Electric Utilities and the Federal PTC for Renewable Energy

We first explore how variation in the federal political environment condition vested interests and then investigate the responses of these vested interests under different constellations of state policy. As articulated in Moe (Reference Moe2015), vested interests are fundamental to an understanding of the stability and change of political institutions, from the formal governmental structures to governmental programs. They can be sources of the stability, but also of the change. It is our attempt here to understand one of the mechanisms through which vested interests can influence the policy outcome. By investigating how utilities respond to the potential change in the federal program, we illustrate one mechanism through which vested interests influence the stability and change of government programs.

To keep the discussion focused, we begin with a brief introduction to the PTC, which was initially enacted on October 24, 1992, as part of Energy Policy Act (H.R.776.ENR) and set to expire on June 30, 1999. Since then, it has been repeatedly under threat of expiration with renewals effective for only one or two years at a time. With the exception of the initial expiration date in June 1999, the PTC has always been scheduled to expire at the end of the calendar year.

Fuel choice is one of the fundamental factors shaping electric utilities’ interests. Depending on the type of fuel in which utilities have invested for electricity generation, climate policies can result in significant benefits or losses. Specifically, by offering credit for electricity generation from wind and other eligible renewables, the PTC boosts their profitability. In 2014, wind, geothermal, and closed-loop biomass were eligible for a rebate of US$0.023 per kilowatt-hour, equivalent to around 15% of the average levelized cost of renewable energy generation in 2014 (US$0.148 per kilowatt-hour), while the rebate for other eligible technologies was US$0.011 per kilowatt-hour.^{Footnote 8} The rebates are available for generation during the first 10 years of a project.^{Footnote 9}

Studies show that the PTC has been a boon for wind energy in particular. In 2014, the National Renewable Energy Laboratory (Lantz et al. Reference Lantz, Steinberg, Mendelsohn, Zinaman, James, Porro, Hand, Mai, Logan, Heeter and Bird2014) conducted a modeling exercise. If the PTC expires without alternative federal-level support, wind electricity generation capacity would only grow by 3 to 5 gigawatts annually versus 5 to 15 gigawatts under the PTC. In other words, the PTC could increase wind energy deployment by up to three times.

For the purposes of lobbying, the timing of PTC expiration and modifications can be considered quasi-exogenous. The legislative history of the PTC shows that the typical PTC extension is one year, and the only exceptions have been instances of major legislative activity that have allowed pro-renewable supporters to demand longer extensions. As we detail below, in neither case is the timing of PTC expiration systematically related to lobbying activity though the outcome may well be.

The expiration of the PTC has historically generated intense lobbying activity. Consider, for example, the threat of expiration on December 31, 2012. On October 26, 2012, “[t]wo dozen investors representing over $800 billion in assets under management sent a letter to Congressional leaders today calling for an immediate extension.”^{Footnote 10} On December 12, 2012, the American Wind Energy Association (AWEA) published a press release^{Footnote 11} saying that

The lobbying is not limited to proponents of the PTC. As noted in the introduction, the electric utility Exelon, a major producer of nuclear power had to resign from AWEA because it lobbied against the extension of the PTC. On November 4, 2013, the conservative advocacy organization Americans for Prosperity sent a letter to the Senate saying that the

Given this logic, we expect the consideration of PTC renewal to increase the lobbying activity of electric utilities. Consider how the PTC influences the electricity demand from utilities and the supply from electricity producers in the wholesale electricity market. (For the graphical illustration of the changes of the demand and the supply, see Section A10 in the online appendix.) When the PTC is about to expire or under legislative debate, electric utilities have stronger incentives to lobby than at other times.

This logic is equally applicable to both expected winners and losers. On one hand, the consideration of PTC renewal makes lobbying more beneficial for expected winners, electric utilities that generate electricity from renewable sources, because the PTC provides rebates for their electricity generation. On the other hand, this would place utilities that do not rely on renewable sources at a comparative disadvantage in the wholesale electricity market. The PTC would increase the profitability of renewable electricity, which raises the supply of renewables and reduces electricity prices. While utilities that do not rely on renewable sources may benefit from reduced electricity prices at which they buy, utilities are in competition in the wholesale market and concerned about the differential effects of the PTC on different types of utilities. This logic is also consistent with why Exelon opposed the PTC for wind energy due to its differential effects on wind versus other energy types such as nuclear.^{Footnote 13} Due to this competition, utilities without any renewable energy generation would be relative losers from the PTC and would have increased incentive to lobby against it.

• Hypothesis 1: When PTC renewal is under consideration, electric utilities are more likely to lobby.

Federal and State Policies

A key contribution of this study is the characterization of the effects of state policy on federal lobbying. In a federal system, firms must consider the effects of state policy on incentives to lobby at the federal level. For renewable energy, the question is whether the presence of an RPS increases or decreases the propensity of lobbying when the PTC is about to expire. Though RPS policies comprise heterogeneous designs across states in various dimensions such as coverage, penalty, requirements for “new” investments in renewables, and enforced through a credit-trading mechanism, they all share a key feature: all RPS policies mandate electric utilities to meet a certain percentage of their total generation from renewable sources by purchasing electricity from renewable energy sources (Shrimali, Lynes, and Indvik Reference Shrimali, Lynes and Indvik2015; Wiser et al. Reference Wiser, Namovicz, Gielecki and Smith2007; Yin and Powers Reference Yin and Powers2010). The utilities that do not own or develop renewable energy assets typically meet this requirement by making long-term contracts with other independent renewable producers or by purchasing renewable energy credits. Without an RPS, both types of electric utilities have increased incentives to lobby on PTC when its renewable is under consideration, as outlined above. When RPS has been implemented in a state, both winners and losers have less incentive to lobby on PTC due to the substitution effects between federal-level PTC and state-level RPS.

We consider how the state-level adoption of an RPS affects the lobbying strategy of both winners and losers of the federal-level renewable energy policy. First, consider winners of the federal-level renewable energy policy—electric utilities with renewable electricity generation capabilities. Without RPS, these utilities have a clear incentive to lobby for the PTC extension because the PTC provides a subsidy for renewable energy generation, thereby reducing the net production cost of renewables. The utilities with renewable electricity generation capabilities in the states with RPS, however, have less incentives to lobby for the PTC extension compared with the ones in the states without RPS. These two policies clearly act as a substitute. Most RPSs use renewable energy certificates that are issued for the production of renewable electricity. The RPS requires electricity retailers to purchase certificates, which act as a subsidy to renewable energy generators who can sell the certificate credit at the market price (Fell, Linn, and Munnings Reference Fell, Linn and Munnings2012). As utilities in the RPS states already enjoy the benefits of state-level RPS that provides quasi-subsidy, they have less incentive to engage in costly lobbying on the PTC extension. Even if they lose benefits from the PTC, they would still enjoy from the RPS requirement.

Next, consider losers of the PTC—electric utilities without renewable energy generation capabilities. Utilities in the states without RPS have a clear incentive to oppose the PTC. The PTC decreases the production costs of renewable energy generation, which depresses electricity prices in general. While the reduction in prices is offset by subsidy for utilities with renewable energy generation, those utilities without renewable energy generation cannot enjoy the subsidy. Therefore, these utilities have incentives to engage in lobbying for the expiration of the PTC. However, in states with RPS, utilities expecting to have to purchase renewable electricity from others do not have any obvious additional incentive to lobby. These utilities now have to purchase renewable electricity under the RPS, and the subsidy from the PTC reduces their costs of doing so. Thus, although the market price for electricity decreases, the cost of meeting the RPS requirement also decreases. Thus, these utilities under the RPS system have both gains and losses from the PTC, so their incentives to lobby are not obvious. Given that engaging in lobbying is costly, we can expect that these utilities would not engage in lobbying unless they expect clear benefits or losses from the policy that will far offset the cost of lobbying.

Moreover, the PTC may even generate some benefits to previous losers, especially if they choose to become renewable electricity generators. If the PTC is sufficiently attractive to make renewable electricity generation profitable, some utilities may choose to actively lobby in favor of the continuation of the policy when under the threat of expiration. In this case, some of the previous losers, who would have lobbied against the PTC without the RPS, now lobby in support of it. Thus, the overall level of lobbying again remains unchanged.

This reasoning can be turned into a testable hypothesis by comparing how utilities affected and unaffected by a state RPS respond to the threat of PTC expiration. Given that a state RPS and the PTC tax credit are essentially substitutes, we hypothesize that utilities with production capacity covered by an RPS are less likely to mobilize in response to the threat of PTC expiration.

• Hypothesis 2: When PTC renewal is under consideration, electric utilities in states with stringent RPS policies are less likely to lobby than in states without.

Before we turn to discuss our research design, it should be noted that the two hypotheses are not mutually exclusive. While the first hypothesis examines the independent effects of the consideration of PTC renewal on the likelihood of lobbying, the second hypothesis focuses on how the size of such depends on the presence of state-level RPS. While we expect utilities (both expected winners and losers) on average to be more likely to engage in lobbying when PTC renewal is under consideration per our first hypothesis, such effects would be expected to be smaller for utilities operating in RPS states per our second hypothesis.

Dependent Variable

As noted above, the dependent variable is a binary indicator for lobbying by utility i at time t. To minimize concerns about the effects of barriers to entry in lobbying, such as upfront costs and returns to experience (Drutman Reference Drutman2015; Kerr, Lincoln, and Mishra Reference Kerr, Lincoln and Mishra2014), we first restrict our sample to the utilities that had engaged in lobbying at least once during the 1998–2012 period. Among 3,389 utilities in our original sample, 2,860 utilities (84.4%) did not engage in lobbying in the given period at all. We exclude these utilities from our sample in the first analysis and focus on the 529 utilities that had lobbied at least once. This analysis is interested in variation in lobbying propensity among utilities that are generally capable of entering the lobbying game. This approach allows us to focus solely on the timing of lobbying, but the results are robust if we consider all nonlobbying utilities by omitting utility fixed effects (Section A6 in the online appendix). The distribution of lobbying over time and the periods when the PTC was under legislative debate, is shown in Figure 1.

Figure 1. Share of electric utilities in the sample that lobby over time, overlaid on periods of legislative activity over the PTC.

Note. PTC = production tax credit.

Based on the lobbying disclosure data, we check the distribution of lobbying reports that fall into relevant lobbying categories: energy, environment, and utilities. The majority of lobbying reports are found to be related these three categories. The category of energy was the most frequently lobbied topics, with 44% of lobbying reports by utilities in the sample found to be related to energy. Next, 32% and 13% of lobbying are relevant to the issue of environment and energy, respectively. It should be also noted that lobbying reports are related to multiple issues: 6% of lobbying reports are related to all three categories, 22% have lobbied on two of three categories, and 31% have lobbied one category among three. Only 41% of lobbying reports did not choose any relevant category.

Explanatory Variable: PTC Activity

The first explanatory variable is an indicator for legislative activity around the impending expiration of the PTC at a given time. Since it was first enacted in 1992, the PTC has experienced only short-term and unpredictable extensions with occasional expirations. The legislative activities and outcomes are summarized in Table 1. We created a binary indicator to denote periods of legislative activity regarding the PTC, be it debate or impending expiration, coding those periods as 1. We coded periods of zero PTC legislative activity as 0. The exogeneity of the timing of PTC expiration and extension is an important identifying assumption for this study, and we will discuss it in depth below.

Table 1. Summary Table of Legislative Activities and Outcomes Regarding PTC Extension, 1999–2013.

Note. PTC = production tax credit.

Explanatory Variable: State Renewable Energy Policy

Our next explanatory variable is a measure of how much a utility is affected by the RPS. Our RPS data are from Shrimali, Lynes, et al. (Reference Shrimali, Chan, Jenner, Groba and Indvik2015). Of the various policy options to promote renewable energy resource development in individual states, the RPS is the most common and important. It requires energy suppliers to procure renewable energy or renewable energy credits to make up a certain percentage of their electricity sales or generating capacity. Figure 2 describes which states have implemented the RPS as of 2012 along with the location of utilities. The comparison of the two panels shows that the location of the utilities is not systematically related to the presence of RPS.

Figure 2. Implementation of RPS across the states in 2012 and the location of utilities: (a) Number of utilities operating in the states that had the RPS (as of 2012). (b) Number of utilities operating in the states that did not have the RPS (as of 2012).

Note. The upper panel (a) features the states that had the RPS in 2012. The lower panel (b) features the states that did not have the RPS in 2012. We rank the states from the one that has the least number of utilities to the one with the most number of utilities. States with more utilities are colored in darker purple. In calculating the number of utilities in a given state, we assigned weights according to the electricity generation in a given state. RPS = renewable portfolio standard.

To capture the extent to which utilities are influenced by state RPS policies, we utilize a simple binary indicator for RPS states. We calculate the share of electricity generation by each utility in those states in each time period. Because our utility data are at the plant level, we assign weights to different state policies across all utilities in our sample. For example, if utility i generates 30% of its total electricity in the states that adopted RPS regulations in period t, it is recorded as 0.3. The distribution of the utilities’ share of electricity generation in the RPS states is presented in Figure 3. The vast majority of utilities (85%) in our sample fall under the two extreme values of 0 and 1. This distribution suggests that using a continuous measurement of the extent that a utility is affected by RPS is problematic, as the distribution is neither continuous nor normal. We therefore construct a dichotomous variable, coded as 1 for a utility that operates only in the RPS states and 0 otherwise. As an alternative specification, we use the average share of electricity generation in the RPS states (0.238) as a cutoff point and coded 1 for utilities above that point and 0 for those below (Section A8 in the online appendix).

Figure 3. Histogram of the share of electricity generation in the RPS states.

Note. RPS = renewable portfolio standard.

The adoption of RPS over time is illustrated in Figure 4. The number of states that have adopted an RPS has sharply increased since 2000. In addition, the stringency of state RPS, measured by the incremental share indicator (ISI), which measures the incremental percentage requirement in renewable generation, increased over years (Shrimali, Chan, et al., Reference Shrimali, Chan, Jenner, Groba and Indvik2015; Yin and Powers Reference Yin and Powers2010). By adjusting the estimates for existing renewable electricity capacity, this measure captures the utilities’ need to add new renewable energy generation capacity to their portfolio (Shrimali, Chan, et al., Reference Shrimali, Chan, Jenner, Groba and Indvik2015). Specifically, we use the incremental requirement in wind energy generation mandated by an RPS policy. Along with the adoption of RPS, the average level of wind ISI has also increased.

Figure 4. Adoption of RPS policy by states and average wind ISI level over time, 1990–2012.

Note. RPS = renewable portfolio standard; ISI = incremental share indicator.

In our empirical analysis, we also estimate all the models using an indicator for wind ISI mandated by RPS policies to account for heterogeneity in RPS policies (Section A7 in the online appendix). We construct a dichotomous indicator for wind ISI as the distribution is, again, not continuous and significantly skewed to the right.^{Footnote 18}

Furthermore, we check whether the results remain robust when using continuous measure of RPS coverage instead of a binary specification of RPS (Section A8 in the online appendix). Specifically, we use a continuous indicator that captures yearly RES-E deployment requirement as a percent of total generation available from Shrimali, Lynes, and Indvik (Reference Shrimali, Lynes and Indvik2015). Again, we weighted this measure with regard to an utility’s share of electricity generation in the states where it operates.

Identifying Assumptions

Our identification strategy is based on the idea that lobbying by electric utilities changes with the quasi-exogenous timing of episodes of PTC consideration in the Congress. The change in lobbying is caused by the heightened salience of the PTC, but PTC activity does not change other incentives to lobby. The size of the effect, in turn, depends on whether the utility has capacity in RPS states. As long as RPS adoption at the state level is quasi-exogenous to PTC consideration, we can identify the interaction term of our regression model. Specifically, we impose three assumptions.

First, we assume that the timing of legislative PTC activity is quasi-exogenous to federal lobbying. This assumption can be validated by looking at the factors that determine whether the PTC is debated or not. As shown in Table 1 above, with the exception of the original expiration date of June 1999, the PTC has always been set to expire at the end of the budget year, a regular scheduling decision that is unrelated to federal lobbying. The most important exceptions to this pattern are shown in Section A1 in the online appendix. The exact timing in all these cases depends on the timing of much broader bills on the legislative agenda. It is unlikely that utilities can have such a short-term influence on the legislative agenda of the same quarter. The quarter during which the PTC was debated is thus not something a utility lobbyist can plausibly influence. Other than these measures, the only notable variation is in the length of each PTC period and whether or not the Congress managed to renew the PTC in December of year t or in January of year t + 1.

One threat to inference that we must consider is the possibility of anticipation effects: if utilities anticipate times of legislative debate, then they may begin to lobby earlier. To quantify the sign of any possible bias, such lobbying would induce a downward bias in our estimates of the importance of PTC consideration for lobbying as some of that lobbying would occur in anticipation.

To assess this possibility of a downward bias, Figure 5 shows the number of press releases by Senate and House members, as well as government agencies, mentioning the PTC. Anticipation effects would presumably accompany considerable public attention to the PTC before the beginning of the legislative debate. As the figure shows, press releases are much more common during periods of legislative debate and related lobbying. The data support our coding and alleviate concerns about anticipation effects.^{Footnote 19}

Figure 5. Number of press releases mentioning “Production Tax Credit” and “energy” anywhere in the documents distributed by members of the U.S. Senate and House as well as government agencies, overlaid on periods of legislative activity over the PTC.

Note. PTC = production tax credit.

Our second, less stringent, assumption is that the creation of RPS policies is not determined by lobbying activity on the PTC in particular. While RPS policies are obviously influenced by lobbying, we only need to assume that variation in PTC consideration does not itself determine the likelihood and timing of state-level RPS adoption. This assumption is not very stringent because the timing of PTC consideration is erratic and influenced by factors irrelevant to state policy adoption as discussed in detail above.

Based on Figure 4 above, this assumption appears valid. The number of RPS policies increased almost linearly between the years 2002 and 2010, with no relationship to the fluctuating fortunes of the PTC. Moreover, the literature on RPS policies emphasizes their origins in state economics and politics (Lyon and Yin Reference Lyon and Yin2010; Rabe Reference Rabe2006), with no evidence to suggest that federal policies or the lobbying surrounding them would either directly or indirectly influence RPS adoption.^{Footnote 20}

The third assumption is that RPS adoption, as opposed to other state-level factors, drives differential lobbying efforts over PTC among utilities. The validity of the assumption is not obvious, so we must test it carefully. As there is no reason to believe that pre-RPS adoption years would modify the effect of PTC activity on lobbying, we conduct a placebo test by running the main regressions with a modified RPS variable. This variable codes RPS adoption as 1 for a state for the five-year period before actual RPS adoption. If unobserved state characteristics correlated with RPS adoption condition the effect of PTC activity,^{Footnote 21} then the modified variable should behave similarly to the actual RPS-adoption dummy. The estimation results are presented in Section A3 in the online appendix. The results show no substantive effects between the placebo RPS and the PTC: across all model specifications, coefficients of the interaction term are not statistically significant. This suggests that RPS adoption, as opposed to other state-level factors, affects utilities’ lobbying behavior when the PTC extension is about to expire or under legislative debate.

When these three assumptions are met, we can estimate the effects of PTC legislative activity on the lobbying behavior of utilities with different levels of exposure to RPS policies. The triple interaction models that further condition the interactive effects of PTC legislative activity and RPS policies on renewable electricity generation capacity are, in turn, based on theoretically partial correlations. Here, we make no claims about causal identification and instead focus on more classical hypothesis testing of ex ante theoretical expectations.

The relationship between state-level lobbying and federal lobbying warrants a comment. It is certainly possible that lobbying on the RPS itself occurs, but such lobbying should not influence the effects of PTC legislative activity on utilities’ federal lobbying. To the extent that lobbying on the RPS reduces resources available for federal lobbying, the resource reductions should affect all kinds of lobbying, as opposed to only lobbying when the PTC is under threat. Thus, resource constraints from lobbying on the RPS should not have a particular effect under PTC legislative activity at the federal level; potential diversion of resources or attention from the federal level would, instead, cause an overall reduction in federal lobbying activity among utilities.

Finally, it should be noted that our analysis may produce a lower estimate of the effects of PTC expiration and RPS, given that the PTC can be credited over a 10-year period. While this does not prevent utilities from developing a new renewable energy project, utilities that had exhausted their 10-year eligibility would have no clear incentives to lobby on the expiration of PTC if they do not have any plans for a new renewable energy project. This suggests that our estimates would be conservative, and the actual effects of PTC expiration and RPS would be higher than our estimate.

Main Results

The main estimation results are shown in Table 2. In each set of models, we start by examining the effects of impending PTC expiration and the RPS independently. We then add the interaction term and a different time variable in each of the other models. We use the linear year trend, biannual dummies, and quarter dummies in Models 1 and 2, respectively, and a year-dummy in Model 3. In Model 4, we include quarterly fixed effects to allow nonlinear heterogeneity across time periods.

Table 2. Effect of PTC Activity on Lobbying by Electric Utilities, 1998–2012.

Note. Clustered standard errors at the state-level in parentheses. Utility FE is included across all models. PTC = production tax credit; RPS = renewable portfolio standard.

*p < .10. **p < .05. ***p < .01.

The results across estimated models provide support for our theoretical expectations. In line with Hypothesis 1, we find that the propensity to lobby goes up when the PTC is about to expire. In Model 1, the coefficient for the PTC renewal variable is positive and statistically distinguishable from 0. The first model indicates that the odds of lobbying by utilities, on average, are 1.2 times larger when there is a threat of PTC expiration, holding other factors constant. RPS on its own reduces lobbying propensity, as indicated by negative coefficients across the models. These findings suggest that the consideration of PTC expiration and the presence of RPS independently influences the probability of lobbying, but only to a modest degree.

More interestingly, the positive effect of possible PTC expiration on the probability of lobbying is substantially reduced in the presence of an RPS, as indicated by the coefficients on the interaction term. The coefficients are negative in their direction and statistically significant at the conventional level across all the specifications. According to Models 2 to 4, when the PTC is about to expire, the probability that utilities operating in RPS states lobby is about 1.3 to 1.9 times those in non-RPS states. These results show the important interactive effect of federal and state policies. PTC expiration at the federal level increases the incentive to lobby in general, but utilities under the influence of state-level RPS are far less likely to be influenced by PTC expiration than utilities operating in a non-RPS state. The results provide empirical support for Hypothesis 2.

Among other variables, one factor that appears to have substantial effects on lobbying propensity is total net generation, which consistently shows strong positive effects across the models. Not surprisingly, large utilities are more likely to lobby, ceteris paribus. This finding is consistent with the previous literature that argues that smaller utilities in general do not individually lobby because they lack the resources and the political power to influence outcomes (Kim, Urpelainen, and Yang Reference Kim, Urpelainen and Yang2016).

Lobbying by Renewable Electricity Generators and Others

The above analysis provides evidence on the substitution effect between the timing of PTC expiration and state-level renewable policies on lobbying behavior. We further hypothesized that the substitution effect might be stronger for utilities with renewable energy capacity than those without. To test this hypothesis, we estimate models with triple interactions to compare the responses of electric utilities with and without renewable electricity generation.

To better describe how possible PTC expiration and renewable energy generation modify the effect of RPS on lobbying probability, we present the marginal effects of RPS under different conditions in Figure 6. The figure clearly demonstrates the difference in direction and magnitude of marginal effects of RPS across the four different conditions. The first notable pattern is the varying effect of RPS depending on PTC legislative activity. For both types of utilities, the negative effect of RPS is more substantial when PTC is about to expire. The second notable pattern is the difference between renewable and nonrenewable energy generators. RPS makes renewable energy generators 14 percentage points less likely to lobby when the PTC is about to expire. This effect translates to a 23% reduction from the average lobbying probability (61.2%). For nonrenewable energy generators, this negative effect is substantially smaller with only a 7 percentage point decrease (or 11% decrease from the average lobbying probability).

Figure 6. Marginal effect of RPS by renewable versus nonrenewable utilities conditional on the consideration of PTC expiration (left); marginal effect of PTC by renewable versus nonrenewable utilities conditional on the presence of RPS (right). The calculation of marginal effects is based on Model 1 in Table A4 in the online appendix.

Note. RPS = renewable portfolio standard; PTC = production tax credit.

We also illustrate how RPS presence and renewable energy generation capacity change the marginal effect of the PTC in the right panel of Figure 6. The effect of possible PTC expiration is strikingly different between utilities in RPS and non-RPS states. For both types of utilities, the possibility of PTC expiration increases the lobbying probability only for non-RPS utilities. Another observed pattern is the higher sensitivity that utilities with preexisting renewable energy generation capacity operating in non-RPS states have toward impending PTC expiration (5 percentage point increase in lobbying probability due to the possibility of PTC expiration) compared with nonrenewable utilities under the same condition (3 percentage point increase).

Intensive and Extensive Margin of Lobbying

In this section, we investigate the effects of the PTC and RPS on the intensive margin of lobbying—how much is spent on lobbying expenditures when the decision to lobby has been made (Kerr, Lincoln, and Mishra Reference Kerr, Lincoln and Mishra2014). To systematically estimate the effects, we employ a “double hurdle model” (Cragg Reference Cragg1971). Unlike a tobit model, which assumes the same decision-making process for the act of lobbying and the amount spent on the activity, the double hurdle model allows these two decisions to be determined by separate processes (Cragg Reference Cragg1971; Wooldridge Reference Wooldridge2010). We use probit regressions with unconditional utility fixed effects to determine the probability of lobbying by utilities in the first tier, and a truncated ordinary least squares (OLS) regression with utility fixed effects for lobbying expenditures using our base Model 2 in Table 2.

Data on lobbying expenditures were extracted from the mandated lobbying reports. Under the Lobbying Disclosure Act of 1995, lobbying firms must report whether lobbying expenditures were less than US$5,000, as well as the total amount if it exceeds that amount. Therefore, we set the lower truncation limit to US$5,000 in the second tier of the double hurdle models. Also, we utilize the logarithmized value of lobbying expenditures.

The results for the first and second tiers are reported in Table 3. Consistent with the results from our main models, the estimated effect of RPS on lobbying decisions when the PTC is about to expire is negative and robust, as shown in the first-tier results. On the contrary, the second-stage results indicate that there might not be substitution effects between RPS and the PTC on lobbying expenditures. However, the coefficient is not precisely estimated at conventional levels. Such null results imply that there is no difference in lobbying intensity between utilities in RPS and non-RPS states. Combined with the robust effects for the extensive margin of lobbying, such results suggest that utilities operating in states with stringent RPSs adjust their lobbying decisions rather than the amount of lobbying expenses when they face the threat of PTC expiration. In other words, as our theory suggests, the substitution effect of the PTC and RPS decrease utilities’ incentives to lobby in the first place rather than the intensity of lobbying activities.

Table 3. Estimation Results Using Double-Hurdle Models: Effect of PTC Activity on Lobbying Amount (Logged) by Electric Utilities, 1998–2012.

Note. Standard errors clustered at the state-level in parentheses. Utility FE is included across all models. For Tier 1, the dependent variable is lobbying. For Tier 2, the dependent variable is the logarithmized value of the lobbying amount. FE = fixed effect; PTC = production tax credit; RPS = renewable portfolio standard.

*p < .10. **p < .05. ***p < .01.

Issues Lobbied

Having demonstrated the effect of federal and state policies on the likelihood of lobbying by electric utilities, we next conduct a validity test to investigate whether electric utilities lobby on issues related to energy policy. While lobbying reports do not contain information on whether firms lobby for or against a given policy, firms report on which issues they lobby. We can use information on the lobbied issues to test the validity of our empirical analysis. If electric utilities do not lobby on the energy issue, our analysis would not be valid in testing the effects of energy policy on their lobbying behavior.

We examine the specific lobbying issues described in lobbying reports. The text data we analyze come from the “specific lobbying issues” section in the lobbying reports. In this section, organizations or lobbying firms describe bills or issues that they lobbied on during a given period. Our analysis focuses on observations where this description is available because some do not provide a description of specific lobbying issues. Among the examined lobbying reports in our analysis, 12.8% provide information on specific issues they lobbied on. The degree of specificities of the contents varies greatly across lobbying reports. For instance, the AES Corporation mentions “climate and energy legislation” without further details in their lobbying report in the first quarter of 2009. Invenergy Wind provides more detailed information that they lobbied on the issue of “legislative and regulatory services related to PTC of section 45: H.R. 7060; the Renewable Energy and Job Creation Act of 2008” in their lobbying report of the same quarter.^{Footnote 22}

We analyze the contents of specific lobbied issues in the lobbying reports using a topic modeling technique. Specifically, we classify the contents into a number of topics using a text analysis technique, the structural topic model (STM), which enables the discovery of topics from given text data that incorporate information about each document (Roberts et al. Reference Roberts, Stewart, Tingley, Lucas, Leder-Luis, Gadarian, Albertson and Rand2014). We briefly discuss the estimation procedure and visually describe the topic model results and provide a detailed discussion on the estimation and the results in the online appendix (Section A11 in the online appendix).

We utilized the STM R package from preprocessing to estimation. After preprocessing our texts (e.g., removing stop words, stemming), we estimated eight topic models with a varying number of topics from five to 10 and 15 to 40 with an interval of five. After examining the results from various topic model estimations, we select the model with 15 topics (as the results from a larger number of topics do not provide new insights into the reports). Figure 7 presents the proportion of each topic. The most frequently lobbied issue (Topic 9) and the third most frequently lobbied issue (Topic 12) are directly related to the issue of energy.^{Footnote 23}

Figure 7. Expected topic proportions from structural topic model estimation results.

Note. Data for the analysis are from specific lobbying issues in lobbying reports by utilities.

Robustness

As a robustness check, we estimate another set of models by focusing on energy and environment-related lobbying behaviors among electric utilities. The goal is to ensure that our results are not driven by lobbying activities targeted at other irrelevant issues, but instead capture utilities’ incentives to lobby in response to the PTC and RPS. In Table A17 in the online appendix, we report results from specifications that utilize lobbying on energy and environmental issues as the dependent variable. We find the results very similar to the regressions using all lobbying activities by utilities. A negative coefficient on the PTC-RPS interaction term is consistently found and obtained a conventional level of statistical significance in almost all the estimated models.

Placebo Test

We next explore whether the effects we documented are driven by lobbying activities on issues unrelated to renewables. We expect RPS and the PTC to only affect utilities’ incentives to lobby on energy and environmental issues, but not other issues. To verify this, we conduct a placebo test using lobbying on trade issues as the dependent variable. The critical condition for a good placebo test is to choose an issue area that is completely irrelevant to the PTC. This was why trade was chosen over tax even though the former is more frequently lobbied on (35.3% vs. 16.2%). Unlike taxation issues, trade is largely about tariffs and other trade policies, which are irrelevant to the purely domestic PTC. For the classification of lobbied issues, we follow the issue codes provided in the lobbying reports. The dependent variable is coded 1 when utilities lobbied on issues related to trade but not energy and environment, and 0 otherwise.

We present the results with lobbying on trade as dependent variable in Table A20 in the online appendix. There is no evidence of the substitution effect between RPS and the PTC on trade-related lobbying activities. The interaction term appears to be negative but close to 0 and not statistically significant at the conventional level in any of the estimated models. These findings provide support for our proposed mechanism.