Lecture draws on:
Allcott and Greenstone (JEP 2012)
Allcott (Annual Review 2014)
Gerarden, Newell and Stavins (JEL 2018)
Source: Allcott and Greenstone (2012)
Serious interest began during oil crisis in 1970s
Engineers concluded that many energy saving technologies were slow to diffuse
Over time interest has centered on apparent "win-win" from environmental perspective
agency issues
resale issues
engineering models wrong
[present bias]
inattention/ salience
cognitive issues /confusion
Consumers have indirect utility
Assuming $g$ is calculated and discounted appropriately, a natural test is to estimate $\gamma$ and test if its equal to 1.
Note
Hausman (1979) estimates a discrete choice model for AC's
However, $g_{ij}$ varies for many other reasons within product
Typical logit share identity (Berry, 1994):
AW actually estimate:
Why do they do this?
Berry equation rearranged is
group all cars by month, above / below median
instrument for $G$ with $Z$
why do they do this?
Jerry Hausman's (2001) "iron law of econometrics": due to measurement error, the magnitude of a parameter estimate is usually smaller in absolute value than expected
What is the null hypothesis that AW want to test?
tradeoff implies discount rate of about 15%
results sensitive to how G is constructed
are people making mistakes?
what questions does this leave open?
Many studies similar to AW
Although panel data helps, interpretation still requires econometrician to construct $g$
If hypothesis is that $\gamma < 1$ due to inattention, imperfect information or bounded rationality, an alternative is to experimentally vary exactly those margins
Experiment #1:
Experiment #2:
can ask people about consideration sets
can ask about beliefs
can elicit WTP
assertion that consumers making large mistakes on average appears incorrect
once you account for product unobservables and use exogenous fuel cost variation, tradeoffs seem close to rational
experimental studies directly educating or directing consumer attention to energy costs have found very small impacts
Heterogeneity in both values and bias
What role do firms play here?
How do consumer's form beliefs?
develop a theoretical framework in presence of heterogenous bias and tastes
implement two experiments informing consumers about CFLs
evaluate welfare effects:
Pigou (1920):
If damages are heterogenous, first best achieved with polluter specific tax
Diamond (BJE 1973): If damages are heterogeneous, but you can only set one tax rate, it should be equal to the average damages at the margin when the tax is implemented.
Setup
unit demand: $j \in {E,I}$
utility: $u_j = v_j + z - P_j$
choose $E$ if $v - b > p$
Proposition 1:
where $B = E(b|v - b = p)$ is the average marginal bias -- ie the average bias for consumers on the margin the (subsidized) price
Optimal subsidy:
Need not be constant
How can we estimate the average marginal bias?
Option 1:
Option 2:
Artefactual field experiment
Give consumers $10
Elicit WTP ($v$) for CFLs
Inform some consumers about cost to own to recover $b$
consumers of type $j$ have distortions $d$ (nests bias from AT)
population average $\bar d = \sum \alpha_j d_j$
define targeting as:
Result 1:
Poor targeting reduces welfare gains
Result 2:
Optimal poorly targeted subsidy could be small, even if average bias is large
[ Optimal subsidy increasing in $\tau(s)$ ]
Optimal subsidy depends on average marginal bias
Not sufficient to know bias and responsiveness separately: need to show biased consumers are actually the onces affected by the policy.
getting institutional buy in was tough
compliance a big issue
Allcott & Sweeney: Sales agents can target. Can we leverage this somehow?
Houde (2014): firms bunch product characteristics around subsidy / label cutoffs
Houde (2018): labeling may further reduce average purchased quality