if the irrigation project were buil t with the incomes of the same farmers if the project were not built. In practice, if the project is built, analysl, cannot dirccliy obse rve what th e farmers' incomes would have been if it had not been built. One way to infer what their in comes wou ld have been wit hout the project is to liS!.'
the incomes of the same fanners before the project was built (a before and after design) or the incomes of similar fanners who did not benefit from an irrigat ion project (a nCHl '
experimental comparison gro up design).lllC before and after design is reasonable only if all other variables that affect farmers' incomes remain constant, such 8S weather cond itions, crop choices, taxes, and subsidies. If these variables change then th e incomc~i observed before the project arc not good estimates of what incomes would have beclI if the project had not been imple mented. Si milarly, the com parison group design i, appropriate only if the comparison group is similar in all important respects to the farmers with irrigation, except fo r the presence of irriga tion.
As mentioned ill Exhibit 13-2, salary differences between those with a college degree and those with a high school degree may depend on ability, intelligence, socio..
economic background and other factors in addition to college attendance. Similarly, in labor market studies of the value of life, differences in wages among jobs may de pend on variations in status among jobs and the barga in ing power of different unions ill
CHAPT ER 13 Valuing Impacts fram Observed Behavior: Indirect Market Methods 349 Ising stock prk(1)j l~ additio n to fatality risk, In simple asset price studies, thc price of a house typically
'iciently capilali "! depend s on factors such as its distance from the centra l busines s district and sil.e, as late of th e vahh: of well as whethe r it has a view, Analysts should take accoun t of all import ant explana-Idily accessi!>k iii tory variables, If a relevant explan atory variable is omitted from the model and if it is correla ted with the included variable(s) of interest, then the estimat ed coefficients will a security, whid\ iij be biased, as discussed in Chapte r 12.
f an event <tlld liut hers estimat e dail V Self-S electio n Bias ing which the Cl't\iil Anothe r potenti al problem is self-se lection bias. Risk -seekin g people turn to the seculit y tend to self-select themselves for danger ous jobs. Because they like to take risks anges in the priuj~ they may be will-ing to accept low salaries in quite risky jobs. Consequently, we may observ e FTSE 100. 11 '1M e only a very small wage premium for danger ous jobs. Because risk seekers are not ggregated to obl""1 represe ntative of society as a whole, the observed wage differential may underestimate
) shareh oldcrs Ihilt the amoun t that average membe rs of society would be willing to pay to reduce risks e an estimate of tlui and, hence, may lead to underestimation of the value of a statistical life.
The self-selection problem arises whenever diffe rent people attach diffe rent val-ues to particu lar attributes, As anothe r example, suppos e we want to use differences in house prices to estimat e a shadow price for noise. People who are not adverse to noise, possibly becaus e of hearing disabilities, naturally tend to move into noisy neighb or-hoods. As a result, the price differential between quiet houses and noisy houses may be
'al potenti allimil a quite small, which would lead to an undere stimation of the shadow price of noise for
!D the omitted \lad the "average)) person.
HEDO NIC PRIC ING METH OD
- ume tbat all olhcl ctice. Consider, for The hedoni c pricing method , sometimes called the hedoni c regressioN melhod, offers a
,ally, analysts would way to overco me the omitted variabl es problem and self-selection bias that arise in the W'ith the incomes nl relative ly simple valuati on method s discussed earlier . Most rece nt wage-risk ect is built, analy~!~; studies for valuing a statistic al life (alSO called labor market studies) apply the hcdoni c n if it had not bel_'ll regression method .
the project is to us\~
re and after design ) Hedon ic Regression tion project (a nOll Suppose, for example, that scenic views can be scaled from 1. to 10 and that we want to 1 is reasonable on ly estimat e the benefits of improving the (quality) "level" of scenic view in an area by one 1t, such as weaUll..'r unit. We could estimate the relation ship between individual house prices and the level
~e then the incom ~s of their scenic views. But we know that the market value of houses depend s on other
" would have bcm factors, such as the si7.e of the lot, which is probab ly correla ted with the quality of on group design is scenic view. We also suspec t that people who live in houses with good scenic views tend ant respects to tIll; to value scenic views more than other people, Consequently, we would have an omitted variabl es problem and self-selection bias.
lose with a coilege. 111e hedoni c pricing method attemp ts to overco me both of these types of prob-
. intelligence, socio e lems 12 It consists of two steps. The first estima tes the effect of a margin ally better Idance. Similarly, in scenic view on the value (price) of houses, a slope parame ter in a regress ion model, Ig jobs may depend while controlling for other variables that affect house prices. The second step estimates differe nt unions in the willingness-to-pay for scenic views, after controlling for "tastes,"
which arc proxied
350 PART III Valuation o(lmpacts by income and other socioeconomic fact ors. From this information, we can calculate the change in consumer surplus resulting from projects that improve or worsen the views from some houses.
The hedonic pricing method can be used to value an attribute, or a change in an attribute, whcnever its value is capitalized into thc price of an asset, such as houses or salaries. The first step estimates the relationship between the price of an asset and all of the attributes (charactcristics) that affect its valueD The price of a house, P, for exam-ple, depends on such attributes as the quality of its scenic view, ViEW, its distance from the central business district, CBD, its lot size, SIZE, and various characteristics of its neighborhood, NBHD, such as school quality. A model of the factors affecting house prices can be written as follows:
P = f(CBD, SIZE, VIEW, NBHD) (13.2)
This equation is called a hedonic price function or implicit price function. t4 The change in the price of a house that results from a unit change in a particular attribute (i.e., the slope) is called the hedonic price, implicit price, or rent differential of the attrib ute. In a well-fu nctioning market, the hedonic price can naturally be interpreted as the addi-tional cost of purchasing a house that is marginally better in terms of a particular attribute. For example, the hedonic price of scenic views, which we denote as r" mea-sures the additional cost of buying a house with a slightly better (higher-level) scenic view. IS Sometimes hedonic prices are referred to as marginal hedonic prices or marginal implicit prices. Although these terms are techn ically more correct, we will not use them in order to make the explanation as easy to follow as possible.
Usually analysts assume the hedonic price function has a multiplicative functional form, which implies that house prices increase as the level of scenic view inereases but at a decreasing rate. Assumin g the hedonic pricing model represented in equation (13.2) has a multiplicative functional form, we can write:
(13.3)
The parameters,{:J P{:J2' {:J3' and {:J4' arc elasticities:TIlCY measure the proportional change in house priccs that resu lts from a proportional change in the associated allrib ute. 16 We expect {:Jl < 0 because house prices decline with distance to the CBD, but {:J2' {:J3' and
{:J4> 0 because house prices increase as SIZE, VIEW, and NBHD increase.
The hedonic price of a particular attribute is the slope of equation (13.2) with respect to that attribute. In general, the hedonic price of an att ribute may be a function of all of the variables in the hedonic price equation 17 For the multiplicative model in equation (13.3), the hedonic price of scenic views, r v ' is: 18 p
r, = f33 ViEW> 0 (13.4)
In this model, the hedonic price of scenic views depends on the value of the parameter
/33' the price of the house, and the view from the house. Thus, it varies from one obser-vation (house) to another. Note that plotting this hedonic price against the level of
CHAPTER 13 Valuing Impacts from Observed Behavior: Indirect Market Methods 35 I can caiculnlc scenic view provides a downward-sloping curve, which implies that the implicit price of
)f worsen l1li' scenic views declines as the level of the view increases.
The prcceding points are illustrated in Figure 13-3. The top panel shows an illus-1 change in HII trative hedonic price function with housc priccs increasing as the level of scenic view h as houst'*; !i( increases, but at a decreasing rate. The slopc of this curve, which equals thc hedonic isset and ;111 IIf price of scenic views, decreases as the level of the scenic view increases. The bottom
~, P, for exam panel shows marc prccisely the relationship between the hedonic price of scenic views distance rr()jil (the slope of the curvc in the top panel) and thc level of sccnic view.
- teristics Ill' jh In a well-functioning market, utility-maximizing households will purchase ffecting 1\(11'" houses so that their willingness-to-pay for a marginal increasc in a particular House
!.14 The cll,H1W:
price (PI I
~
ribute (i.e., 1lIi'
~ attribute. III I!
,d as the addl of a parliculH! Hedonic price
!ote as r v' InUi! function er-level) SCCIlIl
~onic prh:I's ti!
'cct, we wi!1lld!
Hive functiOlwi w increasc~ hlU cd in equ<llhHi v, V3 Level of scenic view (V)
Hedonic price of tortional Cil;lII!:\i" scenic 1attribute,lll Locus of household views (rv)
W1 equilibrium but {32' P. I. "lid /illingn:ses-to-pay 1se.
tion (13.2) wil h (V1 ----------
ay be a fUllc!i( lll I I
icative model In I I
I (v2 -----.--------}---------
I I I I
%-----------~---------~-- rv I I I I I I I I
)f the paranwltH I I from one nbs!;! Level of scenic linst the level ill view (V)
352 PART III Va/u"'ioll o(lmpact' all ribule equals its hedonic price. Consequently, in equilibrium, the hedonic pricc 01 an att ribute can be interpreted as the willingness of housebolds to pay for a mar -
ginal increase in tbat allribute. The graph of the hedonic price of scenic views, f, .,
against the level of scenic view is shown in the lower panel of Figure 13-3. Assuming Dean all households have identical incomes and tastes, this curve can be interpreted as a Biggs 1 household inverse demand curve for scenic views. mate t Yet, households differ in their incomes and taste. Some are willing to pay a consid* Canad.
erable amount of money for a scenic view; others are not. This brings us to the second price e step of the hedonic pricing method. To account for different incomes and tastes, ana -
lysts should estimate the following willingness-to-pay function (inverse demand curve) for scenic views;1 9 lni r" = W(V1EW, 1'; Z) (13.5) where I erty val where f, is estimated from equation (13.4), Y is household income, and Z is a vector 01" ent noi household characteristics that reflects tastes (e.g., socioeconomic background, race, "some" age, and family size). Three willingness-to-pay functions, denoted W 1, W2 , and W3, for 25-40 three different types of households are drawn in the lower panel of Figure 13_3. 20 occurs ~
Equilibria occur where these functions intersect the r" function. Thus, when incomes actcristi and socioeconomic characteristics differ, the r, function is the locus of household equi- Thei librium willingnesses-to-pay for scenic views, Intenu Using the methods described in Chapter 4, it is straightforward to use equation SOllrc;e: P (13.5) to calculate the change in consumer surplus to a household due to a change in and the I:
the level of scenic view. These changes in individual household consumer surplus call be aggregated across all households to obtain the total change in consumer surplus.
Using Hedonic Models to Determine the VSL 111e simple forms of consumer purchase and labor market studies to value life that we described previously may result in biased estimates due to omitted variables or self-selection problems. For example, labor market studies to value life that examine fatality risk (the risk of death) often omit potentially relevant variablcs such as injury risk (the risk of nonfatal injury). This problem may be reduced by using the hedonic pricing method. For example, a researcher might estimate the foHowing nonlinear regression model to find the hedonic price of fatality risk: 21 In(wage rate) = f3 0 + f31In(fat ality risk) + f32In(injury risk) + f33In(job tenure)
+ (3)n(cdueation) + f3sln(age) + (13,6) 111e inclusion of injury risk,job tenure, education, and age in the regression model controls for variables that affect wages and would bias the estimated coefficient of f3 1 if they were excluded. Using the procedure demonstrated in the prcceding seclion, the analyst can convert the estimate of f3 1 to a hedonic price of fatality risk and can then estimate individuals' willingness- to-pay to avoid fatal risks. Most of the empirical esti-mates of the value of life that are reported in Chapter 15 are obtained from labor market and consumer product studies tha t employ models similar to the one presented in equation (13.6).