Last class: Economic efficiency requires we set the marginal benefits from environmental protection equal to the marginal costs.
To do this, we need to measure benefits and costs
Companies use demand curves to set prices and make investments.
Imagine we had internal projections from Bird:
Two points (12, 50K) and (9, 100K)
Need to decide if we care about Bird profits or not.
[If Cambridge cared about Bird profits, they'd use marginal cost not price]
What we'd like to know is how the quantity of environmental quality demanded changes as it becomes more or less expensive.
Challenge: People don't pay for the environment.
Revealed Preference Methods: Use people’s observed behavior in markets to infer their WTP for environmental goods/services
Intuition: Even if people don't pay for environment, often spend money (incur real costs) to get access to clean environment (or avoid poor environment)
Stated Preference Methods - Design surveys that ask people what they would be WTP or WTA
Use Value: The benefits from using a good/service, directly or indirectly
Revealed preference methods preferred for measuring use values
Non-Use Value: Utility people gain from environmental goods that do not benefit them directly or indirectly
For non-use value, we have no choice – must use stated preference methods (measure total value)
– people often combine a private good with an environmental good to produce another good, which is the real source of utility
e.g., travel + wilderness area = recreational day
Idea: If we know how to value the costly input, then we can infer the value of the environmental good.
[Not most useful method, but good introduction to revealed preference models]
Origin of method (Hotelling-Clawson-Knetsch)
– Letter in 1954 to six economists from Director, National Park Service
– Response from Harold Hotelling (UNC)
Conceptual Thought Experiment
– Tuolumne River: to estimate demand function, to ascertain WTP: “Build a fence and charge admission”
– Travel Cost (including opportunity cost of time)
– # of visits from various origins (zones),
– population of zones (for example, treat all zones as having same population)
Intuition: Visiting a park takes time and money.
– Time to drive, gas, etc.
– Plus entrance fees.
– The more awesome the park, the more I’m willing to give up to visit it.
– The closer I live to a park, the more likely I am to visit it.
We can use this to trace out a WTP curve for a park
– Number of visitors from towns at different distances away
– Distance to park gives variation in price
– Variation in visitors gives variation in quantity
• 3165 feet high
• Many hiking trails
• Spectacular views of Boston and Eastern Massachusetts
• Policy question: How much is Mount Monadnock worth?
[Time valued at the average hourly wage in each city (avg $27)]
To get total value, multiply by population in each city
– Hours may be fixed, so the tradeoff is with leisure time.
∗ Need to know the shadow price of leisure time.
– People may have different utility or disutility from traveling vs. working
Other Factors Matter
– Yes, of course: income, education, age, etc.
– Not a problem: use multiple regression
– TCM very useful for understanding concept of revealed preference
– Sophisticated versions still used to value recreation sites
– But the method is going to be of limited value for estimating benefits of many environmental policies
Limited applicability: Only recreational sites or environmental quality associated with recreational sites
Most economic goods are rival -- if one person consumes something, no one else can.
Many environmental goods are non-rival, meaning if one person uses something, it doesn't impact the consumption value for anyone else:
Non-rival goods commonly called public goods
benefit estimation typically occurs at the micro level
by for policy evaluation, we're often interested in calculating total benefits
to do this, we need to aggregate the demand curves of all affected parties
– solve for
– sum up over all at the same to get
– now invert again to get
• A really likes apples:
• B likes them less:
Graph these. What does total (aggregate) demand look like?
Solve for Q
Add curves if Q > 0
[Can't have negative demand]
Now invert back to graph:
– neither A nor B ever see it, but get utility from its existence
What does the aggregate demand curve look like now?
Assume same demand curves:
• A gets utility up to the point where 20 are saved:
• B only cares as long as there are 10:
Now we actually do want to add vertically
To get the total consumer benefits of policy, we want the area under the demand curve.
Often times demand curves are estimated at the individual level
[this is useful for the problem set]
goods are bundles of attributes
even if people only buy a bundle, variation in prices across similar bundles can reveal how much people value a particular attribute
this is useful for environmental economists because many prominent purchases contain the environment as a key aspect of the bundle
imagine we find two houses that are identical in every way except one is in a more polluted area (lower Air Quality Index)
the difference in sales price provides a hedonic estimate of WTP for better air quality
This is approach referred to as hedonic (property) valuation
Estimated relationship is causal
In this example, imagine the house with better air quality also has the best school system in the state.
Common solution: Multiple regression
Challenge is that lots of amenities are correlated (school quality, safety, commute etc), and can't control for everything.
Ideally we'd observe the same “subject” in two otherwise identical states of the world, one where the policy was in effect and one where it wasn't
Drug trials get as close to possible to this ideal by randomly giving some patients a placebo
Often the best we can do in public policy analysis is to look for "quasi-" experiments, where some external factor alters the exposure of some subjects and not others in an essentially random fashion.
Many sections of the U.S. coastline are severely eroding.
Climate change is expected to lead to sea level rise
Under current predictions of a 1.1 foot rise in average sea levels by the year 2100, these erosion rates will accelerate, and between 3,000 and 7,000 square miles of dry land could be lost (IPCC, 2007; Titus, 1989).”
What would be the concern here?
Finds that one-foot beach nourishment adds $42 - $68 to a home's sale price
Large, but considerably smaller results from "cross-sectional" comparison
Cost of beach nourishment is roughly $1,000,000 per mile
Implies would have to be between 210 and 342 beachfront homes per mile in order for the project to generate positive marginal benefits.
Trump administration also recently proposed cutting this program
The 1980 Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA) gave the EPA the right to place sites that pose an imminent danger on the National Priorities List (NPL)
Can we use the hedonic method to value this program?
What other variables are likely correlated with polluted site locations?
In 1983, funding initially allocated for 400 sites
1500 candidate sites identified, 690 finalists
Each finalist was given a Hazardous Ranking System score
– Cutoff: HRS> 28.5 were cleaned up; others weren't
Conventional approach to analyzing Superfund cleanup showed large gains
Using more credible approach, the estimates become economically small and statistically insignificant
Compared to the average clean up cost of $43 million, the Superfund program does not appear welfare enhancing
Can teach a whole class on causal inference in environmental policy
For this class, I mainly want you to understand that correlation isn't causation
Tried to give you some examples of what a more "credible" approach to estimating WTP might look like
conceptually powerful framework, that has been used to value many public policies (like schools, safety, etc)
estimation requires important assumptions
[Source: Berck and Helfand]
In 1987, residents of Perkasie, Pennsylvania, faced contamination of their drinking water by trichloroethylene (TCE), a toxic chemical.
In response, many of those residents bought bottled water or water filters, boiled their water, or hauled water from elsewhere.
The costs associated with these activities were estimated to average between $22 and $48 per household.
Can these values be used to provide an approximation of the amount that people in the community, on average, would pay to reduce the risk?
Premise: people may change their behavior to avoid or lessen exposure to externalities
Intuition: Can infer WTP for risk reduction from expenditures for risk-averting activities
– What information / assumptions would you need?
Assume that air filters remove 50% of indoor PM
In Shanghai, where the price is $200, 25% of households have one
In Xian, closer to the factory, the price is $100 and 50% of hh's have one
For decades China gave heavily polluting boilers to communities north of the river
– Example: bottled water may taste better, be convenient, etc.
– So we may over-estimate WTP for risk-reduction
– Example: bicycle helmets are uncomfortable
– We will under-estimate WTP for risk reduction
In another example, residents of the area around the Nak-Dong River in Korea faced industrial pollutants in the early 1990s. (Source: Berck and Helfand)
• By 1996, water quality was greatly improved and met safety standards
• Yet people still undertook risk avoidance measures.
– People were acting based on their perception of how polluted the water was.
• The estimated willingness to pay to reduce suspended solids was 3X that what would have been predicted based on safety alone
• Many pollutants result in doctor or hospital visits
• Assume we can perfectly observed all medical expenditures
• Can we use these to estimate willingness to pay?
• Chipotle example
• Does not estimate WTP/WTA, but change in explicit market costs resulting from change in incidence of illness:
– Direct health-care costs
– Indirect costs of loss of work time
– What’s left out?
• May be considered a lower bound on WTP, but empirical evidence from comparisons suggests difference can be very large!
• Sometimes used in health economics to value morbidity changes (and by courts in wrongful death cases)
• Method is not theoretically correct nor empirically reliable, but two advantages:
– Cheaper than better approaches
– Easy to explain to policy makers and general public
• So, it’s inexpensive, easy to explain, and wrong.