Economic Valuation Test 2

Exaggerates firm loses

Degredation Production Function

Underestimates firm gains

Improving Production Function

Overestimate change in profit with DEGREATION

Underestimates change in profit with IMPROVEMENT

predicts partial impact of change in E on q: q1-x0-q0

change in consumer surplus is change in profit

x (variable input) is endogenous

Production Function

predicts full impact of change in E on profit: pi1 - pi0

no endogeneity

Profit Function

predicts partial impact of change in E on C: C0-C1|q0

Overestimate change in profit with DEGREATION

Underestimates change in profit with IMPROVEMENT

change in producer surplus is equal to change in profit

q is endogenous

Cost Function

reduce bias from models with endogenous variables (correlated with error terms means biased coefficient estimates)

Why IV?

expected value = true value

Unbiased

predicted value approaches true value

Consistency (what IV gets)

Validity and Relevance

Assumption of IV variables

instrument does not belong in model or error term (exclusion restricion) aka exogenous

only impacts dependent (q) through endogenous variable (x)

Validity

instrument impacts endogenous variable

impacts strongly

Relevance

estat overid, high p = valid

Interpreting Validity

estat first stage, f stat 10+ = strong

Interpreting Relevance

allows us to control for unobserved effects by agent (ci) and time (ot)

xtreg, re/fe

Panel data models

treats unobserved effects as part of the error term

pros: more efficient (if variance estimated well)

cons: biased estimates if ci/ot if correlated with q, w, or E (endogenous)

Random Effects

treats unobserved effects as variables

pros: unbiased estimates if unobserved effects are endogenous

cons: less efficient estimates (still more than OLS)

Fixed Effects

Use random first, if Hausman test is low, use fixed effects

Panel data procedure

to detect endogeneity in estimates

when low, reject, there is endogenity, use FE

hausman fixed/random

Hausman test

FE/RE may be heteroskedastic or spatially/temporally correllated

use robust standard errors

Panel Data Error

Open-ended, payment card, dichotomous choice

Contingent Valuation Formats

pros: greatest precision

cons: protest votes, less incentive compatability

Open-ended

pros: intermediate precision

cons: anchoring (if cluster bids)

Payment card

prefered method

pros: most truthful (incentive compatability)

cons: least precise

Dichotomous choice

validity and reliability

CV Study Validation Concerns

Are predictions true?

Main concern

Validity

Precision

CV already does well

Reliability

Respondents strategically manipulate

Incentive compatability

Respondents fail to take questions seriously

Inconsequentiality

Content, Construct, Convergent, Criterion (increasing rigor)

Validity concepts

correct procedure followed

people make uninformed choices often

Content Validity

measure correlated with theory

scope test for theory validation

Construct Validity

compare meaure to measures of the same thing

CV (SP) estimates are slightly smaller than RP

Convergent Validity

compare measure to true value

use WTP not WTA (less hypothetical bias)

Criterion Validity

-(explanatory coeffs…)/(price coeff)

Logit coefficients maximum WTP

not log-odds

“how much does the likelihood of saying “yes” increase if initial price is $1 higher”

Logit margins

elasticity

“1% increase in income, probability of saying “yes” increases by x”

income elasticity is often less than one (inelastic)

Logit margins eyex

max WTP of households ($/time)

Logit wtpcikr

estat clas

predict pr command (mean is probability of saying yes)

Logit GOF

Hypothetical bias, incentive incompatability, inconsequentiality

CV study Criticisms

respondents do not understand or have information to answer well

Hypothetical bias

Point estimate, central tendency

reported by existing studies

administratively approved

Value transfer methods

find existing studies that estimate similar environmental change

Point estimate

calculate average of set of values

accuracy is limited if policy site is very different

Central tendency

demand/benefit function, meta-analysis

use or develop function

considered more accurate but need more data

Function transfer

out of sample fitted value

viewed as more accurate (more control)

need more data

Demand/benefit function

compile value estimates across a set of valuation studies

Meta-analysis

exponent

consider similarity of study sites

use to value change in E at policy site and have information from other site

Using income elasticities to adjust value transfers

WTP to reduce mortality risk

aggregate of group willingness

Value of a statistical life

health impacts productivity

values productivity of lost work so excludes non-market output

Human capital approach

WTP for reductions in risk

RP and SP methods

VSL

Welfare approach

not limited by observations (can estimate passion/future uses)

provide hicksian measures of CS and ES

CV is effective because…

SP methods (not based on observed behavior)

CV uses…

estat endogenous interpretation

high p = endogenous