This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright
Author's personal copy Energy Policy 39 (2011) 4726–4741
Contents lists available at ScienceDirect
Energy Policy journal homepage: www.elsevier.com/locate/enpol
Are government policies effective in promoting deployment of renewable electricity resources? Gireesh Shrimali a,, Joshua Kniefel b a b
Indian School of Business, Hyderabad, AP, India University of Florida, Gainesville, FL, USA
a r t i c l e i n f o
a b s t r a c t
Article history: Received 23 March 2011 Accepted 26 June 2011
Using a panel data over 50 US states and years 1991–2007, this paper uses a state fixed-effects model with state-specific time-trends to estimate the effects of state policies on the penetration of various emerging renewable electricity sources, including wind, biomass, geothermal, and solar photovoltaic. Renewable portfolio standards with either capacity or sales requirements have a significant impact on the penetration of all types of renewables—however, this impact is variable depending on the type of renewable source: it is negative for combined renewables, wind, and biomass; and positive for geothermal and solar. Further, clean energy funds and required green power options mostly result in increasing the penetration of all types of renewables. On the other hand, voluntary renewable portfolio standards as well as state green power purchasing programs are found to be ineffective in increasing the penetration of any type of renewable source. Finally, economic variables, such as electricity price, natural gas price, and per capita GDP as well as structural variables, such as league of conservation voters rating and the share of coal-generated electricity are found to be generally insignificant, suggesting the crucial role of policy in increasing the penetration of renewables. & 2011 Elsevier Ltd. All rights reserved.
Keywords: Renewable deployment State policy impact Econometric analysis
1. Introduction 1.1. Motivation Renewable energy and the electricity sector are closely linked, with the electric power industry accounting for 93% of renewable energy net generation in 2009 (Energy Information Administration, 2009b). Renewable energy has become an important aspect in the US electricity generation mix and a primary focus of government policy for environmental, energy security, and price volatility reasons. Power plants are responsible for approximately 40% of the nation’s carbon dioxide emissions in the US (Energy Information Administration, 2009b). This has resulted in environmental groups to target the sector in their quest to reduce emissions in response to concerns of global climate change (Yin and Powers, 2010). Furthermore, the public’s growing concern for the environment and progressively stringent regulation of emissions in the electric power industry has driven policies to increase the amount of renewable energy in the electricity generation portfolio. There has been recent uncertainty in the US energy supply due to political concerns in the Middle East countries as well as
Corresponding author. Tel.: þ 1 650 353 8221.
E-mail address:
[email protected] (G. Shrimali). 0301-4215/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.enpol.2011.06.055
volatility in oil and natural gas prices. For example, the delivered price of natural gas to electric utilities has risen from $2.62 per million cubic feet in 1999 to $4.89 per million cubic feet in 2009, while reaching as high at $9.26 per million cubic feet in 2008 (Energy Information Administration, 2009b). This uncertainty and volatility has led to a push to increase US energy independence through a greater domestic energy supply. This view is supported by Bird et al. (2005) and GDS Associates (2001), which assign volatility of natural gas prices as one of the major market factors behind wind power deployment in the US and the enactment of a renewables portfolio standard in Hawaii, respectively. However, the lack of coordinated and comprehensive action by the federal government has led state and local governments to fill this void with a variety of policy approaches (Engel and Orbach, 2008). In fact, in response to global calls for a systematic solution to climate change, state governments are proving themselves as clean energy pioneers (Carley, 2009). State leaders across the country are adopting renewable energy incentives, enforcing integrated resource planning programs, and some have even set carbon abatement levels to reduce future emissions (Wasserman, 2010). Complementing federal policies such as the production and investment tax credit, state governments have taken actions to increase renewable energy capacity and generation, with most of the 50 states enacting policies to encourage the use of renewable energy in their state (DSIRE, 2011). As a result of decentralized policy making, the individual state policies as well as renewable
Author's personal copy G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
power penetration show a great deal of variance. While a number of factors might explain both the growth of renewable energy and the disparity in renewable capacity among states, policies adopted by state governments, including changes in the regulatory environment for electricity, could be expected to play an important role. Yet, despite the increased attention given to the topic, the causal link between state policies and renewable energy development is not well established, with studies finding different, and sometimes contradictory, results. That is, not many studies reliably and comprehensively verify whether state initiatives have been effective in increasing the electricity sector’s diversification of fuel sources—in particular as related to renewable sources. This study aims to add to this growing body of literature (Menz and Vachon, 2006; Carley, 2009; Yin and Powers, 2010) and shed light on the debate surrounding the effectiveness of policy in deploying renewable sources. 1.2. Our work The objective of this paper is to determine which state policies have led to increased deployment of non-hydro renewable energy capacity. It considers the long-run impacts of state renewable energy policies by focusing on capacity construction instead of the short-run impacts on renewable-based electricity generation. Hydropower is not included in the renewable energy capacity because most hydropower was created well before the mid1990s, with few changes in capacity or costs over the time period being analyzed (Energy Information Administration, 2007). Further, hydropower is not only cost-competitive with conventional fuels but also competes with them with respect to baseload power generation. These aspects allow hydropower to be considered a type of ‘‘current generating technology’’, which includes steam or gas turbines fired by natural gas, coal, petroleum, or nuclear power. Due to these, most of the renewable energy policies apply to emerging renewable technologies, such as wind, biomass, geothermal, and solar power. Thus, removing hydropower from the dependent variable allows the focus of the paper to be on the policy effects on the emerging technologies. Most of the previous literature has focused on either total renewable capacity or wind power capacity. Since wind power contributes to more than 90% of renewable capacity, previous literature has focused on the macro-level renewable energy deployment.1 The biggest contribution of our work is that it separately estimates impacts on individual renewable source— wind, biomass, geothermal, and solar—capacity deployment. That is, our work is the first to look at micro-level (or individual) renewable energy deployment. This allows us to not only observe whether policies impact different renewables differently but also how they do so. This paper also expands existing literature (Menz and Vachon, 2006; Carley, 2009; Yin and Powers, 2010) by using the largest panel data set (over 1991–2007) including data for all fifty states. Our paper simultaneously estimates the effects of multiple detailed policy designs while controlling for differences in the electricity market and political environments of a state. That is, it estimates the effectiveness of policy relative to a business as usual (BAU) trend that is defined by the influences of economic, structural, and political variables.2 We emphasize here that 1 Since 1996 91% of new renewable capacity has been wind power capacity while biomass accounts for 8% and geothermal power and solar power combined account for the remaining 1%. 2 Given that policy effectiveness is measured with respect to a BAU trend, all references to increase (or decrease) of renewable capacity would be with respect to the BAU trend hereafter.
4727
our study is focused on detecting whether policies have been statistically significant in impacting the penetration of renewables. However, we do measure the actual impact of the policy in some cases.3 Results of this paper show that three types of policies are found to be effective at expanding non-hydro renewable capacity deployment: a command-and-control policy, known as renewables portfolio standard (RPS)4; a market-based policy, known as required green power option; and an incentive-based policy, known as clean energy funds. The command-and-control policy, the renewable portfolio standard, targets utilities by mandating a specified level of capacity or sales that must come from renewable energy. Our study reveals many findings. First, this policy is effective only when it is in form of a legal requirement. It is not effective when it is a voluntary goal.5 Second, in its nominal form, this policy may actually result in a decrease in the penetration of renewable energy capacity, as opposed to the intended result—an increase. Third, based on the percentage contribution to the total renewable capacity, this policy has different impacts on the penetration of different renewable sources considered in this study. That is, this policy results in a decrease in the penetration of major contributors, such as wind and biomass; and an increase in the penetration of minor ones, such as solar and geothermal. In Sections 5 and 6, we elaborate on these results, and discuss why these results may occur. The market-based policy, the required green power option, mostly results in increasing the penetration of all types of renewable energy capacity. Under this policy, consumers are given the option to express their preferences to buy power from renewable resources at a premium price. This policy works because it creates a differentiated demand by mandating that utilities must offer their customers the choice to purchase green power. This in turn allows consumers to express their preferences through paying an extra, utility commission-approved charge for ‘‘green power’’. The incentive-based policy, clean energy funds, where projects are awarded financial incentives, mostly results in increasing the penetration of all types of renewable energy capacity. Clean energy funds (CEFs) have awarded funding to projects that would have been undertaken to meet future RPS sales requirements regardless of the government’s financial support. CEF funding may have been used as the ‘‘carrot’’ to minimize electric power industry opposition while implementing the RPS sales requirement as the ‘‘stick’’ to ensure renewable deployment. Finally, a command-and-control scheme, state government green power purchasing, which requires state governments to buy a certain fraction of their electricity consumption from renewable sources, is found to be insignificant. Our work builds upon previous work on state policy impacts on the penetration of renewable energy. The econometric results of this paper support the conclusions from the literature that RPS requirements and required green power options have impacted
3 Therefore, for the rest of the paper, when we say ‘‘significant’’ or ‘‘effective’’ we mean ‘‘statistically significant’’. Similarly when we say ‘‘insignificant’’ or ‘‘ineffective’’ we mean were not able to establish statistical significance. 4 Though RPS at first seems like a command-and-control policy, it has a market-based component as well. This is referred to as the renewable energy certificates (RECs). RECs are generated along with renewable energy and can be traded on the market. This mechanism gives utilities the flexibility to buy RECs to satisfy their RPS requirements. As of 2007, all but four states allowed the use of unbundled RECs (Wiser and Barbose, 2008). 5 The distinction between requirements and goals is important: requirements are typically legal in nature, with underlying laws on penalties imposed in case of non-compliance; whereas goals are typically voluntary, with no recourse to law in case of non-compliance (Wiser and Barbose, 2008).
Author's personal copy 4728
G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
the penetration of renewable capacity (Menz and Vachon, 2006; Yin and Powers, 2010). However, as we demonstrate in Sections 5 and 6, our results add a richness to these findings by considering different type of renewables, and showing variable impacts of RPS requirements based on the type of source under consideration. Further, a key difference from previous literature is that our study shows that clean energy funds have a significant impact on the share of renewable capacity—previous literature showed that a related policy, public benefit funds, is insignificant (Menz and Vachon, 2006; Yin and Powers, 2010).
2. Literature review The literature on state renewable energy policies traditionally consisted of case studies on policy effectiveness. However, econometric methods to estimate the effects of various state policies on renewable capacity are becoming increasingly popular. In this section we present findings from previous literature and, where possible, contrast previous results with ours. Though we highlight similarities as well as differences in this section, we hold off detailed discussion until Section 6. 2.1. Case studies Langniss and Wiser (2003) analyze the Texas renewables portfolio standard, including the achievements of the policy mechanism and the design characteristics that allowed the policy to be effective at increasing renewable energy capacity. They find that the clearly defined capacity requirements have been effective in increasing renewable capacity in Texas. Wiser et al. (2005) extend Langniss and Wiser (2003) by considering all renewable portfolio standards, find pitfalls in current policy designs, and assert that a carefully designed RPS requirement policy for promoting in-state deployment of renewable source would have the following features: first, a well-designed RPS should ideally apply equally and fairly to all load-serving entities in a state; second, it should require that it must be filled with generation from new investments in renewable resources; third, it should limit the amount of sales requirements fulfilled by sources external to the state; finally, it should have credible and significant penalties in case of non-compliance. Petersik (2004) provides a non-econometric analysis of the effectiveness of different types of renewable portfolio standards as of 2003 for the United States Energy Information Association (EIA). It finds that only renewable portfolio standards that mandate a certain level of capacity have had any significant impact on renewable capacity deployment. Policies with renewable generation or sales requirements as well as voluntary policy programs were found to have no significant effect. The insignificance of voluntary RPS programs is in line with our results; however, our study goes a step further in identifying the variable significance of RPS requirements, depending on the source under study. Bolinger et al. (2001) describe in detail 14 different state Clean Energy Funds, enumerating the regulatory background, funding approaches, the current status of the fund, and the resulting impacts on renewable energy. Programs that fund utility-scale projects are found to be the most effective at increasing renewable capacity deployment. Bolinger et al. (2004, 2005) summarize the same 14 clean energy funds. They find that due to delays and canceled projects actual capacity often is much lower than initially obligated capacity. Our results support this finding that clean energy funds are mostly effective, despite the shortcomings in program design. Brown and Busche (2008) rate states based on the effectiveness of their renewable energy policies and summarizes the best
practices for state renewable energy policy design. The paper finds a significant correlation of renewable portfolio standards with increased renewable energy generation in a state. Our results support this finding. Production incentives are found to significantly increase both renewable generation and capacity. 2.2. Empirical work Menz and Vachon (2006) are the first ones to econometrically estimate the effects of state renewable energy policy on renewable energy capacity. They use ordinary least squares to estimate state policy effects on wind power capacity and generation with a dataset for 39 states for 1998–2002, while controlling for wind power availability, retail choice, and policy dummy variables for public benefits fund, renewables portfolio standard, required green power option, and fuel mix disclosure.6 In line with our results, RPS requirements and required green power option are found to have a statistically significant effect on wind capacity deployment. However, in contrast to our results, no statistically significant effects are found for public benefits funds. In Section 6, we discuss possible reasons behind this discrepancy. Another major difference from our work is that Menz and Vachon (2006) do not measure the effect of policy on the share of renewable capacity. Carley (2009) uses a variant of a standard fixed-effects model, referred to as a fixed-effects vector decomposition, with statelevel data for 1998–2006. Results indicate that RPS implementation is not a significant predictor of the percentage of renewable energy generation out of the total generation mix, yet for each additional year that a state has an RPS policy, they are found to increase the total amount of renewable energy generation. These findings reveal a potentially significant shortcoming of RPS policies. Political institutions, natural resource endowments, deregulation, gross state product per capita, electricity use per person, electricity price, and the presence of regional RPS policies are also found to be significantly related to renewable energy deployment. A major issue with this study is that it does not take into account the stringency of RPS policies. In the absence of this variation, it may not be possible to account for the impact of RPS on renewable energy generation. However, this study does produce somewhat expected results—i.e., increase in share of renewable capacity—when a proxy for RPS requirement stringency (number of years) is taken into account. In a recent work, Yin and Powers (2010) introduce a new measure for the stringency of an RPS that explicitly accounts for some RPS design features (i.e., existing capacity, incremental requirements, coverage, etc.) that may have a significant impact on the strength of an RPS. They also investigate the impacts of renewable portfolio standards on in-state renewable electricity development using panel data (for 1993–2006) and the new measure of RPS stringency, and compare the results with those when alternative measures are used. Using this new measure, the results suggest that RPS policies have had a significant and positive effect on in-state renewable energy development, a finding which is masked when design differences among RPS policies are ignored. They also find that another important design feature—allowing ‘‘free trade’’ of RECs—can significantly weaken the impact of an RPS in promoting in-state renewable electricity development. The findings of this study are also in line with our findings that nominal RPS requirements (as opposed to their main 6 Public benefits funds often include clean energy funds, but also fund energy efficiency projects. Fuel mix disclosure is a policy that requires the fuel mix a power producer uses in its electricity generation to be disclosed to the public. Menz and Vachon (2006) assert that consumers may use this information to purchase electricity from power producers that use cleaner burning fuels or alternative energy.
Author's personal copy G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
focus – the new measure of stringency) have a negative impact on the share of total renewable capacity, whereas required green power option has a positive impact. However, in contrast to our results, and similar to Menz and Vachon (2006), no statistically significant effects were found for public benefits funds. In another recent work, Sarzynski et al. (submitted for publication) examine the effectiveness of state financial incentives in promoting the deployment of solar technologies: thermal and PV. In addition to the financial incentives, they also look at the effect of RPS. They find that mere presence of policies is not effective in increasing the deployment of solar technologies; however, they do find that policies become effective at increasing the deployment of solar technologies as states gain experience with implementing these policies. In particular, they find that RPS are significant in increasing deployment of solar PV, a finding shared with our study. However, the major difference is that our study uses RPS policy stringency as the independent variable, whereas they does not. Further, Sarzynski et al. (submitted for publication) does not look at other renewable technologies.
3. Model: fixed-effects and time-trends This paper uses a state and time fixed-effects model with state-specific time-trends to estimate policy results. In what follows, we first present a basic pooled model, followed by a state and time fixed-effects model, to justify the use of the state and time fixed-effects model with state-specific time-trends.
4729
other dependent variables, one per renewable source. For ease of exposition, we focus on renratio in this section. The use of renratio as the dependent variable allows us to control for market size across states. It is expected that, ceteris paribas, more renewable capacity will be found in states that generate more electricity to help meet the higher demand for electricity found in those states. For example, given two states with the same renratio, the state with more total generation will have more renewable capacity. Total generation in a state is chosen as a normalizing variable instead of total sales because some of the electricity demand for a state’s power producers may come from other states. Generation is not contaminated with these inter-state sales, which may otherwise inflate or deflate the market size measure. Further, generation and sales are highly correlated, with correlation coefficient 0.951. On the other hand, there is no strong reason to choose total generation as the normalizing variable instead of total nameplate capacity from all sources. Given that the installed capacity can be converted to the generation capacity for renewable resources based on respective power load factors, either can be used without loss of generality.7 In fact, our results are robust to the use of total nameplate capacity as the normalizing variable. Finally, this ratio also serves as a proxy for the share of the renewable resources in the state’s electricity generation mix. For this reason, in what follows, we refer to this ratio as the share of renewable resources as well.
3.2. The fixed-effects model 3.1. The basic pooled model The basic model estimates the ratio of total non-hydro renewable capacity (Cit) to the total net generation (Git) for 1991–2007. Given that different policy designs may be targeting specific types of renewable capacity deployment, and there may also be nonpolicy factors that are driving, or hindering, renewable capacity construction, it includes detailed policy variables, and control variables for a state’s electricity market and political environment. Without variables controlling for differences in market size and political environments, the results could be biased and may lead to incorrect policy interpretations. The model is Cit =Git ¼ a0 þ bnRit þ dnWit þ eit ,
ð1Þ
where ‘‘i’’ is the state, and ‘‘t’’ is the year of the specific observation; Rit is the vector of six regulatory policies (renewables portfolio standard with capacity requirements, renewables portfolio standard sales requirements, renewables portfolio standard with sales goals, state government green power purchasing, required green power option, and clean energy funds); and Wit is the vector of five economic and political control variables, described in Section 4.3. The dependent variable is the ratio of total non-hydro renewable nameplate capacity (rencap), which is the sum of wind, biomass, geothermal, and solar nameplate capacities, to the state’s total electricity generation (gen). We refer to this ratio as renratio. Renewable nameplate capacity (rencap) in the electric power industry includes all renewable nameplate capacity of utilities, independent power producers (IPPs), and industrial or commercial combined heat and power producers. Nameplate capacity is the appropriate inclusion in the definition of the dependent variable because several policies mandate or fund a specific amount of renewable capacity. Policies that do not set specific renewable capacity requirements can be measured in capacity terms by controlling for each state’s market size. Total net generation (gen) is the total amount of electricity generated (in terawatt hours) in a state for a given year. Similarly, we create
The second model includes state-fixed effects and year dummy variables. State dummy variables are used to control for renewable availability and capacity constructed prior to 1996, which is in large part due to the implementation of prior federal policy at the state-level as well as the effects of environmental preferences not captured by other variables.8 The year dummy variables are used to control for time-specific events that may influence all states in a similar fashion—for example, recession, federal election, etc. Inclusion of the fixed-effects avoids any heterogeneity bias that would have occurred from imposing a common constant term. The model estimates the ratio of total non-hydro renewable capacity (Ct) to the total net generation (Git) for 1991–2007: Cit =Git ¼ a0 þ bnRit þ dnWit þ gSi þ yTt þ eit ,
ð2Þ
where Rt and Wt are as defined in (1); Si is the vector of state fixed-effects dummy variables; and ceteris paribas Tt is the vector of year dummy variables.
3.3. State-fixed effects Si State fixed-effects are included in the regressions to control for two factors: capacity installation prior to 1991 in a state and factors that differ across states that are constant over time. For example, a large amount of renewable capacity created before 7 We note that this argument is particularly valid for individual renewable sources, given that the capacity of a renewable source and the corresponding generation is related by a constant, the power load factor. However, this argument may not hold value for the sum of renewable capacity since a conversion to generation would require a power load factor that is a weighted sum (based on the share of renewable sources) of various power load factors. 8 Given the potential heterogeneity of the sample between states, a series of Hausman tests were performed to determine if fixed-effects (FE) models were necessary. The Hausman tests overwhelmingly rejected the null hypothesis of no unit heterogeneity, confirming the need to control for unobserved differences between states that cannot be captured by the explanatory variables.
Author's personal copy 4730
G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
1991 originated from the Public Utilities Regulatory Policy Act (PURPA). For a variety of reasons, the effects of PURPA varied from state to state.9 State dummy variables (Si) measure these effects and other unchanging state factors, such as renewable resource availability (Morris, 2003). 3.4. Time dummy variables: Tt Year dummy variables control for national trends, such as general renewable energy economic viability and federal policy impacts. The average levelized cost of production for renewable energy in the US have consistently decreased over time. As the costs of production decrease, renewable energy becomes more competitive and will enter the market in larger amounts. Federal policies like the Federal Renewable Energy Production Tax Credit aid in renewable deployment by subsidizing production of electricity from renewable resources. The combination of these two factors should increase renewable capacity over time across the entire US. 3.5. The time-trends model The fixed-effect model assumes that the state-specific characteristics, such as the share of renewable capacity before 1991, are unchanging within a state over time. These characteristics cannot be separately identified, and therefore are subsumed in the state fixed-effects. The use of state effects, which is possible only with a panel, is advantageous because it explains diffusion of renewables flexibly, rather than having to include all relevant covariates explicitly. However, the factors which influence diffusion may vary within a state over time, confounding the estimates of the state effects. These factors may not be completely captured by the control variables. This will also bias the estimate of the coefficients on the policy variables if the changing factors are correlated with the policy changes across states and if such factors do not change at a national level uniformly and get picked up by year effects. The specification in (2) can be refined further to allows for such changing influences within a state over time, as follows (Friedberg, 1998): Cit =Git ¼ a0 þ bnRit þ dnWit þ g0 Si þ g1 Si nt þ g2 Si nt 2 þ yTt þ eit ,
ð3Þ
t is the time trend, so the interaction terms Si nt and Si nt 2 are linear and quadratic trends for each state. They capture quadratic trends in state-level characteristics that influence the adoption of renewable energy, with the slopes of the trends allowed to vary across states. Thus (3) represents an extremely flexible way to control for heterogeneous renewable energy adoption behavior, identifying the impact of policies from a jump in the pattern of renewable energy adoption, which is distinguishable from a smooth quadratic. 3.6. Methodology While Ordinary Least Squares (OLS) yields unbiased and consistent coefficient estimates for TSCS models, it does not correctly estimate their standard errors in the presence of panel heteroskedasticity, contemporaneous correlation of the errors, or serially correlated errors (Stimson, 1985). It can be easily verified that all of these are present in our data, and therefore, need to be corrected. 9 The coefficients should be highly correlated to the amount of renewable capacity that existed in 1991. The correlation between a state’s initial total nonhydro renewable capacity and its state dummy variable coefficients is 0.684. The state dummy variables seem to reasonably control for the impact of existing regulation and the market environment prior to 1991.
Table 1 Nameplate capacities. Nameplate capacity (MW)
Mean
Std. Dev
Min
Max
Median
Total renewable Wind Biomass Geothermal Solar
405.01 112.39 223.02 61.48 8.11
889.37 376.85 271.26 386.69 55.31
0.00 0.00 0.00 0.00 0.00
6760.00 4490.00 1217.00 2821.00 413.00
210.00 0.00 138.00 0.00 0.00
We use the typical two step process for addressing these (Parks, 1967). In the first step, serial correlation is removed by estimating auto-correlation from the residuals of an OLS run. In the second step, heteroskedasticity and contemporaneous correlation are removed by Panel Corrected Standard Errors (PCSE) (Beck and Katz, 1995),10 an approach that has been widely applied to political science and public policy econometric analysis. PCSE uses OLS coefficient estimates and residuals to compute standard errors that correct for panel heteroskedasticity as well as contemporaneous correlation.
4. Variables and data 4.1. Dependent variables: Cit/Git and cnit/Git There are five dependent variables based on nameplate capacities of four different sources. These capacities (cnit) are wind nameplate capacity (windcap), biomass nameplate capacity (biocap), geothermal nameplate capacity (geocap), and solar nameplate capacity (solarcap). The total renewable capacity is P simply the sum of these four, i.e., Cit ¼ 4n ¼ 1 cnit . Nameplate capacity is the amount of capacity the generator produces under ideal conditions. Non-hydro renewable nameplate capacities for solar, biomass, geothermal, and wind as well as generation data are derived from EIA Historical State Electricity Databases (Energy Information Administration, 2007). The EIA database does not contain small-scale distributed projects, and our dependent variables represent large scale deployment–this distinction is important to remember for renewable sources (e.g., solar) where a significant chunk of installed capacity may come from distributed source. Table 1 summarizes the data for the dependent variables. Data sources are listed in Appendix. 4.2. Regulatory policy variables: Rt Six of the explanatory variables are policy variables capturing the effects of different types of renewable energy regulation, either by a state’s legislature or Public Utility Commission (Rt): rpscreq, rpssreq, rpssgoal, sggpp, rgpo, and cef, which correspond to a renewables portfolio standard with a capacity requirement, renewables portfolio standard with a sales requirement, renewables portfolio standard with a sales goal, State Government Green Power Purchasing, required green power options, and clean energy fund, respectively. It is typically expected that all of these policies would have a positive impact on the penetration of renewable resources. Most policies are enacted through state legislation, and then enforced by the state’s public utility commission (PUC). There are a few instances, however, in which a PUC adopts guidelines without state legislation. No legislation or PUC action is required for state governors to use executive orders to create a state 10 See Beck and Katz (1995), Beck (2001), and Reed and Ye (2011) for reasons why PCSE is preferable over Feasible Generalized Least Squares (FGLS) (Parks, 1967) for the task at hand.
Author's personal copy G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
4731
25
Table 2 Policy variables. Policy
States with policy (1996: 2007)
Non-zero observations
rpscreq rpssreq rpssgoal sggpp rgpo cef (4 year lag)
1: 0: 0: 0: 0: 0:
42 234 69 50 34 28
04 44 16 11 07 10
rpscapdnum rpssgoaldnum rpssreqdnum
20 15 10 5 0
government green power purchasing agreement or to set voluntary goals for generation. Information on renewable policies is available from the Database of State Incentives for Renewable Energy (DSIRE, 2011), which is a project of the Interstate Renewable Energy Council and funded by the US Department of Energy.11 Table 2 summarizes the data for the policy variables. Policy dummy variable values are determined by a policy’s enactment date, zero before enactment and one after enactment. The ‘‘enactment’’ date is the year that the policy is passed by the state legislator, created through an executive order, or announced as a mandate under new PUC guidelines. Some of these policies allow a grace period for power producers to meet the new regulations. The ‘‘effective’’ date is the year that the policy requirements must be met. The average lag from the ‘‘enactment’’ to ‘‘effective’’ date is a little over one year, but can be longer for renewable portfolio standards. The enactment year is a better choice to determine when the policy begins to impact the power producers. Once a power producer becomes aware of a future requirement, it may begin to construct any necessary renewable capacity. These actions could lead to large amounts of renewable capacity being constructed between the enactment date and effective date. 4.2.1. Renewable portfolio standard The renewables portfolio standard specifies the amount of a state’s electricity sales or capacity that must be renewable-based. Renewable portfolio standards can be differentiated into three main structural forms: policies that set (1) mandatory renewable sales levels, (2) voluntary renewable sales goals, and (3) mandatory renewable energy capacity requirements. Fig. 1 shows the progression of states adopting these measures over time. Requirements or goals on renewable energy sales either within a state or in another state that may accept renewable generation from the state under consideration should increase renewable generation demanded from the state of interest, which should increase renewable capacity within the state. We note that, given the complexities of RPS policies it may not be very straightforward to figure out the ideal measure that incorporates inter-state effect and, for the purpose of this paper, we focus on only the RPS requirements (or goals) within the state.12 The differences between renewable portfolio standards can be accounted for in the model by three variables: two variables that measure the size of the renewable sales requirement (rpssreq) or goal (rpssgoal) within the state, and a variable that measures the size of the capacity requirement (rpscreq) within the state. Capacity requirements range from 50 to 2280 MW for 2007, and sales 11 The information is compiled from many different sources, including federal and state officials, public utility commissions, and renewable energy organizations. The source of the information is included within each policy description. Bolinger et al. (2001) include additional information on the enactment and design of public benefits funds. 12 This assumption may have some negative influence on our results, especially in terms of potentially missing statistical significance of RPS policies.
1990
1995
2000 year
2005
2010
Fig. 1. The number of states with renewable portfolio standard. rpscapdnum: number of states with RPS with capacity requirement; rpssreqdnum: number of state with RPS with sales requirement; rpssgoal: number of states with RPS with sales goals.
requirements range from 0% to greater than 30%. The results from the models will determine how much more effective strict requirements are at increasing renewable capacity relative to voluntary goals where regulators accept in ‘‘good faith’’ that the utilities will in fact increase renewable energy-based electricity sales. The capacity requirement size and date are used to extrapolate the expected requirement for each year assuming a linear function, where the power producers increase capacity by the same amount each year until they meet the final requirements, to form the variable rpscreq. rpssreq and rpssgoal are more complex variables. The sales requirement or goal, which usually sets a target five or more years after enactment, is linearly interpolated backwards to the enactment date of the policy, such that the target at the enactment date meets the actual generation on the date.13 For example, a policy enacted in 1996 with a 0.5% generation in 1996, and a sales requirement of 1.0% beginning in 2000 would be linearly interpolated to be 0.5% in 1996 and increase by 0.1% each year until it reaches 1.0% in 2000. If the policy is updated or overridden by new legislation, the variable value is adjusted to account for the change in policy requirement or goal. This requirement is multiplied by the total electricity sales in the state for that year to estimate the total renewablebased sales that should be sold within the state. Some state renewable portfolio standards with a sales requirement allow the use of some hydropower electricity to meet the requirement. However, there are normally specific requirements as to which facilities will be eligible, including restrictions on a unit’s maximum capacity, type of hydropower, and year of installation. For example, some states (California and Colorado are representative examples) do not allow generating units greater than 30 MW to be eligible. One state (Illinois) does not allow any hydropower to originate from new dammed hydropower plants. Another state (New Mexico) only allows electricity from new (after July 1, 2007) hydropower capacity to be eligible. Despite restrictions, allowing hydropower to meet requirements will decrease the effectiveness of these policies to increase nonhydro renewable capacity in a state. However, the complexities of
13 We acknowledge that this is an approximation of policy strength. However, most states do have a yearly target (DSIRE, 2011). For the most part, this target resembles the one determined by us, and our approach should serve as a reasonable first order approximation. However, it is appropriate to point out the implications of this assumption in case there were no intermediate targets. In this case, the actual RPS implementation may be skewed towards the end of the compliance period, and this impact may be missed by our analysis.
Author's personal copy 4732
G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
8 10
reqgpdnum
gppurdnum
6
5
4
2
0
0 1990
1995
2000 year
2005
2010
1990
1995
2000 year
2005
2010
Fig. 2. The number of states with state government green power purchasing.
Fig. 3. The number of states with required green power option.
the restrictions make it difficult to create an appropriate measure for these effects.
for the guarantee that some of the consumer’s electricity consumption is produced from renewable sources. Consumers purchase electricity at the market price and then pay a premium for blocks of ‘‘green’’ electricity, usually about $2 per 100 kWh. The second type allows the producers to charge consumers a higher rate per kWh, but only to cover the ‘‘additional costs’’ for electricity from renewable sources. Both the premium block rate and premium per kWh rate must be approved by the state’s Public Utilities Commission (PUC). Fig. 3 shows the progression of states adopting these measures over time. Required green power options elicit customer preferences and a crude measure of willingness to pay for renewable energy by allowing consumers to voluntarily pay higher prices for the knowledge that they are supporting renewable-based electricity. The creation of this niche market for renewable energy generation should have a positive impact on renewable capacity. There should be minimal overlapping effects with renewable portfolio standards because green power sold at a premium cannot be counted towards RPS requirements in most states (Bird and Lokey, 2007). The variable rgpo is a dummy variable.
4.2.2. State government green power purchasing State government green power purchasing policies require that some percentage of a state government’s electricity purchases be from renewable sources. These purchase agreements range from 2.5% to 50% of a state government’s electricity purchases. Similar to Renewable Portfolio Standards with sales requirements, a State Government Green Power Purchasing agreement increases the need for renewable-based electricity generation. Fig. 2 shows the progression of states adopting these measures over time. As state government electricity use rises, the renewable generation needed to meet the requirement increases. If the new generation needs cannot be met by current renewable capacity, power producers will need to construct new renewable energy capacity. The size of the State Government Green Power Purchasing requirement is represented in terms of a percentage of the state government’s electricity purchases (sggpp). It is not clear how these requirements are met over time. Further, it has not been possible to determine the percentage of state electricity consumed by state governments.14 In absence of these exact numbers, we have made a couple of simplifying assumptions. First, the data is linearly interpolated backward from the effective year for years between enactment and effective dates. This is similar to the linear interpolation we performed for RPS requirements and goals. Second, it is also assumed that all state governments consume the same constant percentage of total electricity consumed in a state throughout the dataset.15 4.2.3. Required green power option A required green power option requires utilities to offer customers the option to purchase renewable power at a premium. There are two versions of how these options are implemented. The most common type gives consumers the option to make voluntary contributions, called ‘‘voluntary renewable energy tariffs’’ in return 14 For example, the best known source—the EIA electricity database (Energy Information Administration, 2007)—contains information on consumption by certain sector (residential, commercial, industrial, etc.) but not by governments. 15 We acknowledge that this assumption may not be always true—for example, in states with a lot of heavy industry. Nevertheless, these assumptions will be harmless if the deviation from actual implementation does not create any systematic biases—however, in the absence of systematic biases, the eventual coefficient from regressions may be biased.
4.2.4. Clean energy fund A clean energy fund is a state-level program that is often, but not always, created through the restructuring of the electricity market and is used to fund grants, loans, and production incentives for both research and development and actual deployment of alternative energy. Many clean energy funds focus on funding actual renewable capacity deployment, which should lead to more renewable capacity in a state. Funding is usually based on actual production, but it is paid in a lump sum once the capacity has been constructed. Clean energy funds are paid for through system benefits charges (SBCs), which are additional charges paid by all consumers on their electricity consumption. SBCs can be considered a consumption tax on electricity to fund deployment of renewable capacity in the industry. In Minnesota, a settlement with the electric utility Xcel Energy created a similar fund that is paying for renewable energy research and deployment. Maine created a voluntary fund similar to a clean energy fund for the state’s customers to donate money.16 16 Database of State Incentives for Renewable Energy (DSIRE) does not include New Mexico as having a clean energy fund, while Bolinger et al. (2001) verify that New Mexico does have a clean energy fund.
Author's personal copy G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
10
cefdnum
8
6
4
2
0 1990
1995
2000
2005
year Fig. 4. The number of states with clean energy funds.
Similar to renewable portfolio standards, clean energy funds must be differentiated to understand how effective these policies are at increasing renewable deployment in a state. The amount of capacity to date (lagged four years) that has been awarded funding for utility-scale projects from clean energy funds (cef) is the variable used to control for these policy effects. The data for the clean energy fund variables originated from the 2006 Database of Utility-Scale Renewable Energy Projects (LBNL, 2006) from the Clean Energy States Alliance (CESA). Fig. 4 shows the progression of states adopting these measures over time. The use of capacity that was awarded funding may seem unusual at first since our model is using capacity as both independent and dependent variables—the ideal variable would have been the dollar amount dedicated to these projects. However, the use of funded-capacity can be justified due to the following reasons. First, states typically report funds used support renewable energy and energy efficiency together and it is hard to isolate the dollar amounts dedicated to clean energy funds. Second, this is capacity funded and not capacity that actually came online, and there is a clear separation between the independent and dependent variables. The variable is lagged four years because that is the average amount of time as of 2006 it takes between awards being issued and actual capacity construction of those projects. For example, a project that is awarded funding in 2000 will, on average, be online in 2004. This lag is typical regardless of the technology. This variable is further normalized by the total electricity generation to account for the state size.
4.2.5. Policy variability In econometric analysis a key requirement is that there is enough variation in the explanatory variables of interest (Wooldridge, 2002). Given our focus on policies, a brief discussion of policy variability is in order at this point. Further, it is important to establish that policy impacts can be effectively separated from each other as well as from the economic and structural variables. Figs. 1–4 show that except for rpscreq in Iowa, all policies were adopted in the time period of interest. That is, we can rule out the issue of policy changes occurring outside the period of interest, and it is safe to assume that all the change in policies is captured in the period of interest for our analysis. These figures also shows that almost all the policies were adopted after 1995. Thus the period 1991–1995 acts as a control period, allowing us to separate out the effects of economic and structural variables.
4733
Further, Fig. 5 shows how policies were adopted over individual states. The policies were mostly adopted one at a time by the states, except for three instances where two policies were adopted together: Illinois and Minnesota adopted two policies each in 2001; and New Jersey adopted two policies in 1998. However, compared to a total of 91 policy changes, these are not significant. Further, the policy changes seem to have occurred at different times for different states. That is, policy changes do not seem to be correlated across states or time. As a final check, we performed a pair wise correlation analysis on the policy variables to detect whether policy changes occurred together (Table 3). Except for a moderately high correlation of 0.69 between rpssreq and sggpp, no significant correlation was found between any other pair. For this reason, whenever sggpp was found significant, we ran regressions with and without sggpp. Beyond this exemption, it is reasonably safe to assume that the policies are not collinear and, in fact, the policy changes are independent of each other—an assumption whose validity is crucial to ascertain confidence in the results. 4.3. Economic and political variables: Wt Five variables account for non-policy variability (Wt) in nameplate non-hydropower renewable capacity in the electric power industry of each state for 1991–2007: elecprice, ngprice, pcgdp, pctcoal, and lcvscore.17 The first three (elecprice, ngprice, pcgdp) are economic variables measuring electricity prices, natural gas prices, and per capita gross domestic product, respectively. The political variables, pctcoal and lcvscore measure a state’s preferences for renewable capacity. Table 4 summarizes the data for the control variables. 4.3.1. Economic variables High or volatile conventional (fossil-fuel) energy prices improve the potential return on investment for renewable technologies, as consumers look to substitute away from commodities with high or volatile prices. However, as a first order approximation, this model accounts only for average prices and not the volatility of these prices. Separate variables were included in the model for the average price of electricity (elecprice) and the average price of natural gas (ngprice). Energy price data was obtained by state from the DOE’s Consumption, Prices, and Expenditures estimates, which are published annually in the EIA’s State Energy Data System (SEDS). A state’s annual average electricity price is measured in cents per kWh. An a-priori association of this price with renewable capacity ratio is hard to establish. It may be negatively associated with the renewable capacity ratio: the higher the current price of electricity, the less likely is that consumers would want further increases in their electricity bills due to the inclusion of more expensive renewables, which would in turn make it less likely that state utilities will invest in renewable technologies. On the other hand, electricity prices may be positively correlated with renewable capacity ratio: a higher price of electricity has the potential to make renewable energy more economically feasible and therefore competitive. A state’s annual average natural gas price (in 2007 dollars) per million Btus (ngprice) paid by the electric power industry measures the impact the cost of natural gas has on renewable capacity in a state. If renewable capacity is a substitute for natural gas 17 We have made an implicit assumption that these variables are independent of the policy variables. This, in fact, seems to be true given that none of the pairwise correlations between the policy and non-policy variables are even moderately high.
Author's personal copy 4734
G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
Fig. 5. Policy adoption in states. sumdummy: number of policies in a state.
Table 3 Correlation between Policy Variables.
cef rpscreq rpssreq rpssgoal sggpp rgpo
cef
rpscreq
rpssreq
rpssgoal
sggpp
rgpo
1.0000 0.0959 0.3292 0.1220 0.0361 0.0574
1.0000 0.0325 0.1841 0.0241 0.3761
1.0000 0.0243 0.6988 0.0188
1.0000 0.0019 0.1021
1.0000 0.0287
1.0000
natural gas prices for all states excluding Hawaii. The industrial price paid for natural gas is used for Hawaii, which is highly correlated with electric power natural gas prices across the US. There remain missing observations for Hawaii for 1996 and 1997. We also included a variable, pcgdp, to account for wealth effects. This was extracted from Bureau of Economic Analysis Regional Economic Information System (REIS). Consistent with other policy analysis (Ringquist, 1994; Sapat, 2004), we predict that states with greater wealth, other things being equal, will have a higher percentage of renewable energy because they will have the ability to invest more heavily in renewable energy.
Table 4 Control variables. Control variable
Mean
Std. Dev.
Min
Max
Median
gen (TWh) elecprice (2007 $/kW h) ngprice (2007 $/MMBtu) pcgdp (2007 $/capita) pctcoal (%) lcvscore (0–100)
152.59 33.01 5.27 31.28 38.17 43.25
136.78 9.28 2.42 6.79 27.15 26.61
9.88 18.82 1.75 18.09 0.00 0.00
810.99 73.41 19.02 58.79 98.36 100.0
104.22 30.15 4.63 30.62 38.87 38.50
capacity, higher natural gas prices will lead to more renewable capacity as utilities shift production to the relatively cheaper renewable capacity. Natural gas cost data can be found on the EIA website in the Natural Gas Navigator section. The data is deflated using the Consumer Price Index for all goods from the Federal Reserve Bank of St. Louis (FED, 2004). A simple average of natural gas costs of the surrounding states is used as an estimate for
4.3.2. Political variables Coal capacity is primarily used as baseload capacity due to its low operating costs. Coal is the dirtiest fuel used in electricity production and leads to large amounts of pollutant emissions: sulfur dioxide emissions that increase acid rain, nitrogen oxides that increase smog, and carbon dioxide emissions that contribute to climate change. If a state is environmentally conscious, larger amounts of coal capacity should encourage more renewable capacity deployment to decrease the average emissions levels in the state’s fuel mix, lower the negative impacts from electric power emissions, and prepare the state for climate change policy. However, a higher percentage of coal (pctcoal) would represent the strength of coal-based interest groups, and will likely to have negative association with renewable energy deployment. This data was extracted from EIA State Energy Data System (Energy Information Administration, 2007).
Author's personal copy G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
A political variable is included to measure changes in renewable energy preferences in a state. The League of Conservation Voters (LCV) rating is used to determine if policy preferences for environmental protection increase renewable energy capacity independent from its policy effects. Data from the ‘‘National Environmental Scorecard’’ is available from the League of Conservation Voters (2004).18 An average of all the votes by a state’s representatives is taken to get the average House of Representatives score (lcvscore). The scores from the House of Representatives are used instead of the Senate because representatives have a shorter term in office than senators, two years versus six years. The shorter term creates greater pressure on representatives to act according to their constituents’ preferences. A high LCV rating for a state indicates that the state’s constituents are environmentally friendly and are more likely to demand electricity from renewable energy, all other things being equal.19 Policies may be endogenous to higher LCV ratings because states with congresspersons who vote for federal pro-environmental policies may be more likely to enact state pro-environmental policies. The policy endogeneity issue is not addressed in the body of this paper because (a) lagging the variable by one year decreases these concerns, (b) lcvscore is not a strong enough predictor of state policies to be a satisfactory instrument, and (c) removing lcvscore from the regression does not change the results.
5. Results A Wooldridge (2002) test for auto-correlation in panel data could not reject the presence of serial correlated errors in the fixed effect models. Moreover, a modified Wald test for groupwise heteroskeadasticity could not reject the presence of heteroskedastic errors in the fixed-effects models. Consequently, a Prais–Winston (PW) estimation with Panel Corrected Standard Errors (PCSE) was applied. Table 5 reports the regression results for both the state fixedeffects model with state time-trends. The results are summarized below by each of the five dependent variables: total non-hydro renewable capacity ratio in Specification 1, wind capacity ratio in Specification 2, biomass capacity ratio in Specification 3, geothermal capacity ratio in Specification 4, and solar capacity ratio in Specification 5. 5.1. Total non-hydro renewable capacity 5.1.1. Regulatory policy variables: Rt The results were the outcome of the multi-variable regression as outlined in Section 3.6. These results were further verified by running the regressions using only one policy variable at a time. The significance of each of these policy variables did not change due to these single policy regressions. Renewable portfolio standard with sales requirement (rpssreq), required green power option (rgpo), and clean energy fund (cef) were found to have statistically significant effects on total nonhydro renewable capacity in the electric power industry; whereas RPS with capacity requirement (rpscreq), RPS with sales goal (rpssgoal), and state green power purchase requirement (sggpp) were found to be insignificant. However, though we would expect the impact of rpssreq on the renewable capacity ratio (renratio) as positive, it actually has 18 The LCV rating has been used in prior studies, including Baldwin and Magee (2000), Kalt and Zupan (1984), and Nelson (2002). 19 A related hypothesis could be that the LCV score would be correlated with the take up of voluntary RPS goals. However, the pairwise correlation between lcvscore and rpssgoal was found to be low (0.1313).
4735
Table 5 Regressions results. Variables
(1) (2) Total renew Wind
rpscreq
0.014 (0.031) 0.475nnn (0.147) 0.071 (0.135) 0.013 (0.088) 1.529nnn (0.315) 1.725nnn (0.532) 0.057n (0.032) 0.072nn (0.029) 0.066 (0.101) 0.001 (0.015) 0.003 (0.004)
rpssreq rpssgoal sggpp rgpo cef elecprice ngprice pcgdp pctcoal lcvscore
Observations 598 R-Squared 0.9531 Wald test w2 ð172Þ 17730
(3) Biomass
(4) (5) Geothermal Solar
0.046nn (0.023) 0.009 (0.009) 0.031 (0.054) 0.008n (0.004) 1.251nnn (0.275) 0.357nnn (0.129)
0.026 (0.020) 0.509nnn (0.138) 0.023 (0.128) 0.012 (0.085) 0.332nn (0.135) 1.264nnn (0.446)
0.005n (0.002) 0.009nnn (0.003) 0.005 (0.006) 0.000 (0.001) 0.026 (0.019) 0.140n (0.085)
0.000 (0.000) 0.001nn (0.000) 0.000 (0.001) 0.000 (0.000) 0.011nn (0.005) 0.020 (0.019)
0.008 (0.006) 0.033nn (0.015) 0.020 (0.018) 0.001 (0.008) 0.000 (0.001)
0.071nn (0.030) 0.031 (0.022) 0.040 (0.034) 0.008 (0.012) 0.003 (0.003)
0.001 (0.002) 0.006 (0.004) 0.004 (0.005) 0.006n (0.003) 0.000 (0.000)
0.000 (0.000) 0.000 (0.001) 0.001 0.002 0.000 (0.000) 0.000 (0.000)
598 0.9262 7695
598 0.9571 22856
598 0.9890 7875
598 0.9632 7501
Robust standard errors in parentheses;n significant at 10%;nn significant at 5%;nnn significant at 1%. State and time dummy variable coefficients are not reported.
a counter-intuitive negative impact: our results show that every 1% increase in the RPS sales requirement rpssreq results in a 0.475nG MW decrease in the capacity of renewable energy, where G denotes the total electricity generation in the state. Though this counter-intuitive result (i.e., the negative impact) is not new, and has been previously discovered and explored (Michaels, 2007; Yin and Powers, 2010), in Section 6 we discuss the reasons behind this in detail. As an example, it is illustrative to look at Fig. 6 which shows how rpssreq directly influences the renewable capacity ratio in Maine. In particular, the drop in the renewable energy ratio in the period 1998–2002 is directly correlated with the rpssreq variable.20 Furthermore, the coefficients on other RPS related policy variables (rpscreq and rpssgoal) are not significant. On the other hand, the ineffectiveness of rpssgoal is expected because there is no real pressure on the utilities in increasing the renewable energy ratio and, in absence of an regulation for incorporating more expensive renewable energy, utilities would not take any action. That is, voluntary goals where state governments assume in ‘‘good faith’’ that the electric power industry will take action, do not seem to work, as expected. cef, which measures the amount of capacity that the fund has agreed to help finance, has a statistically significant coefficient. This includes capacity that has been agreed upon, but has not yet been built, either because the project has not been finished or the project is later canceled. This is an expected result because over 42% of all capacity funded through CEFs had come online as of 2007. The most interesting result is the effect that required green power options have on renewable capacity. The coefficient for rgpo has a positive and statistically significant coefficient. As an
20 Note that the drop in the renewable energy ratio in Maine is due to the fact that during 1998–2002 the renewable capacity remained largely unchanged but total capacity increased by over 50%.
Author's personal copy G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
rpssalessizeint
rencapperc
4736
40 35 30 25 20 15 1990
1995
2000 year
2005
2010
1990
1995
2000 year
2005
2010
1990
1995
2000
2005
2010
30 20 10 0
reqgpopt
1
−1 year
reqgpopt
rpssalessizeint
rencapperc
Fig. 6. RPS sales requirements driving renewable penetration in Maine. rencapperc: ratio of renewable capacity to generation; rpssalessizeint: RPS sales requirement; reqgpopt: required green power purchase option.
15 10 5 0 1990
1995
2000 year
2005
2010
1990
1995
2000 year
2005
2010
1990
1995
2000 year
2005
2010
1
−1
1 .8 .6 .4 .2 0
Fig. 7. Required green power option driving renewable penetration in Iowa rencapperc: ratio of renewable capacity to generation; rpssalessizeint: RPS sales requirement; reqgpopt: required green power purchase option.
example, Fig. 7 shows how rgpo directly influenced the share of renewable capacity in Iowa. The statistically and economically significant increase implies that required green power options, which create a niche market for green power by requiring states to offer renewable-based electricity to their consumers at a premium, are very useful in increasing the amount of renewable energy capacity in a state. These impacts may persist or even increase over time because participation rates continue to increase across the country, with some local participation rates as high as 20% as of 2007. Finally, the coefficient for sggpp is statistically insignificant. State government purchases of green power are not large enough to significantly increase the total non-hydro renewable capacity ratio in a state.
5.1.2. Economic and political variables: Wt The only two statistically significant coefficients are the economic variables: price of electricity (elecprice) and natural gas price (ngprice). The coefficient of electricity price is positive. This is consistent with the expectation that as price of electricity goes higher, electricity generated from renewables becomes more cost-competitive, resulting in higher deployment of renewable capacity. In addition, given that renewable energy is the closest substitute (in economic terms) for natural gas base power plants, one would expect this relationship to be positive. However, this coefficient has the opposite sign (negative as opposed to positive) than would be expected if substitution effects were driving renewable capacity ratio. The absence of a positive coefficient implies that renewable capacity does not appear to be competing
Author's personal copy G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
with natural gas as a substitute. In what follows, we provide some potential explanations for why we get this counter-intuitive result. One potential reason is that, they are actually complements. The flexible natural gas based plants are used for overcoming the intermittency issues inherent in renewable power generation—in particular wind, the dominant renewable source. In absence of wind generation a gas plant can be employed in a quick manner to support demand. If gas is cheaper, the intermittency of wind base loads is not as big of an issue because the cost of covering the demand on low wind days is lower. Thus, lower natural gas prices would decrease the negative impact of intermittency and encourage deployment of wind power. Another potential reason for this result could be that natural gas prices are negatively correlated with economic activity (Awerbuch, 2003), which would typically be positively correlated with investments in renewables.21 Finally, the coefficient for per capita GDP (pcgdp) is statistically insignificant. Again, these point to the finding that, in general, economic factors do not seem to be driving the renewable capacity ratio so far. This becomes even more obvious as we look at individual types of renewable capacities in subsequent sub-sections. The coefficients on the political variables—lcvscore and pctcoal — are found to be statistically insignificant. These are somewhat surprising since one would expect the renewable capacity ratio to be higher/lower for a higher lcvscore/pctcoal. These reflect that finding that a state’s political environment is yet to impact the renewable capacity ratio.
5.2. Wind capacity Specification 2 estimates the policy and control variable impacts on wind capacity ratio. For wind capacity ratio, the following four policy variables are significant: clean energy funds (cef), RPS with capacity requirement (rpscreq), state green power purchasing requirement (sggpp), required green power option (rgpo), and natural gas price (ngprice). The rest of the variables are insignificant. That is, compared to the results for the renewable capacity ratio, cef and rgpo continue to be significant. These results for wind capacity ratio are similar to the ones for total renewable capacity ratio, except for two differences. First, instead of the RPS sales requirement (rpssreq) being negatively significant, RPS capacity requirement (rpscreq) is negatively significant. That is, though RPS requirements remain significant in driving the wind capacity ratio, the driving policy is capacity requirement and not sales requirement, which turns out to be insignificant. However, the negative result on rpscreq needs to be accepted with caution given the small numbers of states adopting rpscreq in the period of interest: in our sample, only three states have adopted rpscreq in the period of interest, with a total of four states with this policy. Second, state green power purchasing (sggpp) is negatively significant. Given that this variable is highly collinear with the RPS sales requirement (rpssreq), we ran another regression without sggpp. However, rpssreq remains insignificant. In addition, wind capacity ratio is the only capacity ratio that is affected by sggpp. Further investigation revealed that sggpp is highly collinear with the time trends in Maine—these time trends are also negatively significant in our regression. Confounding of this policy variable with the time trends, combined with the fact that this variable is marginally significant (P 4 jzj ¼ 0:09) suggests that this 21 To put this in context, consider the venture capital (VC) investment in the clean tech industry before and after the financial crisis of 2008. During the crisis, total VC investments in clean tech quickly dropped from $8.5 billion in 2008 to $5.5 billion in 2009 (Clean Tech Group, 2010).
4737
result may not be reliable, and should be accepted with appropriate skepticism. These differences (in particular, for rpssreq) from the result for the renewable capacity ratio are somewhat surprising since 91% of new renewable capacity since 1996 has been wind power capacity (in fact, these two capacities show a correlation on 0.85 in our data), and one would expect the results for wind capacity ratio to be very similar to the ones for the total renewable capacity ratio. Some explanation lies in looking at the actual coefficients from the regressions. For example, for the total renewable capacity, the coefficient of rpssreq is actually driven by the changes in biomass capacity (and not wind) due to the same variable: these coefficients are 0.475 and 0.509 (Table 5), respectively. 5.3. Biomass capacity This regression uses an extra control variable. The total tons of sugarcane production (sgcaneprod) is included in the model estimating biomass capacity. This variable was extracted from USDA Sugarcane Production Database. Florida, Hawaii, Louisiana, and Texas use the byproduct of sugar production from sugarcane as a biomass fuel. For example, in Hawaii sugarcane is one of the primary sources of biomass. Due to market conditions most of the sugarcane farms in Hawaii were shut down over the 1990s, removing the fuel source for much of the biomass capacity in the state. Changes in sugarcane production are likely to have an impact on the amount of biomass capacity in a state. Specification 3 estimates the policy and control variable impacts on biomass capacity ratio. The following variables are significant: clean energy funds (cef), RPS with sales requirements (rpssreq), required green power option (rgpo), and electricity price (elecprice). The rest of the variables are insignificant. That is, compared to the results for the renewable capacity ratio, cef, rpssreq (negative significance) and rgpo continue to be significant. These results for biomass capacity ratio are similar to the ones for total renewable capacity ratio, except for one difference. For biomass capacity ratio, the price of electricity (elecprice) is positively significant; whereas the natural gas price (ngprice) is not significant. This suggests that biomass power does seem to act as a substitute for grid-tied electricity. However, in our analysis, biomass power capacity ratio is the only dependent variable affected by the price of electricity. This may have to do with two facts: first, biomass power is one of the least expensive renewable energy sources (EIA, 2011); and second, biomass power lends easily to cogeneration. Thus, rising electricity prices may lead to increased cogeneration and substitution of conventional electricity with biomass power. 5.4. Geothermal capacity Specification 4 estimates the policy and control variable impacts on geothermal capacity ratio. The following variables are significant: clean energy funds (cef), RPS with sales requirement (rpssreq), RPS with capacity requirement (rpscreq), and percentage of coal generation (pctcoal). The rest of the variables are insignificant. That is, compared to the results for the renewable capacity ratio, cef and rpssreq (however, now with a positive significance) continue to be significant. These results for geothermal capacity ratio are similar to the ones for total renewable capacity ratio, except for the following differences. First, both the RPS requirements—sales and capacity—are now positively significant. This is different from the previous results—i.e., from total renewable capacity ratio, wind capacity ratio, and biomass power capacity ratio. Given that geothermal energy accounts for a small fraction of the renewable capacity, this
Author's personal copy 4738
G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
suggests that, though RPS requirements have resulted in a (counter-intuitive) negative impact on not only the overall renewable capacity ratio but also the major contributors (wind and biomass) due to the reasons outlined in Section 5.1, they have resulted in promoting sources with smaller contribution to overall renewable capacity—an expected result. We would expect to see a similar result for the solar capacity ratio as confirmed below. Second, geothermal capacity ratio is not dependent on any of the economic variables. However, it is negatively correlated on the share of electricity generated by coal (pctcoal). That is, geothermal capacity ratio is lower for states with a higher share of coal-generated electricity. Though we do not verify this, we provide a couple of hypothesis on why this may be happening. A higher share of coal-generated electricity is typically correlated with states with higher amount of coal reserves, and this finding may be due to the competition for similar resources (underground space as well as miners) required for geothermal as well as coal. Another potential reason may be that higher amount of coal reserves may be negatively correlated with the amount of geothermal reserves, and since generation from a particular resource would typically be positively correlated with reserves from the resource, we would expect to see the negative dependence between coal generation and geothermal capacity ratio. 5.5. Solar capacity Specification 5 estimates the policy and control variable impacts on solar capacity ratio. The following variables are significant: RPS with sales requirement (rpssreq) and required green power option (rgpo). The rest of the variables are insignificant. That is, compared to the results for the renewable capacity ratio, rpssreq (however, now with a positive significance) and rgpo continue to be significant. These results for solar capacity ratio are similar to the ones for total renewable capacity ratio, except for the following differences. First, solar capacity ratio is the only dependent variable that is not dependent on cef. This points to the fact that solar is the most expensive technology and may not be eligible for cef, which may be typically awarded to more economic options, such as wind and biomass. We discuss this in more detail in Section 6. Second, similar to the geothermal capacity ratio, the solar capacity ratio is positively influenced by the rpssreq. The reasoning for this would follow along the same lines: solar capacity contributes very little to the overall energy mix and, the small market is positively influenced by any RPS requirement. Another reason for this may be the presence of a solar specific requirement in many state RPS (Wiser and Barbose, 2010). This requirement would drive solar adoption up from near zero even if the overall renewable capacity ratio goes down. Third, the solar capacity ratio is not influenced by any of the economic and political variables. This lends support to the belief that solar deployment is primarily driven by policy, and it would not be able to sustain itself in the absence of policy.
6. Discussion The general insignificance of the economic and political variables points towards an important conclusion. The renewable capacity ratio is not influenced significantly by regular economic and political factors. Further, given that, in general, the key significant variables are policy variables, it is clear that the renewable capacity penetration is, to a large extent, driven by policy: a key finding of our study! One reason for this may be that, in absence of policy, renewables are not cost-competitive, and may not be able to gain a significant share of the market. In fact, the share of renewables is still small—less than 5% (EIA, 2011).
6.1. Renewable portfolio standard RPS requirements are always significant in one form or another— either sales or capacity requirements are always significant. However, a key finding of our study is that the impact of these requirement is dependent on the kind of renewable energy installed. On one hand, a surprising finding is that the impact of RPS requirements on total renewable capacity ratio is negative. Further, the impact on major contributors to total renewable capacity—wind and biomass—is also negative. The negative significance of RPS requirements for some renewable sources, though surprising, has been previously discovered in (Michaels, 2007), and explored in detail in Yin and Powers (2010). This may very well have to do with not only pitfalls in policy designs but also related interpretations (Wiser et al., 2005; Yin and Powers, 2010). In particular, Yin and Powers (2010) point out that this result is due to the following factors. First, many state RPS (e.g., Maine and New Mexico) include generation expected from existing capacity that may be eligible under the policy. This may weaken the incentive to promote new renewable capacity. That is, if a state introduces an RPS that is already being met, and new non-renewable capacity is being added in the state, causing the renewable capacity to drop, we could actually see a negative impact. A potential reason for this finding may be the underlying empirical approach. Our work, similar to Yin and Powers (2010), uses a panel regression approach which measures the inter-state as well as inter-temporal effects. The inter-temporal effects may be negative, and may overwhelm the potentially positive inter-state effects, which would make the overall impact negative. In fact, this can be verified by conducting a simple experiment. Fig. 6 (Section 5.1) shows how the drop in renewable capacity ratio in Maine is directly correlated with the RPS sales requirement. One hypothesis is that this negative effect is so strong that it overcomes the expected positive relationship. This is indeed the case: when we exclude Maine and rerun our regression for renratio, the coefficient for rpssreq turns positive. Second, our model ignores various exemptions from the RPS—in particular, related to coverage. For example, RPS in many states (e.g., Montana and Arizona) do not cover the entire electricity market. This again weakens the RPS requirement seen by the state as a whole. In fact, Yin and Powers (2010) explicitly account for these RPS design features (i.e., existing capacity, coverage, etc.) that may have a significant impact on the strength of an RPS, creates an incremental RPS requirement variable that is related to nominal RPS requirements, and shows that the overall impact of this incremental variable is positive, as expected from RPS. This finding also suggests that we need to further investigate the impact of inter-state policies, such as RPS that allow interstate trading of renewable energy credits (RECs) that may impact in-state capacity. On the other hand, the impact of RPS requirements on the minor contributors—geothermal and solar—is positive. This separation of the impact of RPS requirements on major and minor contributors is interesting since it points to the fact that the same policy can have different impacts on the penetration of a renewable energy source, based on the share of that source in the overall mix. In this context, when the share is high, RPS requirements can suppress further deployment of the renewable energy source due to the reasons identified in Yin and Powers (2010); whereas, when the share is low, the same RPS requirements can encourage further deployment of the renewable energy source. It is not clear why the latter effect occurs. Though this requires further work, one potential reason may be the presence of other policies that specifically encourage the adoption of different kinds of renewables, such as solar carve-outs within the RPS requirements that stipulate that a certain fraction of the RPS requirement
Author's personal copy G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
be fulfilled by solar energy (Wiser and Barbose, 2010).22 Another potential reason could be that any absolute increase on a small base (e.g., for solar) would look big in relative terms, whereas the same absolute increase on a larger base (e.g., for wind) may be hard to detect. However, another form of RPS, the voluntary renewable portfolio standard policy—RPS goals—is not found to be significant. This is not surprising, and confirms the belief that expecting utilities to act in ‘‘good faith’’ is not an effective way to promote renewable energy. Given a choice, in the absence of command and control regulation, utilities will always opt to install cheaper, nonrenewable energy sources, and stricter measures are required to ensure increased penetration of renewable energy sources. Thus, a policy implication is that states should not try to increase the share of renewable energy capacity by using RPS goals, and any serious effort has to be in the form of an RPS requirement. 6.2. Clean energy fund A key contribution of our work is that clean energy funds are mostly found to be positively significant except for solar capacity ratio, where they are found insignificant. This result is to be expected, given the fact that these funds reduce one of the main barriers—capital expenditure—for renewable energy deployment (Beck and Martinot, 2004). Further, these funds seem to support less expensive sources over expensive ones, perhaps given the lower perceived risk associated with the more competitive technologies. One potential reason for the insignificance related to solar is that, given that the EIA database does not contain small-scale distributed projects (Energy Information Administration, 2007), our dependent variable is really looking at large scale deployment, whereas a good chunk of clean energy funds targeted towards solar is used to subsidize small-scale installations or to support manufacturing (Bolinger et al., 2001, 2004; Bolinger and Wiser, 2006). This empirical result lends credibility to case studies (Bolinger et al., 2001, 2004, 2005) that have claimed that clean energy funds have been effective based on the fact that hundreds of megawatts have been installed using CEF financing. However, this result contradicts previous econometric studies (Menz and Vachon, 2006; Yin and Powers, 2010) that a related policy, public benefit funds, is not effective in increasing renewable (wind) capacity. This difference may be attributed to the focus of the previous literature (Menz and Vachon, 2006; Yin and Powers, 2010) on public benefit funds, which have a broader applicability, due to the focus on energy efficiency in addition to renewable energy, as opposed to clean energy funds, which focus on only renewable energy. Further, these papers (Menz and Vachon, 2006; Yin and Powers, 2010) do not allow for the four year delay between funds being awarded and the installation of capacity. This delay is typical, and ignoring it may result in the policy showing up as insignificant. 6.3. State government green power purchasing The state green power purchasing requirements are mostly found to be insignificant except for wind capacity ratio, where these are found negatively significant. As mentioned earlier (Section 5.2), the negative significance for wind capacity ratio may be discounted since the state green power purchasing 22
We have not been able to verify this hypothesis as of yet. A preliminary regression that included a dummy variable representing the solar carve out requirement indicated that the solar carve out dummy was insignificant. However, it must be noted that the same dummy variable was found significant in another study (Sarzynski et al., submitted for publication).
4739
agreements are highly correlated with the time trends in Maine, which confounds the results given the collinearity. The overall result—insignificance of this policy—is somewhat surprising since one would expect green power purchases by states to result in increased penetration of renewable energy, as claimed by Collison (2010). Though it has not been possible to pinpoint the exact causes for why this may be happening, a couple of potential reasons may be as follows. This may happen either due to these requirements representing very small fractions of overall renewable energy growth or due to these requirements not being strictly followed. 6.4. Required green power option Finally, the required green power option was found to be mostly positively significant, except for geothermal. The robustness of this influence across all renewable categories is surprising, given that this requirement does not set any requirement on the share of the electricity that needs to come from these programs. Forcing utilities to offer customers the option to purchase renewable-based electricity at a ‘‘reasonable’’ premium rate increases renewable capacity in a state by up to 2% of total capacity. The policy has a greater impact in larger electricity markets and appears to be effective regardless of a state’s political environment or other policy implementations. This indicates the power of ‘‘customer pull’’, which shows that, given a choice, a certain fraction of customers would opt for green power, even if it is expensive. This is classic market segmentation at work (Dickson and Ginter, 1987; Bird et al., 2008)—providing customers with a differentiated product allows us to measure the demand for such a product. Another advantage of this policy is that the renewable energy customer pays for the additional cost of renewable energy, and this cost does not need to be socialized,23 thus removing any market distortions associated with integration of renewable energy into the overall energy mix. Another way to look at this is that the green customers are cross-subsidizing the society—especially if green house gas emissions are not priced. There are major renewable policy implications if these results hold or strengthen over time due to increasing green power program participation rates. Only seven states had implemented required green power options as of 2007 even though creating a statewide ‘‘green power’’ market appears to be effective at increasing renewable energy capacity. It will be important to determine if these impacts persist over time as the supply of renewable energy catches up with current market demand—for example, there may be latent demand in states that do not yet offer this option.
7. Conclusions States have enacted many policies to increase the deployment of non-hydro renewable capacity into the electric power industry in that state. This study adds to the existing literature on evaluating the impact of policy on penetration of renewable energy capacity. It utilizes a state fixed-effects model with a larger panel to explain the deployment of total non-hydro renewable capacity as well as the four different types of non-hydro renewable energy (wind, biomass, geothermal, and solar) capacity in a state. 23 If the renewable energy customer does not pay for the higher cost of renewable energy, this higher cost is passed on to all the customers through increased electricity bills. As an example, consider a 90–10% mix of conventional and renewable energy, with corresponding costs 5 and 15 c/kW h, respectively. Assuming that consumers are priced at cost, this would mean an average price of 6 c/kW h if the cost of renewable energy is passed on to all the customers.
Author's personal copy 4740
G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
We find that policy plays an important role in changing the penetration of renewables, and sometimes can have unintended—in particular, even opposite—effects. The important policy implications that arise from the results indicate policymakers have an array of effective tools—command-and-control, marketbased, and incentive-based—at their disposal to promote renewable energy deployment to meet environmental and energy security policy goals. These policy approaches can be used in conjunction to allow power producers to take advantage of green power market differentiation while encouraging market penetration through new requirements on non-green power market electricity capacity or sales. These policy mechanisms will become even more useful to state governments if the prospect of US climate change/carbon emissions policy becomes a reality. We note that the focus of this paper is on evaluating the effectiveness of policies with respect to deployment of renewable energy sources. There are many avenues to expand on this work. The first avenue is to extend our work in two different ways: (a) diagnosing the observed effects better in the context of political economy and institutional characteristics, and (b) characterizing policy design and implementation best practices. In particular, it would be appropriate to explore the results on RPS further by incorporating the RPS design features as identified in Wiser et al. (2005), and not only reconciling our results with Yin and Powers (2010) but also extending them by examining the inter-state effects. In addition, it would be appropriate to not only identify the statistical significance of policies but also the actual impact of policies (e.g., of required green power options) and comparing them with expected impacts. The second avenue is to translate our results into a more refined effectiveness framework. One dimension of this would be looking into the environmental effectiveness of these policies. That is, it would be desirable to measure the effectiveness of policies in reducing green house gas emissions. This would require using the renewable generation as the dependent variable, and translating the renewable generation to green house gas abatement (in tons) based on the appropriate business as usual trajectory (e.g., underlying average fuel mix or alternative new build). Another dimension of this would be looking into the cost effectiveness of these policies. That is, it would be desirable to measure the relative cost (in $/ton) of these policies. This would require estimation of the actual cost of the policies in addition to the green house gas abatement—a much harder exercise.
Acknowledgments We would like to thank Kath Rowley at Climate Policy Initiative and Andrea Sarzynski at George Washington Institute of Public Policy for providing detailed, insightful comments as well as for suggesting avenues for improvement—in particular, for making the paper not only more rigorous but also better suited for a policy audience.
Appendix: Data sources
Data source
Data description
EIA Historical State Electricity Databases EIA Natural Gas Price Navigator BEA Regional Economic Information System DSIRE
Generation, sales, and capacity Delivered natural gas prices State gross domestic product and population Policy design and enactment/ effective dates
CESA Utility-Scale Ren. Energy Project number/size Proj. Database USDA Sugarcane production Annual tons of sugarcane production Fed. Reserve Bank of St. Louis National avg. annual price CPI inflation LCV National Environmental Congress. voting ratings Scorecard
References Awerbuch, S., 2003. Determining the real cost: why renewable power is more competitive than previously believed. Renewable Energy World 6 (2). Baldwin, R.E., Magee, C.S., 2000. Is trade policy for sale? Congressional voting on recent trade bills. Public Choice 105, 79–101. Beck, N., Katz, J., 1995. What to do (and not to do) with time series cross section data. The American Political Science Review 89 (3), 634–647. Beck, N., 2001. Time-series-cross-section data: what have we learned in the last few years? Annual Review of Political Science 4, 271–293. Beck, F., Martinot, E., 2004. Renewable Energy Policies and Barriers. Encyclopedia of Energy. Bird, L., Bolinger, M., Gagliano, T., Wiser, R., Brown, M., 2005. Policies and market factors driving wind power development in the United States. Energy Policy 33 (11), 1397–1407. Bird, L., Kreycik, C., Friedman, B., 2008. Green Power Marketing in the United States: A Status Report (11th ed.). National Renewable Energy Laboratory. NREL/TP 6A2-44094. Bird, L., Lokey, E., 2007. Interaction of Compliance and Voluntary Renewable Energy Markets. National Renewable Energy Laboratory. NREL/TP 670-42096. Bolinger, M., Wiser, R., 2006. The impact of state clean energy fund support for utility-scale renewable energy projects. Lawrence Berkeley National Laboratory and Clean Energy States Alliance. Bolinger, M., Wiser, R., Milford, L., Stoddard, M., Porter, K., 2001. Clean energy funds: an overview of state support for renewable energy. Energy Efficiency and Renewable Energy, Office of Power Technologies, Department of Energy. Bolinger, M., Wiser, R., Fitzgerald, G., 2004. The impact of state clean energy fund support for utility-scale renewable energy projects. Lawrence Berkeley National Laboratory and Clean Energy States Alliance. Bolinger, M., Wiser, R., Fitzgerald, G., 2005. An overview of investments by state renewable energy funds in large-scale renewable generation projects. The Electricity Journal 18 (1), 78–84. Brown, E., Busche, S., 2008. State of the States 2008: Renewable Energy Development and the Role of Policy. National Renewable Energy Laboratory. NREL/ TP-670-43021. Carley, S., 2009. State renewable energy electricity policies: an empirical evaluation of effectiveness. Energy Policy 37, 3071–3081. Clean Tech Group, 2010. Clean Tech Venture Investment Totaled $5.6 Billion in 2009. Available at /http://cleantech.com/about/pressreleases/20090106. cfmS. Collison, B., 2010. State Leadership in the Procurement of Green Power. US EPA presentation. Available at /http://www.nga.org/Files/pdf/1008GREENPOWER WEBINARCOLLISON.PDFS. Dickson, P., Ginter, J., 1987. Market segmentation, product differentiation, and marketing strategy. Journal of Marketing 51, 1–10. DSIRE, 2011. Database of State Incentives for Renewable Energy: State renewable energy policy summaries. Website /http://www.dsireusa.orgS, accessed March 2004–March 2006. Energy Information Administration, 2007. EIA Historical State Electricity Databases and FERC Form-423 Database. Website /http://www.eia.doe.gov/fuel electric.htmlS, accessed March 2004- March 2007. Energy Information Administration, 2009a. Annual Energy Review 2009. Report DOE/EIA-0384(2009). Website /www.eia.doe.gov/emeu/aer/contents.htmlS, accessed February 2011. Energy Information Administration, 2009b. Emissions of Greenhouse Gases in the United States 2008. EIA Report DOE/EIA-0573, available at /http://www.eia. doe.gov/oiaf/1605/ggrpt/pdf/0573(2008).pdfS, accessed 11/28/2011. EIA, 2011. Annual Energy Outlook 2011 with Projections to 2035. Available at /http://www.eia.doe.gov/oiaf/aeo/electricity_generation.htmlS. Engel, K.H., Orbach, B.Y., 2008. Micro-motives for state and local climate change initiatives. Harvard Law and Policy Review 2, 119–137. FED, 2004. Federal Reserve Branch of St. Louis. CPI index data. Website /http:// stlouisfed.orgS, accessed December 2004. Friedberg, L., 1998. Did unilateral divorce law raise divorce rates? American Economic Review 88 (3), 608–627. GDS Associates, Inc., 2001. Analysis of Renewables Portfolio Standard Options for Hawaii. The State of Hawaii Department of Business, Economic Development, and Tourism. Kalt, J.P., Zupan, M.A., 1984. Capture and ideology in the economic theory of politics. The American Economic Review 74 (3), 279–300.
Author's personal copy G. Shrimali, J. Kniefel / Energy Policy 39 (2011) 4726–4741
Langniss, O., Wiser, R., 2003. The renewables portfolio standard in Texas: an early assessment. Energy Policy 31 (6), 527–535. LBNL, 2006. Database of Utility-Scale Renewable Energy Projects. Lawrence Berkeley National Laboratory and Clean Energy State Alliance (CESA). Website /http://eetd.lbl.gov/ea/ems/cases/Large_Renewables_Database.xlsS, accessed December 2006. League of Conservation Voters, 2004. Description and Data from the LCV National Scorecard. Website /www.lcv.orgS; accessed March 2004. Menz, F.C., Vachon, S., 2006. The effectiveness of different policy regimes for promoting wind power: experiences from the states. Energy Policy 34, 1786–1796. Michaels, R.J., 2007. Intermittent Currents: The Failure of Renewable Electricity Requirements. Social Science Research Network, Mimeo. Morris, G., 2003. The Status of Biomass Power Generation in California. Subcontractor Report. National Renewable Energy Laboratory. Green Power Institute, Pacific Institute for Studies in Development, Environment, and Security. Nelson, J.P., 2002. ‘Green’ voting and ideology: LCV scores and roll-call voting in the U.S. senate, 1988–1998. The Review of Economics and Statistics 84 (3), 518–529. Parks, R., 1967. Efficient estimation of a system of regression equations when disturbances are both serially and contemporaneously correlated. Journal of the American Statistical Association 62, 500–509. Petersik, T., 2004. State Renewable Energy Requirements and Goals: Status Through 2003. Energy Information Association. Reed, W.R., Ye, H., 2011. Which panel data estimator should I used? Applied Economics 43, 985–1000.
4741
Ringquist, E.J., 1994. Policy influence and policy responsiveness in state pollution control. Policy Studies Journal 22 (1), 25–43. Sapat, A., 2004. Devolution and innovation: the adoption of state environmental policy innovations by administrative agencies. Public Administration Review 64 (2). Sarzynski, A., Larrieu, J., Shrimali, G. The impact of state financial incentives on market deployment of solar technology. Energy Policy, submitted for publication. Stimson, J.A., 1985. Regression in space and time: a statistical essay. American Journal of Political Science 29 (4), 914–947. Wasserman, A., 2010. WRI Fact Sheet: U.S. Climate Action in 2009–2010. World Resource Institute Fact Sheet, available at /http://www.wri.org/stories/2010/ 09/fact-sheet-us-climate-action-2009-2010S, accessed January 2011. Wiser, R., Porter, K., Grace, R., 2005. Evaluating experience with renewables portfolio standards in the United States. Mitigation and Adaptation Strategies for Global Change 10, 237–263. Wiser, R., Barbose, G., 2008. Renewable Portfolio Standards in the United States: A Status Report with Data Through 2007. Lawrence Berkeley National Laboratory Report LBNL-154E. Wiser, R., Barbose, G., 2010. Supporting Solar Power in Renewable Portfolio Standards: Experience from the United States. Lawrence Berkeley National Laboratory Report LBNL-3984E. Wooldridge, J., 2002. Econometric Analysis of Cross Section and Panel Data. MIT Press, Cambridge, MA. Yin, H., Powers, N., 2010. Do state renewable portfolio standards promote in-state renewable generation. Energy Policy 38, 1140–1149.