to 49% free-rider fraction. Moving further along the abscissa yields a further
exclusion of households that are marked by the histogram. At its peak, which
corresponds to a share of 50% hidden cost, the estimated fraction of free-riders
is still non-negligible, reaching roughly 20% of the sample of western households.
This fraction approaches zero only when the hidden cost comprises up to 100%
of observed cost.
A similar pattern for a different retrofit alternative is seen in Figure 3(b),
which shows the roof-window-facade-heating option. Even when hidden costs
comprise the sizeable share of 50% of observed costs, the corresponding share of
free-riders is substantial at roughly 38%.
We thus conclude that our results call into question the logic of providing
renovation grants to households. Nearly half of the households show a WTP
larger than the required observed investment cost, a result that is reduced only
marginally when hidden costs are taken into account. As such households cannot
be identified in advance, the awarded grants are likely to be exposed to extensive
free riding.
4 Conclusions
This paper has estimated willingness-to-pay for energy savings that accompany
a building’s retrofit. Using revealed choice data from a survey among German
homeowners, we rely on the random-utility framework to capture individual and
choice alternative attributes that determine the decision process. Starting with
the standard conditional logit model, we augment the model’s flexibility by first
allowing for preference heterogeneity using the random parameters logit model,
and second imposing a structure to capture correlation among the utility of the
alternatives with the error components logit model. We find that the conditional
and the random parameters logit model yield almost identical results, while the
error components logit model gives the best fit to the data at hand. Thus, we
22