Matthew G. Interis, Mississippi State University

Timothy C. Haab, The Ohio State University

Willingness to Pay for Environmental Improvements

in the Presence of Warm-Glow

CNREP Meeting May 28, 2010New Orleans, LA

Why think about warm-glow?Warm-glow: the private benefit a contributor

to the public good gets from the act of giving itself (Andreoni 1990) – i.e. the warm fuzzy feeling one gets from doing something “good.”

In non-market valuation, the value of the good we traditionally seek, WTPA, is implicitly defined by:v(y-WTPA, F1, w(a0)) = v(y, F0, w(a0))

v is indirect utilityy is income F is the level of environmental qualityw is warm-glowa is a vector of past actions creating warm-glow

Why think about warm-glow?In the presence of warm-glow, what gets reported in

a contingent valuation (CV) survey is:v(y-WTPR, F1, w(a0, WTPR)) = v(y, F0, w(a0)) (or more usually, a yes/no response based on the

above)

QuestionsHow much of a factor is warm-glow?

A: it’s more important for people who have engaged in fewer past activities → Diminishing marginal utility of warm-glow actions.

Can we back out WTPA from WTPR? If so, how do they differ?A: Yes. WTPR ~ 73% greater than WTPA

Empirical settingInternet Survey of Ohio AdultsSample size 859537 completed surveysSurvey had several sections, but two are

important here:Task 1) Respondents answered yes/no whether

they would pay a higher per gallon gas price, p1, to lower a Fuel Index (FI)

FI attempts to aggregate effects of different emissions vectors resulting from different mixes of fuel consumption across the U.S.

A higher index is worse – greater risk to human health, greater strain on natural resources, greater threat of environmental damage, etc.

Empirical settingTask 2) Respondents asked whether they

would make a hypothetical contribution to a carbon offsetting organization (e.g. TerraPass)Also:

Rated themselves 0-10, on their self-imageRated a hypothetical other person who gives

some amount to offsettingThis task came after the other task.

Empirical modelIn task 1, respondents are willing to pay the higher

price if:v1 = v(p1, y, F1, w1) ≥ v(p0, y, F0, w0) = v0

where: w1 = w(a0, ∆p)w0 = w(a0)F is the fuel index valuep is the per gallon price of gas

estimated using standard random utility modelp, y, and F are easy to measure. w is difficult to

measure. more specifically, in RUM, we need a measure for ∆w =

w1 - w0 this is where task 2 is used

Empirical modelHow to measure ∆w?

No obvious way, and any attempt will have its flaws We use, from task 2:

∆w = (Rating of self – Rating of other)/(contribution of self – contribution of other) * ∆p

Interpreted as: the change in warm-glow per dollar, γ, times the change in the price of gasoline.

Obvious weaknesses:assumes people rate others similarly to how they rate

themselvescomes from a different context (contribution to carbon

offsetting)can take on a negative value (no constraint that a greater

contribution must mean a higher image)

Empirical modelLet tA be the actual price premium consumer is willing to pay

tR is the reported price premiumAssuming a linear in parameters indirect utility function,then:

where αp < 0 and αw > 0 are the marginal utilities of gas price and warm glow, respectively.

inequality holds assuming γ ≥ 0, and denominator remains positive

note: if γ = 0, then tR = tA note also: for a good with inelastic demand (i.e. gas), WTPi = ti*q ,

where i = A,R, and q is quantity of gas consumed (Johannson 1996)

pR A A

p w

t t t

Empirical modelSurvey contained questions on past environmental

behavior, (vector a): whether respondent had given money to an env. organization, whether they were a member of an env. organization, whether they had performed any env. activities, etc.

Separating people by past environmental activity, a pattern emerged that those who had done less in the past had a higher marginal utility of warm-glow, αw, and, for people who had done more in the past, αw became negative.

Warm-glow measure is composed of a warm-glow per dollar, γ, and a change in price – by themselves, one would expect these to have opposite marginal effects on utility.

Empirical resultsProbit results:

Covariate EstimateStandard Error

Intercept -0.42 0.61

p (low a) -4.63* 2.11

p (high a) -1.44 2.03

FI -0.06** 0.02

w (low a) 18.65* 9.13

w (high a) -6.99 6.74

Badness of ∆FI 0.29** 0.11

Conservative -0.15* 0.07*indicates significance at 5% level, ** 1% level. N = 196. Percent Concordant = 69.70. Sample includes only respondents for whom γ ≥ 0 (196 out of 251)

Empirical resultsLR test that there is no difference between parameters

on p and w across groups is rejected at 95% level.All signs are as expected

Signs indicate direction of marginal change in the independent variable on probability of a “yes” response

Interesting result is parameters on wPositive and significant for those with little environmental

backgroundNegative but insignificant for those with high environmental

backgroundDiminishing marginal utility of warm-glow actions

Including people with γ <0, all signs and significance remain the same except that parameter on w becomes significant for the high environmental background groupMakes sense – these people get the opposite of warm-

glow. Many possible explanations: don’t trust govt. management of funds, think contributing is for chumps, etc.

Warm-glow and WTPRecall that the reported premium, tR is greater

than actual premium, tA , by a factor of :

Using the mean value of γ and estimated values of αp and αw for the low group, the above has a value of 1.73.

i.e. for those who get more utility from the warm-glow of contributing, the reported price premium is ~ 73% greater than the premium they would be willing to pay, were they to receive no warm-glow from contributing.

wp

p

Warm-glow and WTPCalculating premium based on means of data for

the low group: tR = $0.165 , tA = $0.095

Not accounting for warm-glow at all: tA = $0.155

Nunes and Schokkaert (JEEM 2003) – reported WTP 55-270% higher than “cold” WTP

Most research has focused on finding evidence of warm-glow in decisions to contribute to a public good (e.g. Menges et al. ERE 2005, Ribar and Wilhelm JPE 2002), but hasn’t focused on determining “cold” WTP in the presence of warm-glow.

ConclusionsRespondents show decreasing marginal utility of

warm-glow activitiesFailing to account for warm-glow results in an

estimate of WTP that is ~73% higher than “true” WTP i.e. the WTP we would normally think of, that is, the

monetary payment the respondent pays that makes him just as well off as before the improvement, ceteris paribus.

More research neededWhich is the appropriate measure for practical

concerns of benefit cost analysis?It most likely depends on how the environmental good

or service is provided – whether people get a warm-glow.

Questions? Comments?

Thank you.