The diffusion of voluntary green building certification: a spatial approach

Here the shares apply to all existing buildings.

The decision-making process of a commercial building involves more than one person, including building owners, managers, and other relevant personnel. In this paper, we treat the decision making of the people in a commercial building as a whole. So, when we say commercial buildings make decisions, it is equivalent to say building owners or managers in the commercial buildings make the decisions—similar to the concept of a firm’s decision making (Qiu 2014).

The low-hanging fruit refers to building characteristics that are easily observed and should be targeted by policymakers.

On the other hand, this gap may not exist, since technology-based NPV calculations may not capture all factors that economists would include.

Although this paper does not analyze a building's decision to apply for certification or not from a game-theory perspective, there are parallels that can be additive to the analysis. In particular, building certification decisions can be viewed as a simple signaling game, with the signaler being the building decision makers, the signal being whether to become certified (which includes the necessary investment that goes with it), and the "receiver" broadly interpreted as the rental market, which can reward the building with different rental incomes based on the signal (i.e., certification). In this context, the existence of accurate certification essentially constitutes a separating equilibrium of property types (or at least allow for the possibility of a separating equilibrium) in which efficient buildings would separate themselves from inefficient buildings via certification, and hence reap higher rewards from the receiver (i.e., the rental market). In the absence of accurate certification, efficiency would not be observable to the receiver, and hence, a pooling equilibrium would be reached. Consequently, the existence of "peer effects" in certification can also be viewed as a separating equilibrium that reflects the correlation between expected economic returns from the rental market and spatial effects.

While there are similarities between the decision-making process related to energy efficiency and that related to renewable energy, the implications of Reddy and Painuly (2004) might not carry over to energy efficiency.

Given that the share of green buildings is very low, as in the case of this paper, the ratio of the share of green buildings over the share of non-green buildings can function as an approximation for the share of green buildings. However, when applying the models to the case where the share of green buildings is high, such approximation will no longer be appropriate.

In this paper, we cannot use zip code-level fixed effects because the analysis is cross-sectional at zip code level. The fixed effect used in the basic spatial model is at city level.

The land value used as the instrumental variable is the per square foot land value based on the square footage of the property.

Buildings may be both Energy Star- and LEED-certified. Separately identifying the two types of certifications in the sample might be problematic because it could lead to double counting of a building that is certified by both. However in this paper, given that by definition a building is a green building if it is either Energy- Star or LEED-certified or both, we avoid double counting if a building is both Energy Star- and LEED-certified according to this definition.

The number of observations that have the information of per-square-foot-land-value (the instrumental variable) is less than the total number of observations used in the basic spatial models. Thus, the number of observations used in the IV models (Table 3) is less than that used in the basic spatial models (Table 2).

Properties that are already of high value could have lower percentage change in their value. We do not have the data of percentage change in assessed value to test this potential explanation, and thus, the implicit assumption in our models is that increases in value will be correspondingly higher for those properties that are already of high value.