Download presentation
Presentation is loading. Please wait.
1
1 Extreme Events Scott Matthews Courses: 12-706 / 19-702
2
2 Admin HW 5 Due Now Group Project 2 Out today. Due Monday Nov 24 Next week: 2 case study writeups due 2 PAGES MAX !! DO NOT SUBMIT MORE!
3
Recap of Decision Trees When thinking about strategies for decisions we could make 2-way sensitivity graphs. Purpose: if parameters changed, did that affect our intended strategy? i.e., what would have to happen to change our mind about our strategy? 3
4
2-way Simple DA sensitivity 4
5
5 Extreme Events Low probability, high consequence (cost) events Natural disasters (e.g., hurricanes) Catastrophic infrastructure failure Considered hard to assess.. But can assess with sensitivity analysis: On risk tolerance / utility “how risk averse do you need to be for it to matter?” On probability - “how likely is it to happen?” On expected losses (consequences) “how much would you have to lose for it to matter?”
6
6 Relevant Thoughts As you (the decision maker) become more risk averse, you tend to worry ONLY about the worst case As you accept more risk, converge to risk neutral Example: using exponential utility (similar to Deal or No Deal) Recall definition of R parameter in function Equally willing to risk winning R or losing R/2 For individuals, generally R ~ $1000s Recall goal is to maximize CE
7
7
8
8
9
9 Example: Infrastructure Failure Probability (P) of happening: About 0.5% Damage (D) if occurs - $100 million Typical EMV = P*D = ~$500k Can pay to remove large potential cost by buying insurance (cost $20 million)
10
10 A risk neutral decision maker would ride out the risk all the way to p(fail) of about 0.2. What about a risk averse DM?
11
Extreme Events Spreadsheet Lets look at the same example, and effects from changing R, Loss, p(fail),.. What is our base case decision strategy? Does the strategy change as the parameters change? If it does not, then even though the event is “Extreme” we should be comfortable with our decision. 11
12
In-Class Case Study: Clean Air Regulation Scott Matthews Lecture 24 12-706 / 19-702
13
New Type of Problem Handout of Tables included What happens when we cannot/will not monetize all aspects of a BCA? Example: what if we are evaluating policies where a benefit is lives or injuries saved? How do we place a value on these benefits? Are there philosophical problems?
14
In-Class Case Study Consider this ‘my example’ of how to do a project for this class (if relevant) Topical issue, using course techniques As we discuss, think about whether you would do it differently, be interested in other things, etc. Metrics for this case are ugly (literally): morbidity and mortality for human health Effectively I ‘redo’ a published government report with different data
15
Background of CAA Enacted in 1970 to protect and improve air quality in the US EPA was just being born Had many sources - mobile and stationary CAA goal : reducing source emissions Cars have always been a primary target Acid rain and ozone depletion Amended in 1977 and 1990 1990 CAAA added need for CBA (retro/pro)
16
History of Lead Emissions Originally, there was lead in gasoline Studies found negative health effects Tailpipe emissions (burning gas) were seen as a primary source of lead Regulations called for phaseout of lead We have also attempted to reduce lead/increase awareness in paints, etc. Today, new cars must run on ‘unleaded’ gasoline (anyone remember both?)
17
Construction of Analyses Estimate emissions reduced since 1970 For major criteria pollutants (SO2, NOX,…) Estimated ‘no control’ scenario since 1970 Estimated expected emissions without CAA Compared to ‘actual emissions’ (measured) Found ‘net estimated reduced emissions’ Assumed no changes in population distribution, economic structure (hard) Modeled 1975/80/85/90, interpolated
18
Analyses (cont.) Estimated costs of CAA compliance Done partially with PACE data over time Also run through a macroeconomic model With reduced emissions, est. health effects Large sample of health studies linking ‘reduced emissions of x’ with asthma, stroke, death,.. Used ‘value of effects reduced’ as benefits 26 ‘value of life studies’ for reduced deaths Does a marginal amount of pollution by itself kill?
19
Value of Life Studies Used Actually should be calling these ‘studies of consumer WTP to avoid premature death’ Five were ‘contingent valuation’ studies Others estimated wage/risk premiums Mean of studies = $4.8 million (1990$) Different than “Miller” from earlier Standard dev = $3.2 million ($1990) Min $600k, Max $13.5 million ($1990)
20
Putting everything together Had Benefits in terms of ‘Value from reducing deaths and disease’ in dollars Had costs seen from pollution control Use min/median/max ranges Convert everything into $1990, get NB Median estimated at $22 trillion ($1990)! $2 trillion from reducing lead 75% from particulates Is this the best/only way to show results?
21
‘Wish List’ - added analysis Disaggregate benefits and costs by pollutant (e.g. SO2) and find NB Could then compare to existing cost- effectiveness studies that find ‘$/ton’ Disaggregate by source- mobile/stationary Could show more detailed effects of regulating point vs. non-point sources Has vehicle regulation been cost-effective? Why did they perhaps NOT do these?
22
My Own Work I replicated analysis by using only median values, assumed they were exp. Value Is this a fair/safe assumption? See Table 3
23
Implied Results
24
Recall Externality Lecture External / social costs A measure of the costs borne by society but not reflected in the prices of goods Can determine externality costs by other methods - how are they found? Similar to health effects above, but then explicitly done on a $/ton basis
25
Compare to other studies Large discrepancies between literature and EPA results! Using numbers above, median NB = $1 T
26
Source Category Analysis Using ‘our numbers’, mobile and stationary source benefits (not NB) nearly equal ($550B each in $92) See Tables 12 and 13 for costs and NB Up to 1982, stationary NB > mobile After 1982, mobile >> stationary
27
Final Thoughts EPA was required to do an analysis of effectiveness of the CAA Their results seem to raise more questions than they answer The additional measures we showed are interesting and deserve attention Questions intent of EPA’s analysis
28
Other Uses - Externality “Adders” Drop in as $$ in the cash flow of a project Determine whether amended project cash flows / NPV still positive
29
Mutiple Effectiveness Measures So far, we have considered externality problems in one of 2 ways: 1) By monetizing externality and including it explicitly as part of BCA 2) Finding cost, dividing by measured effectiveness (in non-monetary terms) While Option 2 is preferred, it is only relevant with a single effectiveness
30
MAIS Table - Used for QALY Conversions Comprehensive Fatality / Injury Values Injury Severity1994 Relative Value MAIS1.0038 MAIS2.0468 MAIS3.1655 MAIS4.4182 MAIS5.8791 Fatality1.0
31
Single vs. Multiple Effectiveness Recall earlier examples: Cost per life saved Cost per ton of pollution When discussing “500 Interventions” paper, talked about environmental regs Had mortality and morbidity benefits Very common to have multiple benefits/effectiveness Under option 1 above, we would just multiply by $/life and $/injury values.. But recall that we prefer NOT to monetize and instead find CE/EC values to compare to others
32
Multiple Effectiveness In Option 2, its not relevant to simply divide total costs (TC) by # deaths, # injuries, e.g. CE 1 = TC/death, CE 2 = TC/injury Why? Misrepresents costs of each effectiveness Instead, we need a method to allocate the costs (or to separate the benefits) so that we have CE ratios relevant to each effectiveness measure
33
Options for Better Method Use “primary target” as effectiveness Allocate all costs to it (basically what we’ve been doing) Add effectiveness measures together E.g., tons of pollution Is as ridiculous as it sounds (tons not equal, lives not equal to injuries)
34
Improved Method In absence of more information or knowing better, allocate costs evenly E.g., if 2 pollutants each gets 1/2 the cost Easy to make slight variations if new information or insight is available Could use our monetization values to inform this (e.g., external cost values, $/life values, etc.)
35
Recall from previous lecture
36
Another Option For each effectiveness, subtract marginal cost/benefit values of all other measures from total cost so that only remaining costs exist for CE ratios Again could use median $ values on previous slide to do this Examples..
37
Wrap Up There is no “accepted theory” on how to do this. However when we have multiple effectiveness measures, we need to do something so we end up with meaningful results.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.