Residential Solar Photovoltaic Technology Adoption: An Empirical Investigation of State Policy Effectiveness
Christine Lasco Crago∗ and Ilya Chernyakhovskiy
Abstract
State policy incentives for solar power have grown significantly in the past several years. This paper uses county level panel data to investigate whether state policy incentives are effective in increasing residential solar PV capacity. Empirical findings show that tax incentives, rebates, solar-specific mandates, and loan financing programs are important drivers of residential PV adoption. These results suggest that policy incentives play a significant role in encouraging wider use of solar energy. Results also point to a significant positive relationship between hybrid vehicle sales and residential PV adoption, indicating the importance of pro-environmental preference as a predictor of solar PV demand.
Renewable energy has received growing support in the past several years due to concerns
about climate change and energy security. Solar photovoltaic (PV) technology is one of
the most promising sources of renewable energy. Solar energy resource is widely abundant
in the United States, and solar power generation produces less greenhouse gas emissions
relative to energy sources from fossil fuels. According to Lopez et al. (2012), customer-cited
rooftop solar PV has the technical potential to supply roughly 20% of annual electricity
demand in the United States. In 2010 the United States federal government spent $1.1 billion in financial support for solar energy development, representing more than a 5-fold
increase from 2007 levels.1 States have also played a larger role in incentivizing renewables
like solar by establishing mandates broadly for renewable energy and specifically for solar.
In addition to solar energy generation goals and mandates, generous financial incentives
∗Corresponding author. Assistant Professor, Department of Resource Economics, University of
Mas-sachusetts Amherst. Address: 217C Stockbridge Hall, 80 Campus Center Way, Amherst, MA 01003. Phone: +1 413 545-5738. Email addresses: ccrago@resecon.umass.edu (C.L. Crago), ichernya@umass.edu (I. Chernyakhovskiy).
1United States’ federal spending to promote renewable energy technology totalled $14.7 billion in 2010,
growing significantly from its level of $5.1 billion 2007. In comparison, federal spending to promote wind power totalled$4.98 billion in 2010 and$476 million in 2007 (U.S. Energy Information Adiminstration (EIA) 2011).
in the form of rebates and tax credits and exemptions are being offered. In the case of
residential applications, these financial incentives have lowered the cost of installing a solar
photovoltaic (PV) system by as much as 50%. Given the large amount of current and
projected government expenditures that go to solar incentive programs, it is important to
determine what types of incentives are effective, and to what extent they influence growth
in solar PV capacity.
Despite the fact that incentives for solar PV have become widespread, there are few
em-pirical studies that examine the effectiveness of specific policies used to incentivize residential
solar PV adoption. To date, there is no empirical investigation of the impact of solar PV
incentives that parses the effects of different state policies. This paper fills this gap in the
literature.
We use county level data spanning eight years (2005-2012) to examine the impact of state
policy incentives for residential PV while controlling for demographics, solar resources, and
environmental preferences using panel data estimation methods. Newly installed residential
solar PV capacity in kilowatts (kW) is the dependent variable of interest.
This paper improves upon the existing literature examining determinants of solar
adop-tion in a number of ways. First, individual representaadop-tion of policies enables us to investigate
the effectiveness of different policy options. Existing studies that examine the drivers of
res-idential solar energy adoption (solar hot water or PV) either lump policy incentives in one
variable (Kwan 2012) or do not consider them (Zahran et al. 2008). Others focus on the
role of peer effects in the diffusion of solar PV technology (Bollinger and Gillingham 2012;
Noll, Dawes, and Rai 2014). Second, we use panel data to examine the impact of specific
incentives on adoption. Previous studies (Kwan 2012; Zahran et al. 2008) rely on cross
sec-tion data which, while informative, does not exploit variasec-tion over time. Third, we address
a number of estimations issues. We use a Tobit model to account for the non-linear
rela-tionship between our dependent variable and explanatory variables due to the large share of
specification to address potential endogeneity in policy variables.
Our empirical results show that policy incentives are effective at increasing solar PV
capacity at the county level. Sales tax exemptions, rebate availability and loan financing
programs are significantly related to increases in PV capacity. Solar renewable energy credit
prices, which represent stringency of solar-specific mandates are also found to significantly
in-crease residential solar PV capacity. In contrast, we find that Renewable Portfolio Standards,
a technology-blind mandate does not impact PV capacity growth. These results suggest that
policy incentives play a significant role in increasing the demand for solar energy. In
par-ticular, solar-specific policies and policies that are geared toward reducing upfront cost to
consumers are most effective. Results also indicate that Democratic party affiliation as well
as demand for hybrid vehicles are positively related to solar PV adoption, suggesting the
importance of environmental preferences in residential solar PV adoption decisions.
The paper proceeds as follows: section 1 gives a brief background of solar policy incentives
at the state level. Sections 2 presents our analytical framework and section 3 discusses the
empirical model. Section 4 presents data sources. Section 5 discusses our empirical findings
and section 6 concludes.
1
Background
Over the last several years, solar PV technology has enjoyed a steep upward trend in
adop-tion across the United States (U.S.). Total operating solar PV capacity reached 8.3 GW by
the end of 2013. In 2012 the U.S. was fourth in the world PV market for total peak power
capacity behind Germany, Italy and China (Barbose, Darghouth, and Wiser 2012). Net
electricity generation from solar power sources has grown exponentially; between 2009 and
2013 annual net generation increased by roughly 5000% (U.S. Energy Information
Adimin-stration (EIA) 2014). New solar capacity additions in 2013 totalled 4,751 MW, a 41%
second-largest source of new energy generating capacity behind natural gas (U.S. Energy
Information Adiminstration (EIA) 2014).
Residential solar PV systems represents an important market segment. In 2013, 792 MW
of new capacity was added, representing 17% of solar capacity additions that year. Forecasts
of solar PV capacity additions in 2014 project growth to be most rapid in the residential
segment (GTM Research and Solar Energy Industries Association 2014).
Concurrently, solar power industry revenues, employment, and wages have increased
steadily since 2007, with a strong upward trend projected to continue uninterrupted through
2018 (Smith 2013). However, while the U.S. solar power industry has enjoyed substantial
growth in recent years, the share of solar generated electricity is small when compared with
other energy sources. Electricity generated from solar energy accounted for only 0.2% of
total electricity generated in the U.S. in 2013 (U.S. Energy Information Adiminstration
(EIA) 2014).
1.1
State policy incentives
State governments have implemented a range of policy initiatives aimed at increasing the
share of solar PV-generated electricity in the U.S. energy mix. In 2011 alone, legislatures
in ten states enacted various innovative financial incentives directed at increasing in-state
residential PV capacity (DSIRE 2013). Renewable portfolio standards have received the
most attention in case studies and empirical analyses. Renewable portfolio standards (RPS)
are regulations which mandate that a specific share of a state’s electricity sales must be
generated from renewable energy sources. Carley (2009) and Yin and Powers (2010) show
empirically that RPS implementation is associated with positive growth of in-state RE
gen-eration. Findings by Maguire (2011) and Adelaja et al. (2010) provide evidence that RPS
positively influence wind energy adoption. However, Carley (2009) points out that RPS
generally fail at achieving their stated goal of increasing the share of renewable energy
policies thus far, Wiser, Barbose, and Holt (2011) emphasize the need to include
technology-specific standards into RPS designs to promote diversity in renewable energy markets. The
authors cite evidence that RPS alone promote only a limited number of least-cost renewable
technologies. While solar PV technology is a promising source of renewable energy, the
cur-rently high cost of solar panels does not make it competitive with wind energy, hydropower
or geothermal. One strategy has been to amend existing RPS mandates to include a
“carve-out” for solar energy generation. Of the 29 states plus the District of Columbia with active
RPS policies, 16 states and the District of Columbia have included a solar carve-out. New
Jersey currently has the most aggressive support for solar energy with a carve-out of 4.1%
of electricity sales by 2028.
Generally a solar carve-out goes hand in hand with a solar renewable energy credit
(SREC) market. An SREC market provides a financial production incentive to solar-enabled
households. SRECs are solar energy production credits awarded per kilowatt hour (kWh)
to owners of grid-connected solar PV systems. Aggregation firms sell bundles of SRECs to
utilities in the state’s renewable energy credit market. Utilities are obligated to meet solar
energy sales requirements either by installing their own solar power facilities, buying SRECs
in the open market, or paying year-end alternative compliance payments. The average SREC
price per year is used to measure the effects of a state’s solar carve-out
Additional state incentives come in the form of tax policies, cash rebates, and solar
consumer support laws. Tax incentives are designed to reduce the high up-front costs of
installing a solar PV system. Tax credits and exemptions from personal, property, and sales
taxes aim to reduce the tax liability of adopters. Many states use a combination of tax-based
incentive mechanisms. Roughly twenty states offer sales tax exemptions for the cost of solar
equipment, thirty states offer property tax exemptions for the added value to homes with
solar PV systems, and twenty states and the District of Columbia provide income tax credits.
Rebate programs provide direct cash subsidies towards the up-front costs of installing
are more politically charged than other incentives as they require appropriation of state
funds. Sixteen states and the District of Columbia currently have active rebate programs.
In Massachusetts, the Commonwealth Solar Rebate II program is allotted$4 million per year since 2003, funded by a$0.0005 per kWh surcharge on electricity bills in the state. In addition to direct financial incentives, states also provide legislative support which can protect solar
PV consumers. Solar rights policies are state laws which prohibit local governments from
enacting regulations or ordinances restricting a home owner’s ability to install a solar energy
system.
Table 1.1 gives an overview of when key policies were enacted in the states included in this
study. Success with promoting residential solar PV adoption varies significantly across state
lines. As of December 2012, New Jersey had a total of 955.7 megawatts (MW) of installed
solar PV capacity, whereas New Hampshire ended the year with just 5.4 MW (NREL 2013).
After controlling for population size this is roughly a 20-fold discrepancy in solar technology
adoption.
It is unclear whether adoption trends are a manifestation of availability of solar resources
(i.e., sunshine) , individual home-owner characteristics such as income and pro-environmental
preferences, or genuine responses to state incentive policies. In a county-level study of solar
hot water system adoption using cross section data, Zahran et al. (2008) find that age
and county participation in the International Council for Local Environmental Initiatives
(ICLEI) are significant across model specifications. Policy incentives are not considered
in the analysis. Kwan (2012) examines solar PV adoption at the zip-code level using cross
section data. He finds that policy incentives have an impact on solar PV adoption. However,
Kwan (2012) uses a single aggregate incentives index to represent financial incentives, while
mandates, solar consumer rights and loan financing programs are not considered. He also
identifies solar insolation and demographic characteristics such as college graduation rates,
age, and income as strong predictors of solar PV adoption.
Table 1: State solar PV incentives Year Policy State 2005 2006 2007 2008 2009 2010 2011 2012 Solar carve-out CT DE • • • • • • MA • • • MD • • • • • • ME NH • • • NJ • • • • • • • • NY† ◦ ◦ ◦ ◦ ◦ ◦ ◦ ◦ PA • • • • • • • RI VT WV
Sales tax exemption
CT • • • • • • DE MA • • • • • • • • MD • • • • • ME NH NJ • • • • • • • • NY • • • • • • • • PA RI • • • • • • • • VT • • • • • • • • WV Rebate CT • DE • • • • • • • • MA • • • • • MD • • • • • • • • ME • • • • • • • NH • • • NJ NY • • • PA • • • • RI VT • • • • • • • • WV
This table only indicates existence of a policy. Levels of each policy vary by state and time.
research questions which have yet to be comprehensively addressed. This paper seeks to
answer whether policy incentives are effective at promoting adoption, which incentives are
most effective, and what are the prevailing characteristics of homeowners who are likely to
invest in a solar PV system.
2
Analytical Model
The empirical model is guided by an analytical model of a household’s decision to install a
PV system. Assume that the household seeks to minimize the lifetime cost of a fixed level
of electricity consumption by choosing the optimal time to invest in an energy saving device
such as a solar PV system. The household solves the following optimization problem 2:
min T T Z 0 PtEe−rtdt+ ∞ Z T (1−φ)PtEe−rtdt+ (1−π)ITe−rT (1)
where PtE is the price of electricity, φ is the proportion of electricity consumption
gen-erated from solar PV, r is the discount rate, IT is the cost of the solar PV system at the
time of installation, and π is the amount of subsidies received as a proportion of IT. The
first term in (1) is the cost of electricity prior to installation of the solar PV system. The
second term is the cost of electricity when the solar PV is in place, and the third term is the
installation cost net of subsidies.
Taking the derivative of (1) with respect to T, we obtain the condition for the household
to invest in a PV system at time T:
φPE T
r ≥(1−π)IT (2)
Equation (2) shows that a household will invest in a PV system if the discounted present
value of benefits from solar PV is greater than the net investment cost at the time of
stallation. Based on equation (2) we can identify the variables that affect the household’s
decision to install solar PV when economic factors are considered. We expect that factors
that increase discounted benefits relative to net investment cost will increase the likelihood
of adoption by a household at time T.
The higher the price of electricity the greater the likelihood that a household will choose
to invest at a given timeT. A lower discount rate (r) also encourages investment. Hausman
(1979) suggests that there is an inverse relationship between discount rates and income.
Thus, we include median home value and income as explanatory variables in our regression.
The expectation is that the higher a household’s income and home value, the greater the
likelihood of adoption.
The share of electricity generated from solar PV (and thus the cost savings attributed
to solar PV) is determined in large part by the amount of solar resources available to the
household. Thus, we include data on solar insolation. Households with greater access to solar
resources will expect greater energy bill savings, reduced payback periods on the investment,
and potentially greater returns from production-based policy incentives. We expect that
greater solar insolation would lead to greater likelihood of adoption. The lower the net
cost of installing solar PV the greater the likelihood of adoption. The installation cost data
available through the NREL OpenPV database points to limited variation in installation
costs for an average sized residential system (NREL 2013). Residential installation costs are
assumed to be constant among counties. A linear time trend variable is included in order
to capture changes to federal policies as well as falling prices of solar panels over time. The
net cost to the household also depends on the amount of financial assistance obtained from
solar incentive programs. We include data on different financial incentives that lower the net
cost of PV installation. We also include variables to represent support policies such as loan
programs.
Equation (2) helps identify financial factors that are likely to affect adoption. However,
environmen-tal externalities, contribute to the public good, gain prestige, and “feel good” about oneself
could also motivate solar PV adoption (Andreoni 1990; Sidiras and Koukios 2004; Sexton
and Sexton 2014). We use the share of hybrid-electric vehicles in new vehicle registrations
per county, percentage of households that purchase organic food, as well as the percentage of
Democratric Party votes to represent households’ pro-environmental preferences that cause
them to make green investments due to non-financial considerations. Studies have also
iden-tified peer effects and familiarity with technology as factors that influence adoption of solar
PV (Bollinger and Gillingham 2012; Noll, Dawes, and Rai 2014). As more solar panels are
installed in a neighborhood, social interactions and solar panel visibility on rooftops may
cause a reduction in perceived uncertainty about solar technology. We use a one year lag
of the number of solar installations to capture the impacts of peer effects and familiarity
with PV technology. Other studies also find that age, urbanization, and educational
at-tainment can be important factors in determining solar adoption (Kwan 2012; Zahran et al.
2008; Lutzenhiser 1993; Labay and Kinnear 1981). It is possible that these variables affect
purchasing power, discounting preferences, or environmental awareness. We include median
county age and the share of residents with a bachelors degree or higher. As a measure of
ur-banization and as control variables for available rooftop space, we include population density
and housing density for each county.
3
Empirical model and choice of estimation method
The main empirical goal of this paper is to model the drivers of solar PV capacity growth at
the county level. The analytical model shows the variables that affect a household’s decision
to adopt solar PV. We expect that the same variables (averaged at the county level) will
also determine the amount of PV capacity at the county level. Based on the analytical
model, counties receiving larger financial incentives that lower the cost of PV installation
prices, greater solar insolation and number of existing PV systems installed, and higher
proportion of residents that have higher incomes, are better educated and that have
pro-environmental preferences will have greater PV capacity.
Equation 3 is the hypothesized linear model of county solar PV capacity additions.
Yit=α+Xitβ+Pitζ+Gitκ+Iiγ+it (3)
where the dependent variable, Yit, is residential PV capacity additions in county i, in year t,
and it is the error vector. Average installed solar kW capacity across all counties and time
periods, holding all explanatory variables constant at zero, is captured by the intercept α.
Vectors Pit and Git include policy and non-policy variables that vary over time. Vector Ii
includes explanatory variables which do not vary over time: average annual solar insolation
and the percentage of votes for the democratic candidate in the 2008 presidential election.
Explanatory variables can be categorized into three groups: demographics and geographic
controls; state policies; and indicators of pro-environmental preferences. A specification
including only demographics and geographic controls as regressors is analogous to previous
studies of solar technology adoption (Zahran et al. 2008), in which the financial effects of
state policies are not considered. The independent variables included in this specification are:
average annual solar radiation (insolation), yearly average electricity price; the percentage of
homeowners with a bachelor degree or higher; median home value; median age; the number of
owner-occupied homes; population density; lagged number of solar installations; and a time
trend. A state policy specification builds on the previous model by including the following
incentive variables: SREC prices; sales tax exemption; the number of years an RPS has been
in effect; plus four binary variables indicating the availability of: a personal tax credit; cash
rebates; solar rights regulations; and loan programs. The final specification includes two
indicators of pro-environmental preferences: the proportion of hybrid vehicles out of total
auto sales, and the proportion of democratic party voters. 3
3.1
Tobit estimator
In the solar technology adoption context, a potential problem with estimating the linear
model in Equation (3) is the high number of counties with no solar installations for a given
year. Thirty seven percent (37%) of the county-year observations in the data have zero solar
capacity additions, and 17% of counties have no solar installations throughout the study
period. This implies that 37% of linear combinations of explanatory variables included in
the model are associated with a zero value in the dependent variable. An empirical difficulty
arises because a least squares model will falsely impose a linear association to this data,
i.e., combinations of the right hand side variables can result in negative predictions of solar
adoption. Thus, parameter estimates obtained from a linear estimator are likely to be biased
and inconsistent.
We apply a maximum-likelihood Tobit estimator to account for a non-linearity due to the
high proportion of corner solutions in the data. The Tobit model is typically referred to as a
censored regression model. However, Wooldridge (2010) makes a clear distinction between a
censored regression and a corner solution application of the Tobit estimator. It important to
make this distinction for the purposes of this paper. In a censored regression context there
are observations in the sample which are artificially censored at some arbitrary point. This
is distinct from a corner solution context because the dependent variable is fully observed.
Zero solar PV capacity addition in a county simply reflects that no new residential solar PV
systems were installed in the given year. Household solar PV capacity additions are strictly
positive with a probability mass point at zero. When individual household decisions are
aggregated to the county level the lower limit remains at zero.
3.1.1 Unobserved county heterogeneity as a random effect
A random-effects error component is added to the standard Tobit estimator in order to
incorporate unobserved heterogeneity by county. The error term is re-written as an aggregate
error,
µit=νi+it (4)
whereνi is the random parameter which represents unobserved county-specific effects, under
assumptions: νi|Xit ∼ N(0, σ2ν), it|Xit ∼ N(0, σ2). A formal test for random effects is
performed using the Breusch-Pagan (BP) Lagrange Multiplier test. The null hypothesis
of no random effect is rejected. However, the RE Tobit estimator relies on the restrictive
assumption that the error term is strictly exogenous, i.e., cov(νi, Xit) = cov(νi, Xi,t−1) =
0. Violation of this assumption casts doubt on the RE Tobit results. A Hausman test
between a linear fixed-effects and random-effects model shows that the above assumption
is violated. Despite this limitation, the results from an RE Tobit estimator are useful for
making comparisons across different estimators.
3.1.2 Instrumental variable method
A potential source of endogeneity is the SREC incentive. SRECs provide a financial incentive
to drive household demand for PV systems. Meanwhile, each newly installed PV array
generates solar energy, which increases the supply of SRECs and puts downward pressure on
SREC prices. This source of endogeneity can cause the parameter estimates of SREC price
effects as well as the other included explanatory variables to be biased. An instrumental
variable (IV) Tobit estimator is applied in an attempt to mitigate the endogeneity bias.
Three instruments are used for SREC prices: the solar alternative compliance payment
(SACP), a one-period lag of SREC prices, and a binary variable indicating the presence of
a solar carve-out. A Smith-Blundell test of exogeneity confirms that the right hand side
4
Data
A panel of 300 counties over 8 years (2005 to 2012) was constructed from several data
sources. This section outlines the various data sources. The final dataset consists of 2014
observations.
4.1
Solar Adoption
Data on residential solar PV installations were obtained from the National Renewable Energy
Lab (NREL) Open PV Project. The Open PV Project is a public database which provides
location and capacity rating information for all solar PV projects in the United States. We
obtain data from 2005-2012 for the following states: ME, NH, VT, MA, NY, RI, CT, NJ,
PA, DE, MD, WV, and the District of Columbia. Following Barbose, Darghouth, and Wiser
(2012) and Kwan (2012), we use an upper limit of 10 kW capacity to denote residential PV
systems. Geographic Information System (GIS) software was used to spacially reference and
aggregate individual residential PV systems to the county level. County boundary shapefiles
were obtained from the U.S. Census Bereau Tiger/Line shapefile database. Figure 1 provides
Figure 1: Total residential PV capacity (kW), year-end 2012 (1024.30,11122.36] (248.98,1024.30] (113.13,248.98] (27.45,113.13] (0,27.452] [0,0]
4.2
Electricity price and solar resource
Yearly state average electricity prices were obtained from the U.S. Energy Information
Ad-ministration. Residential electricity prices are averaged yearly for the total electric industry
by state, measured in cents per kilowatt hour (¢/kWh). Electricity prices are not indexed or adjusted for inflation.
The average annual solar resource was calculated using data developed by the State
University of New York/Albany satellite radiation model. Solar resource, i.e., insolation is
measured in kilowatt-hours per square meter per day (kWh/m2/day). Ten-kilometer grid
cells of annual average daily total solar resource were spatially referenced and aggregated to
4.3
State incentives for solar PV
State policy incentive variables were constructed using the Database of State Incentives for
Renewables and Efficiency (DSIRE 2013). State sales taxes are used to capture heterogeneity
of sales tax incentive policies. A homeowner receiving a sales tax exemption in a state with
a high sales tax effectively receives a larger incentive than an identical homeowner in a
neighboring state with a lower sales tax. Personal tax credit and property tax exemptions
are included as binary variables, which take on a value of one if the county is located in
a state with these incentives. Rebate and loan program availability are also measured as
binary variables. Following Menz and Vachon (2006), we also include an RPS trend variable
to capture the number of years a state RPS has been in effect.
4.3.1 SREC prices
Solar renewable energy credit market prices were obtained from www.SRECtrade.com. One
SREC is equivalent to one megawatt hour (MWh) of electricity generation. The market price
of SRECs is averaged for each year in the study period. All counties within the same state
are subject to the state market price of SRECs. There are some SREC markets which can
overlap state lines. For example, the Pennsylvania SREC market accepts energy generated
by systems located in Ohio. Solar alternative compliance payment (SACP) in dollar amount
is used as an instrument for SREC prices. The SACP is a per kWh fine levied on utilities
for shortfalls in meeting RPS solar carve-out requirements. Unlike SREC prices which are
determined in the market, the SACP is set by legislature years in advance.
4.4
Indicators of pro-environmental preferences
Three variables are used to capture preferences for positive environmental action. The
percentage of county residents who buy organic food was obtained from Geographic Research
Inc. (2011). Presidential election results from 2008, obtained from National Atlas of the
county in the 2008 presidential election. Yearly hybrid vehicle registrations for personal
transportation by county were obtained from the automotive consulting firm R. L. Polk.
4.5
Demographics and geographic characteristics
County-level, yearly data from 2005-2012 for median age, median household income, median
home value, number of owner-occupied housing units and total population were obtained
from American Factfinder, a web service of the U.S. Census Bureau’s American Community
Survey.
Table 2 presents summary statistics. Column 1 is the variable name with units. Home
density, population density and median home values are used in natural log form to correct
for highly skewed distributions. Column 2 indicates the level of variation for each variable,
where “C” is for county, “S” is for state, and “Y” is for year.
5
Results and Discussion
This section presents marginal effects obtained from the empirical methods discussed above.
Table 3 compares three specifications of residential PV demand. The displayed coefficients
are average marginal effects on the uncensored dependent variable obtained from the Tobit
estimator.
Each subsequent specification represents an improvement to the model fit. Both the
log-likelihood and pseudo-R2 increase from model (1) to model (3). Likelihood ratio tests are
performed to formally confirm that each subsequent specification represents a statistically
significant increase in explanatory power.
The addition of state policy effects and environmental preferences reveals significant
cor-rections of omitted variable bias. Omitted variable bias arises when a significant explanatory
variable is omitted from the specification. Corrections of omitted variable bias are evident
Com-Table 2: Summary statistics
Varies by Mean Std. Dev. Min. Max. Dependent variable
PV capacity additions C, Y 105.891 357.184 0 7467.045
Independent variables
Solar resource (kWh/m2/day) C 4.491 0.199 4.085 5.007
ln(Owner-occupied homes) C, Y 10.381 1.147 7.34 12.908
ln(Population density) C, Y 5.43 1.571 1.035 11.18
ln(Median home value) C, Y 12.033 0.591 10.325 13.816
Electricity price (¢/kWh) S, Y 13.9 3.315 6.21 20.33
Bachelor degree or higher (%) C, Y 29.769 12.542 3.925 83.252
Median Age C, Y 40.203 3.08 27.3 51
SREC Price ($) S, Y 72.397 154.789 0 664.5
Sales tax exemption (%) S, Y 3.641 3.75 0 8.48
RPS trend S, Y 4.872 3.546 0 15
Rebate (d) S, Y 0.457 0.498 0 1
Income tax credit (d) S, Y 0.349 0.477 0 1
Loan program (d) S, Y 0.182 0.386 0 1
Solar rights (d) S, Y 0.196 0.397 0 1
Hybrid vehicle sales (%) C, Y 1.932 1.071 0 10.986
Democrats (%) C 51.679 11.92 23.64 92.5
Instruments
SACP S, Y 176.47 246.75 0 711
Carve-out (d) S, Y .368 .482 0 1
(d) for discrete change of dummy variable from 0 to 1
†Natural log transformed values
*Population density calculated by Total populationLand area , with county land area measured in square miles (mi2).
Table 3: Three specifications of residential PV demand
(1) (2) (3)
∆(System count)t−1 0.0118∗∗∗ 0.00873∗∗∗ 0.00858∗∗∗
(4.77) (3.60) (3.71) Solar resource (kWh/m2/day) 1.243+ 2.146∗∗∗ 1.872∗∗
(1.87) (3.32) (2.94) ln(Owner-occupied homes) 0.760∗∗∗ 0.905∗∗∗ 1.020∗∗∗
(3.66) (4.26) (4.66) ln(Population density) -0.428∗∗ -0.601∗∗∗ -0.678∗∗∗
(-2.84) (-4.77) (-4.89) ln(Median home value) 1.460∗∗∗ 0.931∗∗ 1.113∗∗∗
(4.34) (2.93) (3.52) Electricity price (¢/kWh) 0.431∗∗∗ 0.246∗∗∗ 0.182∗∗∗
(14.99) (5.03) (3.42) Bachelor degree or higher (%) 0.00785 0.0221+ -0.00379
(0.56) (1.88) (-0.27) Median Age -0.110∗∗∗ -0.0894∗∗ -0.0750∗∗ (-3.35) (-2.90) (-2.56) Time trend 0.359∗∗∗ 0.208∗∗∗ 0.195∗∗∗ (10.11) (3.59) (3.45) SREC Price ($) 0.00105∗∗ 0.00104∗∗ (2.38) (2.46) Sales tax exemption (%) 0.342∗∗∗ 0.374∗∗∗
(7.01) (7.21)
RPS trend -0.0566 -0.0510
(-1.49) (-1.32)
Rebate (d) 1.224∗∗∗ 1.111∗∗∗
(7.38) (6.54)
Income tax credit (d) -0.516+ -0.432+
(-1.92) (-1.65)
Loan program (d) 1.539∗∗∗ 1.782∗∗∗
(5.54) (6.51)
Solar rights (d) 0.262 0.129
(1.34) (0.68)
Hybrid vehicle sales (%) 0.276∗∗
(2.96) Democrats (%) 0.0179∗∗ (2.16) Pseudo-R2 0.201 0.250 0.254 Log-likelihood -3404.1 -3195.2 -3178.1 t statistics in parentheses + p <0.1,∗∗ p <0.05,∗∗∗ p <0.001
pared to (1), the effect of lagged system count decreases in (2) and (3), although it remains
significant. The effect of solar resources (insolation) nearly doubles in magnitude and gains
statistical significance when policy variables and pro-environmental preference variables are
introduced in models (2) and (3). The remainder of this paper will focus on results based
on the full specification in model (3).
5.1
Comparing results across estimators
This section compares results across four estimators: a pooled Tobit, a random-effects (RE)
Tobit, an instrumental variable (IV) Tobit, and a pooled ordinary least squares (OLS)
esti-mator. As discussed above, results obtained from the random effects (RE) Tobit estimator
as well as the OLS approach are problematic because of restrictive assumptions about the
error term and non-linearity of PV demand, respectively. Nevertheless, the results from both
estimators are displayed for comparison.
The estimates displayed in Table 4 are marginal effects under the Tobit estimators and
linear β coefficients under the OLS estimator.
Several explanatory variables are robust to model specification. Lagged system count
is significant under all but the RE Tobit estimator. This result points to the existence of
a “snowball effect” in residential solar technology adoption. It can be inferred that social
interactions and solar panel visibility on neighbors’ rooftops may lead to a reduction in
perceived uncertainty about solar technology.
The number of owner-occupied households in a county has a positive effect and is robust
at the 1% significance level. This is primarily a control variable which is used to make
comparisons across counties of different sizes. It’s significance speaks to the importance of
controlling for available roof space. The time trend is also positive and robust to specification.
Time effects are assumed to capture changes to federal policies and to solar panel installation
costs over time. Costs vary over time due to falling solar panel prices as well as decreasing
Table 4: Comparing results across estimators
(1) (2) (3) (4)
Tobit RE Tobit IV Tobit OLS ∆(System count)t−1 0.00858∗∗∗ 0.00222 0.00935∗∗∗ 0.0120∗∗∗
(3.71) (1.09) (3.87) (4.56) Solar resource (kWh/m2/day) 1.872∗∗ 1.306+ 1.840∗∗ 1.275∗∗
(2.94) (1.67) (2.82) (2.65) ln(Owner-occupied homes) 1.020∗∗∗ 1.196∗∗∗ 0.976∗∗∗ 0.839∗∗∗
(4.66) (5.16) (4.49) (4.90) ln(Population density) -0.678∗∗∗ -0.819∗∗∗ -0.659∗∗∗ -0.465∗∗∗
(-4.89) (-5.08) (-4.75) (-4.21) ln(Median home value) 1.113∗∗∗ 1.832∗∗∗ 1.124∗∗∗ 0.672∗∗ (3.52) (5.65) (3.50) (2.74) Electricity price (¢/kWh) 0.182∗∗∗ 0.0930+ 0.211∗∗∗ 0.0549 (3.42) (1.92) (3.68) (1.34) Bachelor degree or higher (%) -0.00379 0.000601 -0.00492 0.000368
(-0.27) (0.04) (-0.35) (0.03) Median Age -0.0750∗∗ -0.0588∗∗ -0.0737∗∗ -0.0454+
(-2.56) (-2.22) (-2.44) (-1.81)
Time trend 0.195∗∗∗ 0.332∗∗∗ 0.167∗∗ 0.226∗∗∗
(3.45) (6.26) (2.85) (5.65) Hybrid vehicle sales (%) 0.276∗∗ 0.0965 0.290∗∗ 0.211∗∗
(2.96) (1.05) (2.98) (2.75) Democrats (%) 0.0179∗∗ 0.0356∗∗ 0.0181∗∗ 0.0109 (2.16) (3.00) (2.08) (1.62) SREC Price ($) 0.00104∗∗ 0.000383 0.00154∗∗ 0.00197∗∗∗
(2.46) (0.90) (2.05) (4.97) Sales tax exemption (%) 0.374∗∗∗ 0.323∗∗∗ 0.360∗∗∗ 0.344∗∗∗
(7.21) (6.29) (6.67) (7.76)
RPS trend -0.0510 -0.0627 -0.0490 -0.0364
(-1.32) (-1.37) (-1.24) (-1.17) Rebate (d) 1.111∗∗∗ 0.890∗∗∗ 1.180∗∗∗ 0.692∗∗∗
(6.54) (5.83) (6.57) (5.18) Income tax credit (d) -0.432+ 0.185 -0.480+ -0.605∗∗
(-1.65) (0.64) (-1.69) (-2.31) Loan program (d) 1.782∗∗∗ 1.994∗∗∗ 1.630∗∗∗ 0.771∗∗∗ (6.51) (6.83) (4.68) (3.53) Solar rights (d) 0.129 -0.568∗∗∗ 0.0708 0.134 (0.68) (-3.63) (0.36) (0.78) Constant -37.14∗∗∗ -46.55∗∗∗ -36.10∗∗∗ -20.74∗∗∗ (-6.41) (-9.29) (-6.14) (-5.01) Observations 2014 2014 1911 2014 t statistics in parentheses + p <0.1,∗∗ p <0.05,∗∗∗ p <0.001
a solar PV system, including the cost of mounting equipment, interconnection and labor.
As the industry develops, experience and competition among installers drives BOS costs
downward. Consumers’ growing familiarity with the technology, through word-of-mouth
and solar energy marketing, may also be driving the positive time effect.
The negative parameter estimate for population density, robust across specifications at
the 1% significance level, suggests an inverse relationship between urbanization and
resi-dential solar PV adoption. This result is not consistent with Zahran et al. (2008), where
empirical evidence pointed to a positive relationship between urbanization and solar hot
wa-ter adoption. However, the urbanization measure used by Zahran et al. (2008) is a dummy
variable for Census-designated urban areas which may include sub-urban environments with
relatively low population density when compared with metropolitan city areas.
Of the variables measuring differences in demographics, median home value and median
age are fairly robust across specifications. The strong positive result for median home values
speaks to the importance of income status as predictor of residential PV demand. Higher
income homeowners have more disposable income to invest in a solar PV system. Estimated
effects of educational attainment, on the other hand, are weak and sensitive to model
specifi-cation. Median age has a negative effect which is significant at 5% and 10% levels under the
Tobit and linear OLS estimators, respectively. The negative effect is somewhat unexpected;
it is possible that younger homeowners are more willing to invest in solar PV because they
expect to stay in their homes for many years and are thus able to reap long-term benefits
from their solar PV system.
The state policy variables are very strong predictors of residential solar PV adoption.
Parameter estimates for sales tax exemptions, rebates, and loan programs are robust across
estimators at the 1% significance level. SREC prices are significant at least at the 5% level
under all but the RE Tobit. Each of these state incentives impacts the financial decision of
residential PV adoption because they each decrease the cost of installing a solar PV system.
markets decrease the amount of time it takes for the PV investment to pay for itself.
Unexpectedly, the income tax credit effect is negative and significant at 10% under all
but the RE Tobit estimator. There are two possible reasons why income tax credits fail
to boost residential PV demand. Firstly, the income tax credit incentive may be too small
to significantly impact the adoption decision. This is because the state income tax credit
amount is calculated based the net expenditure of an installation. The net expenditure is
defined as the expense less the 30% federal tax credit and any other state financial incentives.
Secondly, the income tax credit does not reduce the out-of-pocket expense of a residential
installation because homeowners must wait to claim the credit during tax season. The lack
of an immediate price reduction may be rendering the income tax credit incentive ineffective.
The RPS trend effect is consistently not significant across estimators. This suggests
that a technology-blind RPS alone, holding all else constant, does not provide a significant
incentive to residential PV adoption. This is consistent with Wiser, Barbose, and Holt
(2011)’s concerns that without a solar-specific mandate, RPS programs only favor a limited
number of least-cost technologies. To our knowledge, this is the first empirical evidence of
the failure of a technology-neutral RPS to promote diversity in renewable technology.
Demand for hybrid vehicles and political party affiliation are found to be significantly
related to residential solar PV demand. Voting rates for the democratic party are
signifi-cant at 5% across specifications. This is strong evidence of a relationship between political
party affiliation and pro-environmental preferences, and suggests that Democratic-leaning
constituencies, holding all else constant, are more likely to invest in solar PV technology.
The estimated effect of hybrid vehicles sales is positive and significant at 5% under all but
the RE Tobit estimator. This is further empirical evidence of findings by Dastrup et al.
(2012) which point to a positive relationship between these two “green status” consumer
5.2
Marginal effects
Marginal effects are evaluated based on results obtained from the instrumental variable (IV)
Tobit estimator. The IV Tobit is the preferred model because it simultaneously deals with
two empirical issues: 1) the non-linearity of residential PV demand; and 2) the endogeneity
of SREC prices. The linear maximum-likelihood parameter estimates, ˆβ, are also displayed
for comparison.
The following tables are divided into three groups of variables: demographics and
geog-raphy; state incentive policies; and pro-environmental preference indicators. In Table 5 the
marginal effects are evaluated at the sample means of each independent variable. Coefficient
estimates are interpreted as proportional changes to residential PV capacity, given a one
unit change in the independent variable. For variables in natural log form the coefficients
are elasticities.
The largest change in PV capacity comes about from a one-unit change in solar resource.
An additional kilowatt-hour per meter squared per day is expected to increase residential PV
capacity by 141%. This helps to explain the wide discrepancy in adoption between states
like California and New Jersey. Both states have aggressive incentives, but average solar
resources in California exceed sunshine in New Jersey by 2-3 kWh across the state. Solar
resource also has the largest per-unit effect (20.5%) on the probability of positive adoption.
The time trend indicates that with each year, holding all else constant at zero, residential
PV capacity is expected to grow by roughly 13%. This effect is driven by falling solar panel
prices and competition among a growing number of installation firms. One-cent increases
in electricity prices have a 16% impact on PV capacity. An additional system installed in
the previous year is expected to increase capacity by about 7/10 %. A county with ten new
systems installed in the current year could expect new capacity to increase by 7% the next
year, holding all else constant.
Residential PV demand is fairly inelastic to differences in the number of owner-occupied
Table 5: Marginal effects at means: demographics and geography
Unit change At (1) (2) (3) (4)
in Mean βˆ ∂E[Y]/∂X ∂E[Y|Y >0]/∂X ∂P(Y >0)/∂X
∆(System count)t−1 13.276 0.0107∗∗∗ 0.00935∗∗∗ 0.00718∗∗∗ 0.00104∗∗∗
(3.81) (3.87) (3.91) (3.45)
Solar resource (kWh/m2/day) 4.495 2.109∗∗ 1.840∗∗ 1.415∗∗ 0.205∗∗
(2.83) (2.82) (2.81) (2.83)
Electricity price (¢/kWh) 13.985 0.242∗∗∗ 0.211∗∗∗ 0.162∗∗∗ 0.0235∗∗∗
(3.65) (3.68) (3.70) (3.42)
Bachelor degree or higher (%) 30.194 -0.00563 -0.00492 -0.00378 -0.000548
(-0.35) (-0.35) (-0.35) (-0.35)
Median Age 40.177 -0.0845∗∗ -0.0737∗∗ -0.0567∗∗ -0.00821∗∗
(-2.43) (-2.44) (-2.46) (-2.31)
Time trend 5.045 0.191∗∗ 0.167∗∗ 0.128∗∗ 0.0185∗∗
(2.87) (2.85) (2.82) (2.95)
Percent change in Elasticities
ln(Owner-occupied homes) 10.435 1.119∗∗∗ 0.976∗∗∗ 0.751∗∗∗ 0.109∗∗∗
(4.55) (4.49) (4.43) (4.62)
ln(Population density) 5.495 -0.756∗∗∗ -0.659∗∗∗ -0.507∗∗∗ -0.0734∗∗∗
(-4.80) (-4.75) (-4.70) (-4.75)
ln(Median home value) 12.053 1.287∗∗∗ 1.124∗∗∗ 0.864∗∗∗ 0.125∗∗∗
(3.55) (3.50) (3.46) (3.65)
t statistics in parentheses
Table 6: Marginal effects: environmental preferences
Unit change At (1) (2) (3) (4)
in Mean βˆ ∂E[Y]/∂X ∂E[Y|Y >0]/∂X ∂P(Y >0)/∂X
Hybrid vehicle sales (%) 1.943 0.332∗∗ 0.290∗∗ 0.223∗∗ 0.0323∗∗
(2.98) (2.98) (2.98) (2.92)
Democrats (%) 51.917 0.0207∗∗ 0.0181∗∗ 0.0139∗∗ 0.00201∗∗
(2.09) (2.08) (2.08) (2.09)
t statistics in parentheses
+ p <0.1,∗∗ p <0.05,∗∗∗ p <0.001
number of owner-occupied homes is expected to increase residential capacity by about one
percent. PV demand is more inelastic when it comes to population density. A one percent
increase in the population density is expected to decrease capacity by .65%, while a one
percent increase the county median home value is expected to increase capacity by 1.12%.
Next, table 6 presents the marginal effects of pro-environmental preference indicators.
The marginal effects are again evaluated at the sample means.
Hybrid vehicles sales are found to be positively related with residential PV adoption. An
increase from 2% to 3% of hybrid vehicle sales is associated with a roughly 30% increase in
residential capacity. Political party affiliation has a statistically significant effect, but small
in magnitude. If voting rates for the democratic party increase from 52% to 53%, residential
capacity is expected to increase by about 2%.
5.3
Policy impacts
Table 7 presents the per-unit change effects of SREC prices and sales tax exemptions. Sample
means were not used as evaluation points because the averages are skewed due to an absence
of these incentives from several states. Wherever the incentive is not available these variables
take on a value of zero, causing sample means to be far from representative of real world
values. Instead, the effects are evaluated at three representative points of each variable.
SREC prices are evaluated at the mean of observed positive values ($280), and one standard deviations away in each direction ($95, $465). Sales tax exemptions are evaluated at 6% and
Table 7: Marginal effects: SREC prices and sales taxes Unit change (1) (2) (3) (4) in βˆ ∂E[Y]/∂X ∂E[Y|Y >0]/∂X ∂P(Y >0)/∂X SREC Price ($) at$95 .00176∗∗ 0.00130∗∗ 0.00115∗∗ 0.000125∗∗ (2.04) (2.03) (2.03) (2.01) at$280 .00176∗∗ 0.00134∗∗ 0.00120+ 0.000119∗∗ (2.04) (1.97) (1.96) (2.13) at$465 .00176∗∗ 0.00138+ 0.00124+ 0.000112∗∗ (2.04) (1.92) (1.90) (2.26)
Sales tax exemption (%)
at 6% 0.412∗∗∗ 0.347∗∗∗ 0.304∗∗∗ 0.0292∗∗∗ (6.81) (5.98) (5.98) (7.57) at 8% 0.412∗∗∗ 0.368∗∗∗ 0.330∗∗∗ 0.0228∗∗∗ (6.81) (5.83) (5.69) (13.35) t statistics in parentheses + p <0.1,∗∗ p <0.05,∗∗∗ p <0.001
8%, which are the sales tax rates in Vermont and New York, respectively.
The SREC price effect is significant, although the coefficient is not large. The small effect
may be due to the significant variability in SREC prices since the opening of SREC markets.
A one hundred dollar increase from $95 is estimated to give a 13% boost to residential PV adoption. Negative shifts in SREC markets can likewise negatively impact demand.
These results support anecdotal evidence from Pennsylvania that plummeting SREC prices
slackened demand for new PV installations. Large drop-offs in the SREC incentive may
cause homeowners to question the reliability of these markets for providing a production
incentive throughout the lifespan of the PV system. Future solar carve-out policies may
need to consider price support mechanisms to avoid price shocks and mitigate uncertainty
about SREC markets. Where a sales tax exemption is available, an increase in the sales tax
rate from 6% to 7% is expected to increase residential PV adoption by roughly 35%. States
with higher sales taxes will see somewhat greater gains to adoption as a result of the sales
tax exemption policy.
Next, table 8 presents the marginal effects due to the availability of policies represented
Table 8: Marginal effects: dummy policy variables
Discrete change (1) (2) (3) (4)
from 0 to 1 βˆ ∂E[Y]/∂X ∂E[Y|Y >0]/∂X ∂P(Y >0)/∂X Rebate (d) 1.351∗∗∗ 0.997∗∗∗ 0.873∗∗∗ 0.108∗∗∗
(6.37) (6.51) (6.58) (5.85)
Income tax credit (d) -0.555+ -0.401+ -0.357+ -0.0387+
(-1.68) (-1.70) (-1.69) (-1.70) Loan program (d) 1.785∗∗∗ 1.369∗∗∗ 1.228∗∗∗ 0.125∗∗∗ (4.77) (4.69) (4.61) (4.95) Solar rights (d) 0.0809 0.0595 0.0527 0.00572 (0.36) (0.36) (0.36) (0.36) t statistics in parentheses + p <0.1,∗∗ p <0.05,∗∗∗ p <0.001
in residential PV capacity due to a discrete change from zero (no policy) to one (policy exists).
Loan programs have the largest impact on residential PV capacity. Availability of a
low-interest solar loan is expected to increase capacity by 137%, holding all else constant.
Rebates also have a strong positive effect, with availability of a rebate expected to increase
capacity by roughly 100%. Income tax credit has an unexpected sign (for reasons discussed
above) and is weakly significant. Solar rights regulations are found to have no significant
effect on residential PV demand.
6
Conclusion
This paper examines the effect of state policy incentives on residential solar PV capacity
growth. It is possible that solar adoption is mainly being driven by pro- environmental
preferences and altruism. In this case, solar policies are inefficient because adopters would
have installed solar PV absent policy incentives. Previous literature has not explored state
policy effects on growth of residential PV capacity, while controlling for demographic and
geographic characteristics, and pro-environmental preferences.
Our empirical results show that policy incentives are significant drivers of solar PV
ca-pacity growth. Solar-specific policies such as the rebate, loan program, sales tax exemption,
not significant. This result suggests that for PV capacity to grow, policies directly targeted
at solar are needed. Our findings also suggest that policies that reduce the up-front cost
of solar PV installation have the most impact. Of solar-specific policies, the coefficient of
loan program availability has the greatest magnitude, suggesting that credit constraint is a
deterrent to adoption and the availability of financing is just as important as (if not more
so than) direct financial incentives.
Similar to previous studies, we find evidence that demographic characteristics such as
lower median age and income are correlated with PV adoption. The lagged number of solar
installations is also significant suggesting that familiarity with solar PV technology and
in-fluence of adopters within one’s network positively inin-fluence solar PV adoption. In addition,
pro-environmental preferences as indicated by political party affiliation and ownership of
hybrid vehicles are positively related to growth in PV capacity.
We also find evidence that non-policy variables that affect the financial viability of a
solar PV system such as solar insolation and electricity prices are significant, underscoring
the importance of financial considerations in a household’s adoption decision.
The focus of this paper is measuring the impact of solar incentive policies. These policies
are certain to have welfare consequences. While we do not consider the efficiency of solar
PV policies, the magnitude of policy impacts estimated in this study could be used as an
References
Adelaja, A., Y.G. Hailu, C.H. McKeown, and A.T. Tekle. 2010. “Effects of Renewable Energy
Policies on Wind Industry Development in the US.”
Andreoni, J. 1990. “Impure altruism and donations to public goods: a theory of warm-glow
giving.”The Economic Journal, pp. 464–477.
Barbose, G., N. Darghouth, and R. Wiser. 2012. “Tracking the Sun V An Historical
Sum-mary of the Installed Price of Photovoltaics in the United States from 1998 to 2011.”
Environmental Energy Technologies Division, Lawrence Berkeley National Laboratory.
Bollinger, B., and K. Gillingham. 2012. “Peer effects in the diffusion of solar photovoltaic
panels.”Marketing Science 31:900–912.
Carley, S. 2009. “State renewable energy electricity policies: An empirical evaluation of
effectiveness.”Energy Policy 37:3071–3081.
Dastrup, S.R., J. Graff Zivin, D.L. Costa, and M.E. Kahn. 2012. “Understanding the Solar
Home price premium: Electricity generation and Green social status.”European Economic
Review 56:961–973.
DSIRE. 2013. “Database of State Incentives for Renewables and Efficiency.” available at
www.dsireusa.org.
Geographic Research Inc. 2011. “SimmonsLOCAL Consumer Survey Data.” Retrieved from
SimplyMap databasewww.simplymap.com.
GTM Research and Solar Energy Industries Association. 2014. “US Solar
Mar-ket Insight (TM): 2013 Year in Review (Executive Summary).” Retrieved from
http://www.seia.org/research-resources/solar-market-insight-report-2013-year-review.
Hausman, J.A. 1979. “Individual Discount Rates and the Purchase and Utilization of
Kwan, C.L. 2012. “Influence of local environmental, social, economic and political variables
on the spatial distribution of residential solar PV arrays across the United States.”Energy
Policy 47:332–344.
Labay, D.G., and T.C. Kinnear. 1981. “Exploring the consumer decision process in the
adoption of solar energy systems.” Journal of Consumer Research, pp. 271–278.
Lopez, A., B. Roberts, D. Heimiller, N. Blair, and G. Porro. 2012. “US Renewable Energy
Technical Potentials: A GIS-Based Analysis.”Contract 303:275–3000.
Lutzenhiser, L. 1993. “Social and behavioral aspects of energy use.”Annual Review of Energy
and the Environment 18:247–289.
Maguire, K. 2011. “What’s Powering Wind? The Role of Prices and Policies in Determining
the Amount of Wind Energy Development in the United States (1994-2008).” In 2011
Annual Meeting, July 24-26, 2011, Pittsburgh, Pennsylvania. Agricultural and Applied
Economics Association, 103599.
Menz, F.C., and S. Vachon. 2006. “The effectiveness of different policy regimes for promoting
wind power: Experiences from the states.” Energy Policy 34:1786–1796.
National Atlas of the United States. 2009. “2008 Presidential General Election, County
Results.” available at www.nationatlas.gov.
Noll, D., C. Dawes, and V. Rai. 2014. “Solar Community Organizations and active peer
effects in the adoption of residential PV.” Energy Policy 67:330–343.
NREL. 2013. “National Renewable Energy Laboratory.” available at openpv.nrel.gov.
Sexton, S.E., and A.L. Sexton. 2014. “Conspicuous conservation: The Prius halo and
will-ingness to pay for environmental bona fides.” Journal of Environmental Economics and
Sidiras, D.K., and E.G. Koukios. 2004. “Solar systems diffusion in local markets.” Energy
Policy 32:2007–2018.
Smith, D. 2013. “IBISWorld Industry Report 22111e Solar Power in the US.”
U.S. Energy Information Adiminstration (EIA). 2011. “Direct Federal Financial
Interven-tions and Subsidies in Energy in Fiscal Year 2010.”
—. 2014. “Electric Power Monthly with Data for January 2014.” avaliable at www.eia.gov.
Wiser, R., G. Barbose, and E. Holt. 2011. “Supporting solar power in renewables portfolio
standards: Experience from the United States.” Energy Policy 39:3894–3905.
Wooldridge, J.M. 2010. Econometric Analysis of Cross Section and Panel Data.
Yin, H., and N. Powers. 2010. “Do state renewable portfolio standards promote in-state
renewable generation.” Energy Policy 38:1140–1149.
Zahran, S., S.D. Brody, A. Vedlitz, M.G. Lacy, and C.L. Schelly. 2008. “Greening local
energy: explaining the geographic distribution of household solar energy use in the United