2 1. Intuition and StoriesRevealed preference studies using housing (or land markets) and labor marketsSimple hedonics story w/ 2 housesAssume houses are identical in every respect except that B is located in a polluted areaIf they sell for the same price, everyone buys AAir pollutionAB
3 Say B is offered at some discount (to entice buyers) Now, individuals implicitly face the decision: Is the clean air worth $5,000?If yes, buy AIf no, buy BIndividuals reveal something about the value (wtp) of clean air in the market decision.
4 How would markets clear in a simple case like this? Imagine many A’s and B’sImagine all houses start at $100,000Excess demand for A Prices riseExcess supply for B Prices fallGap opens up and continues to widen until individuals are sufficiently compensated by price drop to live in polluted area. Settle into an equilibrium in the housing marketImplicit value of clean airis $30,000/house. We call thisa “hedonic” or “implicit” price.AB$105,000$75,000
5 Commentsopen mkt and governing the priceidentical people and perfect compensationrent vs. asset valueValue for environmental quality “revealed” in housing or property values and based on individual preferences (and supply). More important air quality is the high the hedonic price will be.
6 Each has an implicit value Real housing (land) markets are not so simple. Housing choice and price depends on many characteristics: structural, neighborhood, and locational.Each has an implicit valueHedonic price technique is a method for sorting out these valuesThink about the $30,000 being due to more than just pollution.Works for any differentiated product: cars, food, comic books
7 Environmental Goods in Land Markets Air pollution (37 studies)Landfill/Hazardous waste sitesProximity to open spaceCrop damage/AgriculturalAirport noise“Capitalized” valuesCrimeFrat House
8 2. The Mechanics of the Hedonic Price Method (Property Values) Postulate a functional relationship between housing prices and characteristics of houses:Hedonic price functionVector of Environmental Attributes:Air qualityDistance to waste siteNoise levels and so onVector of Structural Attributes:# RoomsLot sizeAgeBasement (yes/no)Design and so onVector of Neighborhood Attributes:School qualityOpen spaceDistance from downtownDistance form shopping and so on
9 You may set this up in a linear or nonlinear way You may set this up in a linear or nonlinear way. Here is a typical linear form:Focus on the parameter .It tells us the relationship between Price and Air Quality.Take the first derivative of f(.) with respect to AirQ and you getCoefficients tell us the relationship between eachcharacteristic and housing prices. Note expected signsabove the coefficients.Implicit Price of a Unitof Air Quality in theHousing Market
10 So, we set out to estimate the coefficients in the Hedonic Price Function. To do this we need data on housing sales: price and characteristics of houses sold in the market. Data set looks something like thisHouse Price Age Lot Size Air Quality and so on ………………… .(Dollars) (Years) (Acres) (Visibility Days)1 $85,2 65,3 211,4 92,5 300,.,More characteristicsNumber of observations
11 Focus on AirQ for an estimate of the implicit value of air quality. Regress House Price on characteristics and this gives estimates of the parameters inFocus on AirQ for an estimate of the implicit value of air quality.TheoryThis is the implicit price function for characteristic#1, holding all other characteristics fixed. Note: need not be flat!
12 3. Some Theory House Mkt Equilibrium gives P(z). Individuals take P(z) as fixed.Individuals choose z and x (composite good) to maximize utilityNote that the individual ischoosing attributes
13 Shopping chart analogy, “implicit” decision Graphically This is the implicit price function for characteristic#1, holding all other characteristics fixed. Note: need not be flat!There is a similar graph for each characteristic.It is as though the person is choosing n separate goods.
14 What if the population is homogeneous? Different people w/ different income and taste will consume different levels of each characteristicYou end-up with different people at different points on the implicit price line.What if the population is homogeneous?Person #2Person #1
15 5. Issues Functional form Probably not linear, so often estimated as a nonlinear functionOne nonlinear case
16 Measurement of quality variables Flexible forms (Box-Cox)Measurement of quality variablesAir pollutionTSP,SO2Visibility indexesClear daysLandfillDistanceViewsCrime LevelsNoiseColinearityDisamenities and amenities “run” together
17 Discreteness, lumpy bundles PerceptionHealth vs. VisibilityMarket segmentation2 or more housing mkts
19 The identification problem We estimate the blue line (the implicit price function) when we estimate the hendonic price function.But, we can not in a single housing mkt estimate the individual marginal value (MV) functions. These are the demand curves for the attributes.
20 Everyone the same, again. Implications Because we only observe each individual at one point on his or her demand curve, we can not determine the shape of the entire demand curve.Everyone the same, again.Implicationsreally only have marginal values for each personforced to use hedonic to approximate non-marginal changesneed multi-market data to identify marginal value’s (at least this is the major way to fix the problem) See next slide.
21 Now we can begin to get at the shapes (“identify”) of the mv functions, and begin to think about consumer surplus w/ hedonics.
22 Repeat sale analysisHaving data on the same houses before and after an episode of some sort. New landfill, airport ….Controlling for other factors difficult
23 Wages and Land interacting C.A.Kolstad, Environmental Economics,Oxford U. Press, 2000.C.A.Kolstad, Environmental Economics,Oxford U. Press, 2000.Pollution is productivePollution is not productive
24 5. Valuation/Consumer Surplus Marginal changesUse marginal values from the hedonicNon-marginal changesShort-runABConsider an increase in quality(as a housing attribute)Area A is the short run benefitsof the improvement. Discuss.Since mv is difficult to measure, often useArea A+B as an upperbound.What if z declines?
25 Long run Gets complicated because hedonic shifts and people move Consider this shift from a clean-up
27 6. Wage Hedonics and the Value of Life The BasicsSame basic analysis w/ job mkt dataInterested in valuing small reductions in risk of deathMany regs are concerned w/ health and associated risk reductionair pollutionhazardous waste removaltransportation safetyhealth care initativesHedonic wages studies are one way to infer value for risk of death
28 Intuition and Story Scaffolding story Wage compensating differential DynamicsRisk of InjuryRisk of DeathVector of Job Characteristics:Union (yes/no)SkilledManagementLocation and so onVector of worker characteristics:AgeExperienceEducation and so onWage Hedonic
29 Specify functional form and focus on parameter on the rd variable Many of the same issues here as in property value studiescollinearityidentificationStatistical LivesCommon ParlanceDefinition (by example)Population 100,000Risk of Death per Person .003In this case we say 300 statistical lives will be lost.100,000 * .003 = 300We say that these are not “identified” lives
30 $500,000 is the Value of a Statistical Life Assume we can reduce risks by introducing some regulations. Say risks drop from .003 to .002.After the regulation, 200 statistical lives are lost in the populationWe say that the regulation saved 100 statistical livesAssume in a top notch wage hedonic study that we discover individuals are willing to pay $500 (revealed in the job market) for a reduction of .001.If so, in a population of 1,000 people, one statistical life will be saved. Each of the 1,000 persons has an ex ante value for risk reduction .001 of $500.Ex antereally value risk reduction$500,000 is the Value of a Statistical Life
31 Cost per life saved studies and use of value of life estimates Issues EPA, OSHA, and so onCost per life saved studies and use of value of life estimatesIssuesVoluntaryIllness and quality of life