Cost-Benefit Analysis March 30 2017 MSSM Program Columbia

Cost-Benefit Analysis March 30, 2017 MSSM Program Columbia University Satyajit Bose

Announcements • Topic Choices: – – – Multifamily solar Vertical forests California High Speed Rail Fairtrade certification Cheesecake Factory caloric reduction • Group presentation schedules posted. • Outlines due Apr 2. (No grade, just feedback). • Please schedule feedback appointment (4/3 to 4/7) for Outlines.

Announcements • General Advice for Outline: – Review detailed instructions for Outlines on Canvas – Identify audience and write with the appropriate tone and language – Work on identifying data sources for quantification early in the process – Search Econlit and use a variety of sources – Eliminate typos • Today: Stated Preference

Stated Preference • Component of value (“Nonuse value”) that does not leave behavioral trace • No obvious way to value preferences by observing behaviour (“revealed preference”) • Only approach is direct elicitation of valuation (“contingent valuation”)

Values and Behaviour V. Kerry Smith: Values influence behaviour in at least 3 ways: 1. Utility function market choices 2. Values conversations 3. Values adaptations made upon learning Economist gingerly attitude to 2. and 3.

Measuring Environmental Demand • Measuring demand without markets and without utility functions: 1. Revealed preference or market methods. (use values only) a) Hedonic price models b) Household production models (travel cost, defensive expenditure, ecosystem service models) 2. Stated preference methods (use and non-use values) a) Contingent valuation methods b) Experimental markets c) Political referenda

Stated Preference • • • Revealed preference methods can only be used in a limited subset of valuation problems Existence values leaves little or no behavioural trail Examples: 1. Exxon Valdez: concern for sea otters greater than concern for recreational fishing. 2. Groundwater contamination 3. Biodiversity

Stated Preference Steps 1. Identify a sample of respondents from the population. 2. Ask respondents to anticipate reactions to conditions that have not occurred or ask questions directly about valuation. 3. Estimate respondents' WTP for the good using information from the survey. 4. Extrapolate the results to the entire population.

Stated Preference Methods 1. Open-ended WTP Method: ask to state max WTP (high variation). 2. Close-Ended Iterative Bidding Method: Bid up or down till respondent switches from yes to no. (very sensitive to initial bid) 3. Contingent Ranking Method: Ask respondent to rank many feasible combinations of good and money. (very sensitive to order of alternatives) 4. Dichotomous Choice Method.

Dichotomous Choice

Survey Question Construction 1. Include personal demographic items: age, sex, education, occupation, perhaps income and race (ask in the middle or end) 2. Usage, satisfaction, political attitudes (ask with discretion) 3. Do not ask for long term recall (greater than 6 months) 4. Grouping and ordering to avoid response sets while easing responses

Survey Design Issues 1. Interviewer Bias: In person vs Telephone vs Mail vs Internet 2. Sample and Non-Sample Bias: Random sampling, with or without stratification. 3. Exclude the following (Build in questions to detect these types): 1. ‘conscientious objectors’ 2. Respondents who are not serious about survey 3. Respondents who do not understand the survey

Survey Delivery Simple Surveys Complex Surveys

Pitfalls: Protest Zeroes Atkinson et al (2012) Some respondents say WTP = 0 even if WTP > 0 because: • Objection to policy intervention or payment vehicle • Lack of comprehension • Ethical objection to valuation • Free riding • Etc. Hence, need to identify and eliminate protest responses Sometimes, ‘cheap talk’ or ‘entreaties’ can reduce protest responses (but might also bias responses).

Pitfalls: Protest Zeroes Atkinson et al (2012) Study 1: • WTP of UK residents for income tax funded conservation for biodiversity in tropical countries. • Cheap talk designed to eliminate impact of protests over implementation, prevention of corruption etc. Study 2: • WTP of Thames water catchment residents for water storage infrastructure funded by higher water bills. • Cheap talk designed to eliminate perception of profiteering by privatized water companies.

Pitfalls: Protest Zeroes

Question Design Exercise You survey online purchasers of jeans at gap. com. Design up to 3 questions intended to elicit how much they are willing to pay to ensure that their jeans are sourced from factories which comply with building and fire codes as stringent as those in the consuming country? 1. Do you ask before, during or after purchase? 2. What considerations, other than information gathering, will affect your survey design?

CV Design Issues 1. 2. 3. 4. 5. 6. 7. 8. Hypotheticality, meaning and context problems Neutrality of phrasing Noncommitment bias Order effect Embedding effects Anchoring WTP vs. WTA Strategic responses

Stated Preference Methods • Initial concern that there would be strategic responses turns out to be unnecessary • Careful design and implementation necessary for defensibility (NOAA panel recommendations): – clearly communicated definition of the ‘good’ – in-person interviews preferred to telephone calls

Heuristics for Survey Design There are five questions for evaluating CV surveys: 1. Do respondents understand the good being valued? 2. Do respondents have experience in valuation and choice procedures? 3. Is there clarity about the details of the project? 4. Does the survey ask for WTP rather than WTA? 5. Does the survey instrument avoid anchoring and starting point bias?

Heuristics for Using Previous Surveys • Specify which studies and specific estimates are used. • Specify assumptions made in the extrapolations. • Specify any quality changes involved. • Specify distinctions between use and nonuse components. • Perform sensitivity analysis. • Specify any potential remaining biases.

Software Competencies • • Survey Monkey Question Pro Qualtrics Google Forms

Example Quick Survey • 1. How important is it to you where your clothing is produced? (5 point scale) • How important are the working conditions under which your clothing is produced? (5 point scale) • How much more would you be willing to pay for your clothing to ensure they are produced under the same working conditions you receive? – – – $5 $10 $20 $30 $0 ASK AFTER the purchase

Experimental Markets • Construct a laboratory or field experiment where real resources are committed to choices: – E. g. Intrade (shut down), Iowa Electronic Market, Predict It • Political referenda with policy proposal and method of finance

Critiques • Overstatement of WTP relative to market estimates • Violation of rationality? • Total WTP for many programs adds up to more than budget • Unclear how much information provided is absorbed • Embedding problems

Critiques • The “Solomonic” justice system creates incentives for plaintiffs and defendants to take extreme positions • Orders of magnitude differences supported by plausible differences in assumptions: – e. g $900 million vs $100 million in valuation of commercial fishing losses in Exxon Valdez spill • False starts and genuine differences of opinion will be magnified by lawyers shopping for opinions.

Benefits Transfer • No need to invent the wheel • Use previously estimated shadow prices, rather than re-estimate for every CBA. • Databases of previously estimated shadow prices. • Search for comparable sites. • Apply appropriate adjustments. • Validate calibrated models.

A Starting Point Now subscription-based at esvaluation. org Source: Mc. Comb, Van Lantz et al. 2006 29

Estimate Consumer Surplus • Need to estimate demand curve and price elasticity. Own-price elasticities, crosselasticities and income elasticities of demand for various goods available in topic-specific journals. • Needs to predict the project impacts and multiply them by the appropriate market price or shadow price. Market prices are readily available. Shadow prices are not.

Values of a Statistical Life Definition • Implied value of life computed from amount of wealth an average individual would give up for a given reduction in risk of death. • NOT amount of wealth any specific individual would accept for certain death.

Values of a Statistical Life Methods: • Hedonic pricing method: examine wage premia for risky jobs or evaluate consumer behaviour • Contingent valuation: Directly elicit willingness to pay or accept valuation amounts with survey questions. • Meta-Analyses: Surveys of previous estimates.

Value of a Statistical Life Compensation Risk of Death

Value of a Statistical Life • Suppose the market wage increases by $600 to compensate for an increase in risk from 10 -4 to 2× 10 -4. The VSL would be 600 ÷ (2× 10 -4 -10 -4 ) = $6, 000

Hedonic Function for VSL

Values of a Statistical Life • Speed vs. Risk • Confounding Factors: – Road condition – Congestion – Type of Car – Suggested benchmarks – Enforcement

Values of a Statistical Life Measurement Problems: • Endogeneity of risks: a minesweeper in the army vs. a bank CEO. • Risk appetite variation. • Unknown risks. • Agency problems.

Values of a Statistical Life

Values of a Life Year VLY is the constant annual amount which, taken over a person’s remaining life span, has a PV equal to VSL. Assuming the VLY is constant it can be computed from an estimate of the average VSL: where, A(T-a, r) is the annuity factor based on the expected number of remaining years of life, T-a, and the discount rate, r.

Questions 1. Does VSL decline with age? 2. Should VLY be assumed constant?

Value of Time

Value of Recreation

Value of Nature

Income Adjustment To compute the VSL for Canada, denoted VCAN, given the VSL in the US: VCAN = VUS + e. IVUS(ICAN - IUS)/IUS where e. I is the income elasticity of the VSL; and ICAN and IUS denote average incomes in Canada and the US, respectively.

Income Adjustment • Example: computing adjusted damages from artificial nitrogen fertilizer emissions in Mexico, using data from the EU. – Estimates from Von Blottnitz, Rabl et al. (2006): damage 0. 3 Euros per kg of Nitrogen – EU per capita GDP 2013 $35, 551 – Mexico per capita GDP 2013 $10, 307 – PPP vs current $ – Ratio of means vs ratio of medians

Example For Wildfarm • See spreadsheet.

Other Adjustments Physical Characteristics: population density, climate, topography, and present level of the good. Project Differences: price and availability of substitutes, or non-linearity in the relationship between the good and the WTP Temporal Changes: Index to GDP Deflator

Cultural Adjustments Source: Hynes, Norton & Hanley. Adjusting for Cultural Differences in International Benefit Transfer. Environ Resource Econ (2013)

Quick Exercise • What would be the VSL to use for the group studying the value of cancer drugs pricing? – Use the DOT guidance for a specific year in future, with $9. 1 million in 2012 dollars with a 1. 07% annual increase.

Quick Exercise • What would be the social cost of carbon to use for the Hi-speed Rail groups? Source: Technical Update of the Social Cost of Carbon for RIA under EO 12866 (2013)