EFFECTIVENESS OF STATE FINANCIAL INCENTIVES AND REGULATIONS IN PROMOTING SOLAR ENERGY TECHNOLOGIES

Transcription

1 EFFECTIVENESS OF STATE FINANCIAL INCENTIVES AND REGULATIONS IN PROMOTING SOLAR ENERGY TECHNOLOGIES A Thesis Submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Master of Public Policy in The Georgetown Public Policy Institute By Oxana Petritchenko, B.S. Washington, D.C. April 15, 2012

2 Copyright 2012 by Oxana Petritchenko All Right Reserved ii

3 EFFECTIVENESS OF STATE FINANCIAL INCENTIVES AND REGULATONS IN PROMOTING SOLAR ENERGY TECHNOLOGIES Oxana Y. Petritchenko, B.S. Thesis Advisor: Shaun D. Ledgerwood, J.D., Ph.D. ABSTRACT This thesis evaluates whether state-level solar energy incentives and regulations promote the market growth of small-scale photovoltaic (PV) technologies in commercial and residential sectors. The results of this analysis are important for effective design and implementation of policies intended to achieve the increased deployment of small-scale solar energy technologies. A fixed effects model is used to assess effectiveness of solar policies while controlling for unobservable time-invariant factors that affect the deployment of PV system installations unique to each state. In particular, the capacity of PV technologies installed per 100,000 people during a period of 2000 through 2010 across twenty-four states and the District of Columbia was regressed against a variety of state policies directed at promoting PV technologies. Specifically, feed-in tariffs, rebates, solar and distributed generation (DG) carve-outs in states renewable portfolio standards, solar renewable energy credits (SREC), net metering as well as sales, residential, commercial and personal tax exemptions were evaluated. The results of this analysis support claims that feed-in tariffs are significantly more effective than any other policy in promoting the deployment of solar energy technologies. Establishment of a SREC market and solar or DG carve-outs are also found to be effective in promoting the growth of PV installations. Although rebates are found to be ineffective during the first year of their implementation, observing the policies effect several years later reveals that they become significant with time. However, net metering and tax incentives are found to be ineffective in promoting deployment of PV technologies regardless of taking into account the effect of the time lag of policy implementation. iii

4 Thank you for advice and assistance to my thesis advisor Shaun Ledgerwood, Georgetown Public Policy Institute s faculty and staff, and my parents for support that made the research and writing of this thesis possible. Sincerely, Oxana Petritchenko iv

5 Table of Contents I. Introduction 1 II. LiteratureReview 3 III. DescriptiveStatistics 7 A. PVInstallationsData 7 B. Policies 10 IV. ModelSpecifications 13 A. StateandYearFixedEffectsModels 13 B. Model1 14 C. Correlations 16 V. ModelRefinement 18 A. Model2 18 B. Models3and4 19 C. Models5,6,and7 20 VI. AnalysisofResultsandPolicyImplications 21 A. Feed intariffs 21 B. SolarRenewableEnergyCredits 23 C. SolarRPS 23 D. Rebates 24 E. NetMetering 25 F. ElectricityandNaturalGasPrices 25 G. GDPpercapita 26 H. RPSandTaxExemptions 26 VII.SensitivityAnalysis 27 VIII:Conclusion 29 References 30 v

6 I. Introduction Renewable energy is the future of clean, reliable, secure and competitive energy supply. Governments of many developed and developing nations have established national goals or have shown support for moving toward greater generation of energy from alternative resources. Although the United States does not have a federal goal for generating a specified percentage of electricity from renewable resources nationally, President Obama proposed in his State of the Union 2011 address to achieve a goal of generating 80 percent of U.S. electricity from renewable energy sources by This is an ambitious goal, and it may seem unrealistic. However, it is only unrealistic if the tools and methods to achieve that goal do not exist. There are a number of tools at the disposal of the U.S. government to achieve the renewable objectives it establishes. Specifically, there are financial incentives available for targeting deployment of renewable energy technologies, such as rebates, feed-in tariffs (FIT), personal and corporate income tax exemptions as well as sales tax and property tax exemptions. There are also regulations, such as net metering rules, Renewable Portfolio Standards (RPS), establishment of solar or distributed generation (DG) carve outs within the portfolio standards or implementation of a Solar Renewable Energy Credit (SREC) market. This list is not exhaustive as there are other policies available such as government loans at low interest rates, public benefit fund, required green power options, or interconnection standards. The financial incentives and regulations studied in this thesis are some of the most popular and promising policy tools in the European Union and in the United States that are directed at promoting independent generation of solar electricity at a small-scale residential or business level. FITs are of particular interest because they are widely used throughout the European Union (EU) and are often 1 Remarks by the President in State of Union Address. The White House. Office of the Press Secretary. January 25,

7 claimed to be the most effective policy for increasing the widespread use of renewable energy technologies known to date. However, although many academic papers studying the EU claim that FITs are the most effective policy, there have been no empirical studies contrasting this policy with others in the U.S. A reason for this may be that state governments have only recently experimented with implementing FITs on a statewide basis, previously leaving insufficient data for analysis because the success of policy outcomes can only be accurately measured with the progression of time. Lack of data may drive discordant empirical results. For example, many researchers studying renewable energy policies in the U.S. claim that RPS policies, which are essentially a minimum quantity mandate, are the most effective policy at promoting renewable energy generation, while others claim it is not effective at all. Although there have been a number of case studies and empirical studies in the field of renewable energy policies, there is no comprehensive statistical analysis of a combination of various regulations and financial incentives across many states or countries. Statistical analysis that controls for bias and institutional, political and environmental factors that may affect results of policy effectiveness is necessary to obtain accurate conclusions on which government tools are best at promoting the deployment of renewable energy technologies. By comparison, recent claims of policy effectiveness are often based on case-by-case studies of individual policies in individual states or countries, which do not control for political, institutional or environmental concerns that can significantly affect policy outcomes. The following analysis seeks to contribute statistically supported results to the existing literature concerning policy effectiveness in the renewable energy field. It accomplishes this by providing a statistically robust approach to compare combinations of widely different policies implemented across a large number of states over a period of ten years, while controlling for political and environmental factors that can bias the measurement of policy outcomes. 2

8 II. Literature Review There are several recent empirical studies of state policy effectiveness in promoting renewable energy generation, specifically studying the effectiveness of RPS or policies promoting wind energy generation. However, there have been few empirical studies on the effectiveness of state financial incentives in promoting solar energy capacity in particular. While a range of the literature available involves case studies claiming FITs to be the most effective policy, the only comprehensive statistical analysis to support those claims was performed regarding FIT policies in the European Union. Groba et al. (2011) employ panel data to assess the effectiveness of FIT policies in promoting solar PV and onshore wind power developments across the EU. This is the first research study that uses a fixedeffects model to control for country-specific characteristics, accounting for policy design and market features, to determine that FIT policies have been effective in Europe since The study concluded that the design of each FIT policy is a more important determinant of renewable capacity development than the policy enactment alone. When controlling for fixed effects, the link between enactment of FIT and deployment of PV technology is statistically insignificant, and the link is only reestablished once the unique design of the FIT policy is used as an explanatory indicator. Furthermore, Groba et al. found that GDP per capita and fuel mix used for electricity generation are significant determinants of PV capacity development, where the effect of fuel mix can be positive or negative depending on the types of fuels used. However, the analysis failed to establish a relationship between tax incentives or grant investment policies and renewable energy development. This was likely a result of the fact that because each policy was classified as price-driven, the two policies were combined and represented as one binary indicator. (Groba, p. 19). However, representing a combination of two separate financial incentives to measure policy effectiveness is not a good methodology because effect of one may be weakened by the presence of the other. It is necessary to look at each policy separately to gauge whether either one of them is actually effective. 3

9 A paper by Sarzynski et al. was submitted for publication in 2011 but not yet published, and is the only literature to examine the joint effectiveness of state financial incentives and RPS in promoting deployment of solar technologies, including both photovoltaics and solar thermal. A cross-sectional timeseries approach was used to evaluate the extent to which state financial incentives systematically promoted market deployment of solar technology from The results demonstrate that while the stand-alone presence of individual financial incentives is not effective in increasing deployment of solar technologies, such policies became more effective as states began gaining more experience with their implementation. These findings suggest that design of financial incentives may play a more significant role in promoting deployment of solar technologies than a simple implementation of those policies. Some of Sarzynski s previous works provide in-depth analysis of state financial incentives, but they do not utilize statistical analysis of empirical evidence to support her conclusions. She discusses general trends using Pearson s r and ANOVA correlations of states financial incentives and other factors that affect PV generation. Simple correlations are not an effective tool for measuring policy effectiveness because they ignore other important factors and policies that need to be controlled for in order to accurately measure the impact of individual policies. However, they provide a useful preliminary analysis because these observations can be used as a guide for controlling for factors that may affect PV capacity growth beyond the possible impact of financial incentives or regulation. A number of general trends among several economic and societal factors were found to influence penetration of PV technologies: 1) States with higher average incomes tend to have more solar deployment even without incentives because more people are able to afford the technology; 2) States with higher electricity or natural gas prices tend to have more solar installations, due to a larger potential energy savings than might be obtained by installing a solar system in these states. Energy prices exhibited some of the strongest effects on solar deployment or energy conservation in previous impact studies (Lazzari, 1983); 3) States that need to import energy tend to have stronger solar deployment (Sawyer, 1984); 4

10 4) States with higher solar exposure have more solar panels, likely because it takes a smaller sized PV system to generate the same amount of energy as in locations with less solar resources; 5) The relationship between states having net-metering for a longer period and stronger PV markets is weak; and 6) States with some of the most generous solar incentives score high on a scale of citizen liberalism. In her studies of ten state incentive programs, Sarzynski finds that although income tax incentives had greater participation than state grant/rebate programs, the tax incentive programs typically spent far less on each individual participant than the cash incentives. Most of the other research evaluating solar financial incentives was performed in the 1980s and 1990s. Durham et al. (1988) used a regression analysis from a survey of households in 1983 from eight Western U.S. states to find that the availability of a state s tax incentives and higher energy prices significantly improved the likelihood of a homeowner installing solar hot water heaters. Hassett and Metcalf (1995) analyzed federal income tax returns for the residential energy conservation tax credits in The models tested the impact of effective tax price for investment (adjusted for variation across states with additional energy tax incentives), income, homeownership status, climate, unemployment rates, and the overall time trend on the probability of claiming the tax credit. The models controlled for variation through fixed effects and found that the tax credits positively and significantly impacted the likelihood of making an investment in energy conservation technology. Two research papers studied the effectiveness of Renewable Portfolio Standards (RPS) in promoting renewable energy in general. Shrimali and Kniefel (2011) use a fixed-effects regression model for panel data for 50 U.S. states during to conclude that RPS with either capacity or percentof-sales requirements, clean energy funds, and required green power options have a significant positive impact on the penetration of solar energy technologies, while voluntary RPS and state green power purchasing programs are ineffective in increasing usage of any type of renewable resource technology. 5

11 Furthermore, economic variables such as electricity price, natural gas price, per capita GDP, as well as structural variables such as league of conservation voters ratings and shares of coal-generated electricity are found to be insignificant. These findings suggest that the role of state policies is more crucial in increasing penetration of renewables than economic, social, and energy cost factors. Yin and Powers (2009) also study of the effectiveness of RPS in promoting renewables by accounting for the stringency of RPS policies among states. Using a fixed-effects model, they determine that not only do RPS policies have a significant positive effect on renewable energy deployment, but also that this effect would not be observed if RPS stringency is not accounted for. Specifically, when RPS is modeled as a binary variable, its effect is determined to be insignificant. However, when RPS is modeled as an incremental percentage requirement of the target share of renewable capacity, its effect on renewable energy generation is found to be positive and statistically significant. The authors measure incremental percentage requirement as a mandated increase in renewable generation from one year to the next in terms of the percentage of total electricity generation. (Yin and Powers, p. 1142). Fredric Menz (2004) acknowledges that most existing evaluations of states green policies at the time of his analysis were case studies that judged the effectiveness of different policy instruments in terms of their impact or their anticipated effect on stimulating investment in green electricity production. Menz s summary of findings of Gouche et al. (2002) on financial incentives in six states reveals that income tax credits were not a primary motivating factor for purchasing renewable energy systems and that caps on eligible costs or low maximum amounts for high-cost technologies may have limited their effectiveness. Rebate programs in three of the states increased the market for PV systems, but low-interest loan programs did not appear to play a large role in renewable energy purchases. Other factors that influenced renewable energy investment included the level of difficulty in connecting renewable systems to the grid; the adequacy of the distributor and installer infrastructure; and noneconomic factors such as customer awareness of environmental and energy conservation issues, interest in independence from 6

12 utilities and foreign energy sources, and general awareness of renewable energy technologies. Also, a set of complementary measures - net metering rules, low interest loans, tax credits, tax exemptions and buydowns - appear to be necessary to ensure market growth of renewable energy technologies. (Menz, p. 2406, quoting Gouche et al. (2002)) Menz concludes that among the most widely used instruments, financial incentives used in conjunction with mandatory regulations such as net-metering and RPS have been most effective in promoting renewable energy technologies. III. Descriptive Statistics A. PV Installations Data The data used as the dependent variable for the analysis was obtained from OpenPV Project, which was started in 2009 and administered by the National Renewable Energy Laboratory (NREL). This database is comprised of voluntarily contributed data on capacity of photovoltaic systems installed by each generator from state run incentive programs, utilities, installers, and the general public. It provides data on the number, size, cost and location of PV installations in each U.S. state as well as the dates when each PV system was installed. This dataset is a sample of PV capacity installed in each state, and is ostensibly a direct representation of the full number of installations taking place in each state throughout the period of 2000 through Only the twenty-four U.S. states with the largest PV capacity installed as recorded by the NREL OpenPV Project and the District of Columbia were considered in this analysis. These represent most of the PV capacity installed in the United States during the ten year period considered. The remaining 26 states were not considered to have a sufficient amount of solar capacity installed over the last ten years to present a significant contribution to the data analysis. Hence the descriptive statistics apply only to the twenty four states and Washington D.C.. Only PV installations of 50 kw capacity or less were included in the sample because this analysis only considers policies targeting small-scale PV systems of the residential and commercial sectors. 7

13 Although the NREL database does not specify whether the PV installations are used in the residential or commercial sector, their general use can be approximated according to the size of each installation. Among the states of interest during period, the sample included 119,580 PV systems of capacity of 10 kw or less, with a mean capacity of 4.6 kw per installation. These systems can be considered to represent the residential sector. For comparison, the average household in California uses about 6,500 kwh per year, and a system of 3-4 kw is estimated to be adequate to meet most of its electricity requirements. 2 Environmental Law and Policy Center estimates that in Chicago metropolitan area, an average household uses 8,640 kwh per year, and a 3.5 kw PV system would offset more than half of a household s electricity consumption. 3 Many state policies set a maximum capacity limit for PV installations eligible to receive an incentive. In many cases, legislation specified that a residential system is eligible for an incentive of up to 10 kw of installed PV capacity, whereas a commercial system is eligible for incentives for installations of up to 50 kw per system. Some state programs used 100 kw as the maximum capacity as a cut-off point for commercial systems, but most were designed to target PV installations of 50 kw or less. Hence the cut-off capacity of 50 kw was used for this analysis and can be considered as a rough estimate representative of installations by the commercial sector. In the data sample during the period of , there were 12,153 installations of 10 kw - 50 kw capacity with a mean of 17.1 kw per installation. Over the last ten years, a trend has been observed where the number of solar panels being installed in the U.S has been increasing at an increasingly higher rate since Figure 1 below shows the increase of PV installations over the past ten years, as recorded by the NREL. In general, the installations of PV systems are increasing at an increasing rate each year, with only a slight leveling off during 2008 likely due to the financial crisis. Of note, people continued to install PV systems at an 2FAQsforBuyingaSolarElectricSystem.AlternativeEnergyTechnologies. panelfaqs.php 3FAQsSolarPower.EnvironmentalLaw&PolicyCenter. 8

14 increasing rate once the crisis was stabilized throughout 2009 and Figure 1. PV Capacity Installed During Each Year in the 24 States and DC. Whereas Figure 1 represents total installations each year in all 24 states and Washington D.C., Figure 2 shows where most PV systems were installed over the last ten years. As one might expect, California has significantly more PV capacity installed than most other states combined, including that of the next four states after California New Jersey, Arizona, Pennsylvania and Massachusetts. The key benefit of using a fixed-effects model is the ability to determine policy effects despite the presence of such variances. As discussed in Section VII, a sensitivity analysis performed by excluding California from the sample did not significantly alter the conclusions as to the results or effectiveness of the various policies studied. 9

15 Figure 2. PV Capacity Installed in Each State During a Period. B. Policies A wide range of policies directed at small-scale solar electricity generation were implemented among the twenty-four states and the District of Columbia over the last ten years. To be included in the sample, the policies considered had to meet four criteria. First, photovoltaics have to be listed as one of the eligible renewable system technologies. Second, PV systems of up to 50 kw had to be within the specified minimum and maximum capacity requirement and eligible for policy coverage. Third, the policy had to be applicable to both residential and commercial applications of PV installations. Fourth, the policy had to be implemented and in effect before October Table 1 provides the dates of implementation of these policies by state. The variation of policies and implementation dates in these states allows for a useful comparison of their effectiveness in promoting solar electricity generation. 10

16 Table 1. Implementation of Regulatory and Financial Incentive Policies by State State Feed-in Tariff SREC Rebate RPS (Solar or DG + ) RPS Net Meter Prop. Tax CA 2/14/ Sales Tax NJ / /01/ Personal Tax Comm. Tax AZ /1/09 4/10/00 1/1/97 1/1/95 1/1/06 PA 06/01/06 5/18/ /30/04 MA /01/ CT 10/1/ /7/07 HI 11/17/ /1/09 7/1/09 NY 01/01/ /1/ TX NV 08/ /1/83 WI MD 1/1/08 01/01/ / /07 7/1/08 7/1/00 7/1/00 FL 03/ CO 01/ / /1/09 DE 06/01/08 01/11/ VT 9/20/09 6/17/ /08 MN 02/ /05 TN 6/30/10 OR 07/10 5/1/03 1/1/ /1/ DC 01/01/07 02/ /1/06 5/9/00 NH 2010** 7/11/ / IL /1/07 06/08 OH 01/01/ /1/ /1/10 UT 05/02 7/1/04 1/1/07 1/1/07 MI 10/1/08* 06/09 1/1/03 Total * Compliance does not begin until ** Policy is in effect, but market has not been established yet. 11

17 Net metering laws were implemented among most of the states included in this study. It is one of the most popular and one of the oldest policies directed at promoting installations of renewable energy systems, with most states having enacted the legislation in the 1990 s and some even in the 1980 s. The RPS is the second most popular policy to be implemented among the states in the sample, with 80 percent having a mandatory RPS requirement. In general, RPS programs have been implemented much more recently than net metering, with only three states Texas, Nevada and Wisconsin having established the policy before the year Not all states that have established the RPS have specified a solar or distributed generation carve out as a percent goal of total RPS. In fact, only 9 out of 20 states in this study established this provision. (Wiser (2010), p. 25). Property and sales tax exemptions are also some of the more popular and oldest policies among these states. Rebates have been implemented in 60 percent of the states in the sample, a SREC market by seven states, and RPS with solar or DG provisions by nine states. Feed-in tariffs are the newest policy directed at generation of solar electricity in the United States. FITs have been widely implemented throughout the EU and are considered to be very effective in deployment of renewable energy systems. However, only four states among the twenty-four considered in this study California, Hawaii, Vermont and Oregon implemented FIT from The state of Washington also has a statewide FIT policy in effect since 2006, but this state was not included in the sample for this analysis. Washington was excluded from the sample because it was not one of the top twenty five states with the highest deployment of PV capacity. 12

18 IV. Model Specifications A. State and Year Fixed Effects Models This study uses a state and year fixed-effects model with data for twenty-four states and the District of Columbia during the period of 2000 through The goal of the fixed effects regression analysis is to accurately assess and compare the effect of state policies in promoting demand-side market growth of photovoltaic technologies. A fixed effects model controls for unique characteristics of individual states that are not changing over time, eliminating large sources of bias. Due to the fixedeffects transformation, any independent variable that is constant over time for each state is eliminated. For example, because annual solar exposure in each state does not change significantly over time, the model controls for and eliminates this characteristic, thus allowing for a better comparison of the effects of state incentives, regulations and other control variables on the dependent variable. The year fixed effects also allow for elimination of influences external to the model that may affect all states in the sample at the same time. For example, the state of the U.S. economy and passage of federal policies will have an effect on the number of PV systems installed in each state; however, since these outside factors are predicted to affect all states equally at the same time, the model eliminates these factors, allowing only measurement of effects of each state s policies and other control variables. As a result of these attributes, a fixed-effects model is the most effective methodology of examining effectiveness of state-level policies in promoting installations of PV technologies and eliminating bias that is difficult or impossible to measure. In this analysis there are seven models presented to assess and compare the effect of various state policies. Models 1-4 assess the effect of these policies by modeling them as taking effect at the time of implementation. Models 5-7 assess policy effectiveness by taking into account possible changes in their effect through time. 13

19 For all models several policy and control variables were regressed against the PV capacity installed in kilowatts (kw) per 100,000 people in a given quarter of a year in each state. In order to obtain the final measure of a dependent variable, the PV capacity installed in each state was divided by the total population of each state during that year using population data obtained from the U.S. Census Bureau of the U.S. Department of Commerce. This allows for the dependent variable to represent PV capacity installed for a particular subset of a state s population while eliminating the effect of changes in overall electricity demand and changes in PV system installations that may result from significant population changes. The resulting dependent variable PV Capacity therefore captures the growth in PV installations for a segment of population through time. B. Model 1 The initial regression model used to measure effect of state financial incentive and regulations on PV capacity installed for state i, during a quarter of a year q, is: Model 1: PV CAPACITY i,q = β 0 + β 1 *REBATE i,q + β 2 * FEED-IN TARIFF i,q + β 3 *SOLAR/DG RPS i,q + β 4 *SREC i,q + β 5 * NET METER i,q + β 6 * RPS i,q + β 7 *PROPERTY TAX i,q + β 8 *SALES TAX i,q + β 9 *COMMERICAL TAX i,q + β 10 *PERSONAL TAX i,q + β 11 *ELEC PRICE i,q +β 12 *NG PRICE i,q + β 13 *GDPpc i,q + µ i,q + γ i,q + e i,q Where µ i,q represents state fixed effects, γ i,q represents year fixed effects, and e i,q is the error term. This equation uses ten policy variables obtained from the Database of State Incentives for Renewables & Efficiency (DSIRE) funded by the U.S. Department of Energy. REBATE is a binary variable indicating whether a state implemented a program offering a financial rebate, usually offered as a fixed dollar amount for capacity installed ($/Watt). FEED-IN TARIFF is a binary variable indicating whether a state had implemented a feed-in tariff policy, which is a long term guarantee for small renewable developers to sell power to the utility at predefined terms and conditions without contract negotiations. 14

20 SOLAR or DG RPS is a binary variable indicating whether a state had implemented a solar or distributed generation (DG) carve-out provisions within its RPS. The RPS solar requirement distinguishes solar energy form other renewable energy resources. A separate target is set as a fixed capacity or as a percentage of the total electricity that a state has to produce specifically from solar energy as a portion of its RPS. A DG carve-out also promotes solar technology because it sets a specific target for utilities to obtain electricity from a resource on-site of a customer, requiring it to be interconnected to the customer s side of the meter. SREC is a binary variable indicating whether a state had established a Solar Renewable Energy Credit (SREC) market. This market allows owning and trading of solar credits by generating facilities within a state, where 1 SREC = 1 MWh of solar electricity. This market is designed to function under two conditions: an established solar carve-out provision within the RPS; and a form of penalty for non-compliance with a state set solar RPS target, commonly known as a Solar Alternative Compliance Payment (SACP). The SACP drives the values of SRECs above other renewable energy credits, providing an incentive to buyers to pay higher prices that promote solar growth. NET METER is a binary variable indicating whether a state implemented a law which allows customers who generate their own electricity to offset their electricity use by allowing the excess electricity generated to flow back to the grid. Net metering enables customers to later use the amount of excess electricity generated earlier without having to purchase it at the utility s full retail rate. RPS is a binary variable indicating whether a state had established a Renewable Portfolio Standard program, which is a minimum mandate that requires utilities to use renewable energy or renewable energy credits to account for a certain percentage of their retail electricity sales according to a set schedule. PROPERTY TAX is a binary variable that indicates whether a state passed legislation allowing the value added to a property by a PV system to be excluded from property taxes. SALES TAX is a binary variable indicating that a state provides an exemption from or refund of the state sales tax on PV technologies. COMMERCIAL TAX is a binary variable that indicates whether a state provides tax credits, deductions and/or exemptions to corporations and businesses that install eligible renewable energy technologies. There is usually a maximum limit set for electricity generation capacity for a project to receive this incentive. 15

21 PERSONAL TAX is a binary variable indicating whether a state offers income tax credit or deduction, reducing the expense of purchasing and installing PV technologies. There is usually a maximum limit set for the amount of the credit or deduction. The original regression equation also uses three control variables, which were obtained from U.S. Energy Information Administration and the U.S. Bureau of Economic Analysis: ELEC PRICE measures the average annual price of residential and commercial electricity in cents/kwh in each state throughout the time period of This variable can affect the dependent variable because higher electricity prices make investment in PV technologies more attractive as a means to avoid buying electricity off of the grid. NG PRICE is a measure of an average cost of residential and commercial natural gas in dollars per million BTU in each state throughout the time period of This variable can affect the dependent variable because higher natural gas prices can make investment in PV technologies more cost effective. GDPpc is a measure of a real Gross Domestic Product per capita by state in 2005 dollars. This variable is considered because it may have an effect on PV installations as a result of people and businesses in states with higher growth of GDP per capita being able to afford greater investment in solar energy. C. Correlations The simple correlations between the dependent and independent variables considered for this analysis are shown in Table 2. Simple correlations between individual variables is a useful tool for preliminary analysis that reveals the strength of individual relationships of variables to one another. It also provides evidence for multicollinearity issues in the regression model. As expected, all of the state policy variables designed to promote solar energy growth have a positive relationship with the PV capacity installed. Similarly, the higher electricity and natural gas prices are, the more competitive solar energy becomes compared to other sources of energy, revealing a positive relationship with the number of solar 16

22 panel installations in each state. Table 2. Bivariate Correlation Matrix between Dependent and Independent Variables Variables PV Rebate Feedin Tariff Solar or DG RPS SREC Net Meter RPS Prop. Tax Sales Tax Com. Tax Pers. Tax Elec. Price NG Price Rebate 0.26 Feed-in Tariff Solar or DG RPS SREC * 0.60 Net Meter RPS Property Tax 0.06* * Sales Tax * * * -0.03* Com. Tax 0.06* * -0.06* -0.02* 0.05* * Personal Tax Electricity Price Natural Gas Price GDP per capita 0.01* * * * * * 0.01* * Note: * Correlation is not significant at 95% confidence level. Table 2 shows that property tax, sales tax, commercial tax, personal tax exemptions, and GDP per capita have an extremely weak and mostly insignificant relationship with the dependent variable. This means their effect on increasing PV installations in each state is very weak and will be difficult to distinguish among the effects of other variables on the dependent variable, especially since these variables have a higher correlation with other independent variables in the model. As a result, these policy variables were not considered in most of the final regression models. By comparison, solar or DG carve out provisions are established within a state s RPS, which explains why these variables have a higher correlation, with Pearson s r of 0.47, compared to other policies. Also, considering that policies that 17

23 specifically target solar energy technologies are of special interest to this analysis, it is the solar or DG carve out within RPS that is included for analysis in all the models, while the general mandatory RPS variable is dropped from the final model, minimizing the effects of multicollinearity and improving the model. Overall, among the remaining policies strongly correlated with the dependent variable, the collinearity amongst each policy is relatively weak, with the exception of SREC and solar or DG RPS. These policy variables are also correlated, with Pearson s r of 0.60, as a result of the fact that establishment of a solar carve out within a state s RPS is a necessary precondition for establishment of a SREC market. However, since determining the effect of each in promoting PV market growth is important in this analysis, both variables will be included in the final fixed effects model for more representative comparison. Similarly, a high correlation exists between natural gas and electricity prices, with Person s r of 0.76, however, because controlling for both factors is important in order to obtain more accurate results of policy outcomes, both variables are included in most models. V. Model Refinement A. Model 2 Upon testing and evaluation of correlations and regression results, several of the policy and control variables used in the original regression were found to be insignificant in describing the dependent variable. The removal of these variables yielded the revised regression Model 2, used to measure the effect of state financial incentives and regulations on PV capacity installed. Further analysis justifying the removal of several of the policy variables from the following models is further discussed in the previous section on correlations. Model 2 is considered the final model among those that account for policy effect at the time of policy implementation. 18

24 Model 2: PV CAPACITY i,q = β 0 + β 1 *REBATE i,q + β 2 * FEED-IN TARIFF i,q + β 3 *SOLAR/DG RPS i,q + β 4 *SREC i,q + β 5 * NET METER i,q + β 6 *ELEC PRICE i,q + β 12 *NG PRICE i,q + µ i,q + γ i,q+ e i,q Where µ i,q represents state fixed effects, γ i,q represents year fixed effects, and e i,q is the error term. B. Models 3 and 4 SREC and net metering variables are dropped in Model 3 in order to better gauge significance of Solar or DG RPS effect alone. Since Solar or DG RPS variable has a correlation of 0.60 with the SREC variable, it may be useful to look at how the results are affected if one of the two variables is removed from the regression. This action introduces an omitted variable bias into the equation, which affects the accuracy of the coefficient on Solar/DG RPS variable. However, this model can be used for comparison to results of Model 2 to observe how this bias would affect the final results of the effectiveness of rebates, feed-in tariffs and Solar or DG RPS policies. Using the same principle in attempting to avoid high correlations between independent variables to obtain more accurate results on policy effectiveness, Model 4 attempts to gauge effectiveness of the sales tax incentive, which is the only tax incentive that was implemented by states that has a significant correlation with the dependent variable. Hence Model 4 regresses sales tax exemption policy with other significant policies against the PV capacity installed. However, since the sales tax exemption policy variable is correlated with the electricity price, the electricity price control variable is dropped from the model. 19

25 C. Models 5, 6, and 7 Models 5, 6 and 7 measure the effect of state financial incentives and regulations on PV capacity while also accounting for a lagged effect of policies one, two or three years after their implementation. These models show how the effect of each policy changes through time, whether it becomes more or less effective as time progresses. It is important to observe effectiveness of policies through time because it takes time to properly advertise and develop each program. Therefore, as the public becomes more informed of the initiatives that a state offers, the effectiveness of each policy may increase the longer the policy is implemented. Model 5 adapted the revised Model 2 to take into account a policy implementation lag effect over time: Model 5: PV CAPACITY i,q = β 0 + β 1 *REBATE i,q + β 2 *REBATE_1yr i,q + β 3 *REBATE_3yrs i,q + β 4 * FEED-IN TARRIF i,q + β 5 * FEED-IN TARRIF_1yr i,q + β 6 *SOLAR/DG RPS i,q + β 7 *SOLAR/DG RPS_1yr i,q + β 8 *SOLAR/DG RPS_2yr i,q + β 9 *SREC i,q + β 10 *SREC_1yr i,q + β 11 * NET METER i,q + β 12 * NET METER_1yr i,q + β 13 *ELEC PRICE i,q + β 14 *NG PRICE i,q + µ i,q + γ i,q + e i,q Where µ i,q represents state fixed effects, γ i,q represents year fixed effects, and e i,q is the error term. Model 6 tests for the robustness of Model 5 s results by dropping several of the statistically insignificant variables in order to observe how much its results may be altered by these changes. Model 7 is included for general comparison and extra verification of the robustness of results. It includes all variables from the original model, Model 1, while also accounting for the time lag effect of policy implementation. 20

26 VI. Analysis of Results and Policy Implications Based on the results of all seven models developed above, the following sections provide a discussion of the results for each policy and control variables. Table 3 provides a summary of the coefficients and significance level of the independent variables in each model. Each policy is compared against the others, and policy implications of the observed effects are discussed. In general, unless specified otherwise, when the effect of each policy is interpreted, results of Models 2 and 5 are usually stated because these two are considered the final and most accurate models. A. Feed-in Tariffs These models reveal that feed-in tariffs (FIT) are the most effective policy in promoting deployment of PV technologies. From Models 2 and 3, it can be concluded with 99% confidence that implementation of a FIT results in an increase of 27.2 kw in PV capacity installed per 100,000 people in a quarter of a year, all else remaining equal. This is approximately an increase of 109 kw in PV installations per year due to the implementation of this policy alone. Models 1 and 4 also support the result that FITs are the most effective policy with similar statistically significant coefficients. However, according to models 5, 6 and 7 where a time lag effect is included, it can be seen that the FITs become significantly more effective a year after they are implemented. If the time lag is considered, FITs during the first year of implementation yield an increase in PV installations of approximately kw per 100,000 people per quarter of a year; yet a year later, the policy yields an increase of 40.1 kw of PV capacity per 100,000 people per quarter of a year. The effect of the FIT on increasing PV installations in a given state is quadrupled a year after it is first implemented. 21

27 Table 3. Year and State Fixed Effects Model Results Variables Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Rebate Rebate (1yr) 3.32** 2.65** 4.52*** Rebate (3 yrs) 6.87*** 6.79*** 6.79*** Feed-in Tariff 22.9*** 27.3*** 27.2*** 25.5*** 7.90** 7.60** 2.02 Feed-in Tariff (1yr) 40.1*** 40.2*** 38.31*** Solar/DG RPS 3.03** 3.25** 4.89*** 3.37** Solar/DG RPS (1 yr) Solar/DG RPS (2 yrs) 4.50** 4.68** 2.54 SREC 3.60** 3.80** 4.32** SREC (1yr) 7.58*** 6.66*** 8.74*** Net Metering ** 2.31 Net Metering (1yr) RPS -2.61** -2.38* RPS (1yr) 0.24 Property Tax Exemption *** Prop. Tax (1yr) 8.88*** Sales Tax Exemption -4.79*** -3.65*** -5.25** Sales Tax (1yr) 1.10 Commercial Tax Exemption 6.31*** 6.25** Comm. Tax (1yr) 5.05 Personal Tax Exemption -12.4*** -7.48*** Personal Tax (1yr) -11.0*** Electricity Price (c/kwh) 1.64*** 1.30*** 1.45*** 1.24*** 1.12*** 1.75*** NG Price ($/mil BTU) 0.91*** 1.22*** 1.15*** 1.71*** 1.26*** 1.32*** 0.74*** GDPpc ($/person) Constant -20.1*** -20.9*** -21.7*** -13.1*** -22.1*** -20.5*** -19.7*** R Observations *** Coefficient is significant at 99% confidence level. ** Coefficient is significant at 95% confidence level. * Coefficient is significant at 90% confidence level. 22

28 B. Solar Renewable Energy Credits According to all seven models, establishment of the SREC market is the second most effective policy for increasing solar energy utilization by the residential and commercial sectors. According to Model 2, implementation of a SREC policy results in an increase of 3.8 kw in PV capacity installed during a quarter of a year per 100,000 people, all else remaining equal. This effect is approximately seven times less than the effect of implementation of the FIT policy. Nonetheless, it is consistently statistically significant in each of the models at a 95% confidence level. Taking account of the time lag significantly changes the final results. If the effect of the SREC market is measured a year after the policy was established, the effect of the policy almost doubles to 7.6 kw per 100,000 people during a quarter of a year. C. Solar RPS An implementation of a solar or DG carve-out provision in a state s RPS is found to have a significant effect on increasing deployment of PV technologies. However, the effect is not as strong as the effect of a SREC market. This may be a result of two factors addressed by Models 3, 5 and 6. First, establishment of a solar RPS is a prerequisite to creation of a SREC market, demonstrated by a moderate simple correlation between the two policies of The correlation is not very high, but it may play a role in the analysis by introducing a bias where the effect of solar/dg RPS is captured by the SREC variable, biasing the coefficient. Model 3 attempts to show that solar/dg RPS does increase in value and becomes more significant once the SREC policy is not considered. However, in this case, the solar/dg RPS coefficient is possibly a victim of omitted variable bias, capturing the effect of SREC policy. Second, the effect of a solar/dg RPS may be weaker if a time lag of policy implementation is not considered, regardless of the existence of a SREC market. Models 5 and 6 support this conclusion. A 23

29 solar/dg carve out effect increases in magnitude two years after its implementation, causing an increase of approximately 4.5 kw of PV capacity installed per quarter of a year per 100,000 people, all else being constant. This time lag effect is still lower than the effect of the SREC policy once SREC policy is also modeled with a time lag, where it remains highly significant and stronger than the effect of solar/dg RPS in each model. In fact, the effect of solar/dg RPS becomes comparatively weaker than the effect of SREC as time progresses. Most likely, since there is no penalty associated with non-compliance with solar/dg provisions, utilities in a given state take their time and are not as aggressive in meeting specified targets by offering incentives and promoting PV development by customers and businesses. Meanwhile, SREC policy shows a strong effect right away most likely due to its penalty for non-compliance clause, which requires utilities to pay the set price of SACP for not meeting specified objectives. Utilities are likely to act faster in encouraging customers and businesses to deploy PV technologies if they know for a fact they will have to pay a higher price if specific targets are not met. D. Rebates A state s rebate policy was not found to be statistically significant in increasing market growth of PV technologies when the policy is first implemented, as shown in Models 1-4. However, if a time lag in policy implementation is taken into account, rebates show to have a significant effect on increasing PV installations, which increases in magnitude as years go by. According to Models 5 and 6, a year after its implementation, the rebate policy appears to cause an increase in PV capacity installed of kw per 100,000 people during a quarter of year, all else constant. This effect increases to approximately 6.9 kw three years after the rebates are first offered, making it on par with the efficacy of a SREC policy a year after its implementation and stronger than solar/dg carve out in RPS even after two years after policy implementation. Again, this result is likely due to the fact that it takes time for the public to become aware of the incentives their states offer. Once they do, they find the incentives to be financially attractive and cost effective to install in the long run, encouraging greater deployment of PV technologies. 24