1. 1 Cost-Utility Analysis Cost-Effectiveness Analysis Scott Matthews Courses: 12-706 / 73-359 / 19-702 Lecture 21 - 11/9/2005

2. 12-706 and 73-3592 Utility  Last time: eliciting and using individual utility functions to make decisions  Is there a similar concept to help us make decisions at the social level?

3. 12-706 and 73-3593 Specifics on Saving Lives  Cost-Utility Analysis  Quantity and quality of lives important  Just like discounting, lives are not equal  Back to the developing/developed example  But also: YEARS are not equal  Young lives “more important” than old  Cutting short a year of life for us vs  Cutting short a year of life for 85-year-old  Often look at ‘life years’ rather than ‘lives’ saved.. These values also get discounted

4. 12-706 and 73-3594 Contingent Valuation  Analysis method used when there is no observable market  Example: water quality at national parks  Asks questions to population  Is a last resort option!  Called ‘contingent’ since you never really pay  Valuing use non-controversial  Valuing ‘non-use’ VERY controversial

5. 12-706 and 73-3595 Example  Asked for valuations of a certain good  Then estimate overall WTP for it - similar to travel time demand functions  Extrapolated to entire population  Assumes random sample!

6. 12-706 and 73-3596 Criticisms of CV  Extrapolation of ‘all CV studies’ to average consumer would take over their budget  Normal statistical problems (sampling, non-response, biases, etc.)  Surveying opinions is imprecise  Problems tend to be complicated

7. 12-706 and 73-3597 WTP versus WTA  Economics implies that WTP should be equal to ‘willingness to accept’ loss  Turns out people want MUCH MORE in compensation for losing something  WTA is factor of 4-15 higher than WTP!  Also see discrepancy shrink with experience  WTP formats should be used in CVs  Only can compare amongst individuals

8. 12-706 and 73-3598 Measuring Lives Saved  Life years (prevented fatalities) not equal  Qualitative and quantitative issue  Need to consider tradeoffs  Simple example from Boardman  Status quo: no newborns survive a condition  Alt. A: 5 live, but with permanent disability  Alt. B: 2 live, but with low levels of disability  Which option (SQ, A, B) is preferable?  Assume Y increasing, H increasing  Equal costs, no relevant uncertainty

9. 12-706 and 73-3599 Simple Example

10. 12-706 and 73-35910 The Quality/Quantity Game  Assume “preference” for  Increased number of years lived  Increased level of health  Would your preferences be the same?  If so, SQ “dominated” by A and B  Note different horizontal/vertical preference  But which of A or B is better?  We all understand difference in years  Need an index of health status

11. 12-706 and 73-35911 Health Status Index Death 0 Severely Disabled Minimally Disabled HealthModerately Disabled 0.150.470.921  Measures utility, derived from experts  But this says nothing about tradeoff!  Can perform tradeoff survey  Value of “shorter Y, higher H” vs. opposite

12. 12-706 and 73-35912 Methods  Health Rating method (see above)  Time tradeoff method  Standard gamble method  Discounting life years  Can/should we discount them?  Unlike cash values, we can’t make a decision to trade 1 year today for 10 yrs from now

13. 12-706 and 73-35913 Cost-Effectiveness Testing  Generally, use when:  Considering externality effects or damages  Could be environmental, safety, etc.  Benefits able to be reduced to one dimension  Alternatives give same result - e.g. ‘reduced x’  Benefit-Cost Analysis otherwise difficult/impossible  Instead of finding NB, find “cheapest”  Want greatest bang for the buck  Find cost “per unit benefit” (e.g. lives saved)  Allows us to NOT include ‘social costs’

14. 12-706 and 73-35914 Why CEA instead of CBA?  Similar to comments on MCDM  Constraints may limit ability to perform  Monetizing maybe difficult or controversial  Easy to find lives saved, hard to judge value  Monetizing can’t capture total social value or distorts its value

15. 12-706 and 73-35915 The CEA ratios  CE = C/E  Equals cost “per unit of effectiveness”  e.g. $ per lives saved, tons CO2 reduced  Want to minimize CE (cheapest is best)  EC = E/C  Effectiveness per unit cost  e.g. Lives saved per dollar  Want to maximize EC  No practical difference between 2 ratios

16. 12-706 and 73-35916 An Obvious Example

17. 12-706 and 73-35917 Another Obvious One

18. 12-706 and 73-35918 Comments on Obvious Examples  Each had 2 dominated alternatives  Could easily identify best CE/EC option  Also had fixed scale  Fixed cost scale in first  Fixed effectiveness in second

19. 12-706 and 73-35919 Interesting Example

20. 12-706 and 73-35920 Lessons Learned  Ratios still tend to hide results  Do not take into account scale issues  CBA might have shown Option B to be better (more lives saved)  Tend to only consider budgetary costs  CEA used with constraints?  Minimize C s.t. E > E*  Min. effectiveness level (prev slide)  Find least costly way to achieve it  Minimize CE s.t. E > E*  Generally -> higher levels of C and E!  Can have similar rules to constrain cost

21. 12-706 and 73-35921 Sample Applications  Cost-effectiveness of:  New drug/medical therapies* very popular  Pollution prevention  Safety regulations

22. 12-706 and 73-35922 Definitions  Overall cost-effectiveness is the ratio of the annualized cost to the quantity of effectiveness benefit.  Incremental cost-effectiveness is the difference in costs divided by the difference in effectiveness that results from comparing one option to another, or to a benchmark measure.

23. 12-706 and 73-35923 Incremental CE  To find incremental cost-effectiveness :  Sort alternatives by ‘increasing effectiveness’  TAC = total annualized cost of compliance  PE = effectiveness (e.g. benefit measure)  CE = (TAC k – TAC k-1 )/( PE k – PE k-1 )  CE = incremental cost-effectiveness of Option k  Use zero values (if applicable) for base case

24. 12-706 and 73-35924 Incremental CE Example  Inc CE here only relevant within control categories (metals v. oils v. org’s)  ** Negative CE means option has more removals at lower cost  Source: US EPA Office of Water EPA 821-R-98-018, “Cost Effectiveness Analysis of Effluent Limitations Guidelines and Standards for the Centralized Waste Treatment Industry”

25. 12-706 and 73-35925 Definitions (2)  Marginal cost-effectiveness refers to the change in costs and benefits from a one- unit expansion or contraction of service from a particular intervention (e.g. an extra pound of emissions, an extra fatality avoided).

26. 12-706 and 73-35926 Why is CEA so relevant for public policy analysis?  Limited resources!  Opportunity cost of public spending  i.e. if we spend $100 M with agency A, its $100 M we cannot spend elsewhere  There is no federal rule saying ‘each million dollars spent must save x lives’

27. 12-706 and 73-35927 Gray Areas  How to measure cost-effectiveness when there is a single project cost but multiple effectiveness categories  E.g. fatalities and injuries, CO 2 and SO 2  Alternatives:  Keep same cost, divide by each benefit  Overstates costs for each  Keep same cost, divide by ‘sum of benefits’  Allocate cost, divide by each benefit separately  Weight the costs and/or benefits

28. 12-706 and 73-35928 Another CEA Example  Automated defribillators in community  http://www.early-defib.org/03_06_09.html http://www.early-defib.org/03_06_09.html  What would costs be?  What is effectiveness?