Essay！！

Guessing Benefits

Richard Martin

University of Victoria

October 16, 2020

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 1 / 18

Review

Review:

If emissions are continuous we equate the marginal damages (MD) and the marginal abatement cost (MAC) to find the socially optimal emissions. If options are discrete we calculate the net benefits of all the options and choose the option with the largest net benefit. In both cases we need to have a measure of benefits... but there is no mention of benefits when emissions are continuous??? the marginal damages of emissions = marginal benefit of abatement.

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 2 / 18

Review

$

emissions

Figure 1: Total benefit of reduction in emissions.

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 3 / 18

Estimating Benefits

In this chapter we illustrate some of the techniques environmental economists use to estimate

1 marginal damages 2 marginal benefit of abatement 3 marginal willingness to pay for an improvement in environmental

quality.

...which are actually all the same thing.

Valuing the environment is difficult.

For ordinary goods we can use market data to estimate the demand curve. The problem is that there is no market for environmental quality. Estimating WTP for improvement in environmental quality requires us to be clever: we typically infer willingness to pay based on individual behaviour that is related to environmental quality.

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 4 / 18

Main Approaches

There are two main approaches to estimating damages: 1 Direct methods (using market prices of goods and services affected by

the emissions.) 2 Indirect methods (imputing willingness to pay based on behaviour or

answers to questioning.)

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 5 / 18

Direct Methods

In direct methods that focus on health costs “we”: 1 measure emissions. 2 measure the resulting ambient quality. 3 estimate human exposure. 4 measure the impact of exposure. 5 estimate the dollar value of these impacts.

“we” because economists only do 5: physical scientists do 1-3 and epidemiologists do 4.

We have already seen that the jump from 1-2 is difficult in situations with multiple sources of the pollutant.

The leap from item 3 to item 4 is the real challenge: this is known as dose-response.

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 6 / 18

Direct Methods Dose-Response

Suppose your goal was to discover how much your own health has suffered due to exposure to pollution.

A problem you face is that you only observe the factual: your health given the pollution that you have been exposed to. You do not observe the counter-factual: what your health would have been if you were exposed to a different level of pollution. This problem is known as “The fundamental problem of causal inference”: it is impossible to observe a causal effect on a single entity.

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 7 / 18

Direct Methods Dose-Response

We can however observe the average causal effects via a controlled randomized experiment.

In a controlled randomized experiment we would randomly allocate individuals to either be exposed to pollution (the treatment group) or a clean environment (the control group). If we observed a difference in average health outcomes this would identify the average causal effect of being exposed to pollution. Can we randomly allocate individuals to different doses of pollution?

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 8 / 18

Direct Methods Dose-Response

Maybe we could just compare the health of people that are exposed to pollution to the health of those people who are not exposed to pollution?

If people are not randomly allocated to clean/dirty environments then we are almost guaranteed to have selection bias. Selection bias is due to systematic differences between control/treatment groups responsible for differences in health outcomes beyond the treatment effect. If these other differences are observable they can be controlled for in your statistical analysis. ... but there is always the possibility that there is an unobserved variable that is biasing the results.

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 9 / 18

Direct Methods Dose-Response

Example of selection bias: Don’t go to the hospital, they make people sicker!

If you naively compare the health of people who have had a hospital visit in the last year to the health of people who have not been to the hospital, the people who have gone to a hospital are less healthy. Is this because they went to the hospital?

Controlled randomized experiment are the gold standard when investigating causal effects.

Given successful randomization a significant difference in average outcomes must be due to the treatment. But we can’t do controlled randomized experiments on humans with things that hurt people...

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 10 / 18

Direct Methods Dose-Response

But the government can! We can look for natural experiments:

E.g. consider location choice for a new garbage incinerator. Suppose you knew where a new garbage incinerator was going to be built before anyone else and you wanted to measure the impact it would have on health. You would take the following steps.

1 measure the health of the people who live around the future site and the health of a similar group located elsewhere.

2 build the incinerator and start polluting. 3 wait. 4 measure health again for everyone you initially measured.

If the health of those who were exposed to the pollution degraded more than the health of those who were not exposed to the pollution have we identified the average causal effect of exposure to the incinerator? There are clever statistical techniques to overcome this problem.1

1 http://www.nber.org/papers/t0136.pdf

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 11 / 18

Direct Methods Valuing effects

Even if we get the dose-response figured out, how do put a dollar value on sickness (morbidity) and premature death (mortality)?

While an individual might be willing to sacrifice everything to avoid immediate death, as a society we make different choices. Motor vehicle accidents are the leading cause of accidental death, even for men.2

How much would you be willing sacrifice to guarantee that no one would every die in a motor vehicle accident? A universal 5km/h speed limit would do it. We can use motor vehicle fatality data when compared to travel cost to get a dollar value on how much our society values a life / traffic fatality.

2 http://www.bitrebels.com/entertainment/

this-is-why-men-die-before-women-14-pics/ Richard Martin (University of Victoria) Chapter 7 October 16, 2020 12 / 18

Direct Methods Valuing effects

The dollar value of sickness can be estimated using direct methods. E.g.

Diminished productivity either at work or due to absence. Increased medical expenditures (hospitals, doctors, drugs) to treat illness.

How good of a job would these measures do at capturing your maximal willingness to pay to avoid sickness?

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 13 / 18

Indirect Methods ∆ Producer surplus

Damages can also be suffered by firms:

The loss in producer surplus associated with the pollution yields the firm’s maximum willingness to pay to avoid pollution. An easier method is to give the dollar value of output reduction, but this does not correspond to the firm’s maximum willingness to pay to avoid the pollution... so it is wrong.

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 14 / 18

Indirect Methods ∆ Producer surplus

$

Q

Figure 2: Loss in producer surplus due to pollution.

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 15 / 18

Indirect Methods ∆ Consumer Surplus

Same concept for consumers: Max WTP for improvement in environmental quality equal to difference in consumer surplus.

Consumers surplus is difference between the max WTP and the price paid for each unit of the good. Recall that environmental quality is a public good: it is non-excludable, which implies that price is zero. Consumer surplus associated with a change in environmental quality is the area under the demand curve (for environmental quality) between two different levels of environmental quality. Figure 7.3 in the textbook is wrong, as is some of the discussion pertaining Figure 7.3.

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 16 / 18

Indirect Methods ∆ Consumer Surplus

$

environmental quality

Figure 3: Gain in CS due to improvement in environmental quality.

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 17 / 18

Indirect Methods Techniques

The only problem is that there is no market for environmental quality, so no direct way to estimate the demand curve for environmental quality.

So instead we use differences in consumer surplus in related markets as a (partial) measure of the willingness to pay for an improvement in environmental quality. We consider the following 3 techniques:

1 Mitigating expenditures. 2 Hedonic estimation. 3 Surrogate markets.

Richard Martin (University of Victoria) Chapter 7 October 16, 2020 18 / 18