A new measure of energy policy position

When measuring a member’s position on environmental or energy policy, scholars often use the League of Conservation Voter (LCV) scores (e.g. Shipan and Lowry Reference Shipan and Lowry2001). The LCV score, ranging between 0 and 100 for each member of Congress for each year with higher scores reflecting more pro-environment positions, has been the best available measure and is more focused than the Americans for Democratic Action (ADA) scores used in previous analyses of legislative positions of members of Congress. However, the LCV score is problematic because it tends to utilise only a small number of selective energy and environment policy votes. Since consensus votes and less controversial votes are excluded, the ideological differences among members will be estimated to be much larger than they are, making liberals more liberal and conservatives more conservative than they really are (Snyder Reference Snyder1992; Poole and Rosenthal Reference Poole and Rosenthal1997). Another problem of this approach is that the measure is unable to distinguish among nonmoderate legislators, producing less informative measures of ideology. Since only a small number of controversial votes are used to measure members’ ideology, there will be few votes that distinguish among nonmoderate legislators. As a result, many nonmoderate legislators will be estimated to have the same or similar positions, even when there are ideological differences among them. Finally, the LCV score is not a focused measure of a member’s energy policy position because it is based on environmental policy votes as well as energy votes. For example, in the 2016 LCV scorecard, roughly one-third of the Senate votes and one-fourth of the House votes focus on energy. The remaining votes focus on issues, such as the protection of endangered species, that are often only tangentially related to energy policies.

To address these problems, we measure energy policy positions of members of Congress by using all energy policy votes.Footnote ¹ By utilising all energy policy votes, this approach produces a finer and more informative measure than the measures based on a small number of selective energy policy votes. By utilising only energy policy votes, this measure avoids conflating a member’s energy policy position with his or her environmental policy position. To measure the energy policy position of a member that is comparable across time, we use a dynamic item response theory (IRT) model or ideal point estimation technique developed by Martin and Quinn (Reference Martin and Quinn2002). By combining information from previous Congresses in estimating policy positions in the current Congress, this approach allows us to directly compare measures from one Congress to another. We fit a one-dimensional dynamic ideal point estimation model to all energy policy votes from 1971 to 2016 for the House and Senate.Footnote ² Most estimated energy policy positions are in the range between −5 and + 5, with smaller scores indicating more liberal positions on energy policy (i.e. support for more energy conservation and renewable energy).

Before we use this measure, we compare the performance of this measure with that of the LCV scores. In Figure 1, we plot the LCV score of senators in select Congresses against their IRT measure. Note that a high score in the IRT measure indicates an opposition to more regulation on energy policy, whereas a high score in the LCV score indicates support for more regulation. Also, the IRT measure ranges typically from −5 to 5, whereas the LCV score ranges from 0 to 100. The figure shows that there are notable differences between the IRT measure and the LCV score. First, the LCV score has a tendency to overemphasise the degree of polarisation, compared to the IRT measure. This is indicated by the fact that the LCV score tends to be higher than the 135-degree line when the IRT measure is negative, and lower than the 135-degree line when the IRT measure is positive. For instance, in the 92nd Congress, senators are estimated to be more polarised in the LCV score than in the IRT score. On the vertical axis, there are several senators with their LCV scores greater than 80. However, their IRT score on the horizontal axis indicates that they are not that far from the centre of the Senate. Although this trend of overemphasising the degree of polarisation in the LCV score becomes weaker in more recent Congresses, we are likely to make an incorrect conclusion about the degree and pattern of polarisation if we rely solely on the LCV score. There is another difference between the IRT measure and the LCV score. The IRT measure is better at distinguishing ideological differences among extreme legislators than is the LCV score. For instance, in the bottom right panel of Figure 1 (the 111th Senate), there are a group of legislators with an 80% LCV score, while their positions are well sorted out on the horizontal axis. This is explained by the fact that the use of selective votes in producing the LCV score makes it a less informative measure of the position, especially among extreme legislators. By using less controversial or lopsided votes as well as controversial votes, our approach is able to capture subtle differences even among nonmoderate legislators.

Figure 1

Comparison of LCV scores and dynamic ideal point estimates of senators in select Congresses.

Note: A high score in the IRT estimate indicates a conservative position, whereas a high score in the LCV score indicates a liberal position. LCV scores are adjusted using the Groselose et al.’s (1999) procedure. Dots represent Democrats, whereas crosses represent Republicans.

Historical development of party polarisation on energy policies

Using this measure of members’ positions on energy policy, we examine the historical pattern of party polarisation on energy policy in Congress. First, to determine how much energy policy polarisation has been distinctive from the broad pattern of party polarisation, we compare the pattern of party polarisation on energy policy with that of general ideology. We use the DW-NOMINATE first-dimension score to measure the degree of party polarisation on general ideology because DW-NOMINATE scores are computed using all votes in Congress (Poole and Rosenthal Reference Poole and Rosenthal1997). However, since DW-NOMINATE measures are on a different scale, a direct comparison of DW-NOMINATE scores and our IRT measures is not possible. Thus, to improve the comparability of the measures, we standardised both measures before we compute party positions and polarisation scores.

We examine the degree of party polarisation on energy policy and general ideology by the difference between the respective means of the ideal points of Democratic and Republican members, following McCarty et al. (Reference McCarty, Poole and Rosenthal2006). Figure 2 demonstrates that the pattern of party polarisation on energy policy has been distinctive from that of general ideology. In particular, the figure shows that energy policy used to be a nonpartisan issue in the early 1970s. In comparison to the degree of party polarisation on general ideology, the political parties in both chambers were much less polarised on energy policy in the 1970s. While the party polarisation over energy policy increased consistently in the 1970s in the House, the Senate has a different pattern with a steep increase happening in the 1990s and mid-2000s. While a systematic analysis of the inter-chamber difference is beyond the scope of this article, one possible explanation for the inter-chamber difference is the influence of Robert C. Byrd (D-WV) and Senators who represented fossil fuel producer states. In particular, as the majority (1977–1980; 1987–1988) and minority (1981–1987) leaders of the Senate and also with his knowledge of parliamentary procedure, Senator Byrd had a significant influence on the Senate. Since he represented one of the most significant coal-producing states, he was keen on blocking legislation that would have imposed tougher regulations on energy production. Indeed, historical accounts of American energy policy refer to Byrd as “eastern coal interests’ greatest ally” (Graetz 2011: 51). The fact that the Senate was able to pass the Clean Air Act of 1990 only after he was no longer the leader of the Democrats in the Senate illustrates his influence on the energy policy (CQ Almanac 1990). We think this is plausible to a certain degree and thus test the influence of members who represent fossil fuel producer states in the next section. However, we do not think a single senator can explain all the partisan dynamics on energy policy. Unlike the House, party leaders in the Senate have few procedural tools to control the agenda of the Senate.Footnote ³ Also, the fact that the degree of partisan polarisation over energy policy has been gradually increasing since the early 1970s indicates that there were other factors that made energy politics more partisan. In addition, when we plot the median position of each party in Figure 3, we see that Republican Senators were also less partisan on energy policy than on general ideology until the early 1990s. Combined with the fact that this pattern of party polarisation in the Senate is generally consistent with the history of energy policymaking in Congress described in the previous section, suggests that there were more systematic reasons for a different pattern in energy policy polarisation.

Figure 2

Polarisation indices.

Note: Polarisation indices = |Dem mean − GOP mean|. General ideology is based on DW-NOMINATE first-dimension score. To improve comparability, DW-NOMINATE scores and IRT measures are standardised before the computation of polarisation indices.

Figure 3

Median positions of political parties.

Note: General ideology is based on DW-NOMINATE first-dimension score. To improve comparability, DW-NOMINATE scores and IRT measures are standardised before the computation of party median positions.

We suspect that the strong party leaders and their control over the agenda in the House can explain the inter-chamber difference. This is because the timing of the convergence between energy policy and general ideology in the House is consistent with the increased powers of the party leaders due to the House reforms in the 1970s. In the House, energy policy positions have converged to the general ideology of the parties by the late 1970s. Since then, each party’s position on energy policy has been consistent with its general policy position. The reforms in the House in the 1970s gave the party leaders the powers to control the floor of the House, which increased the partisanship across the board in this chamber (Rohde Reference Rohde1991).

Notably, this inter-chamber difference in the polarisation pattern of energy policy adds to the need to study energy politics separately, as the polarisation literature finds that the pattern of party polarisation has been similar between the House and the Senate (e.g. Theriault Reference Theriault2010; Theriault and Rohde Reference Theriault and Rohde2011; also see Figure 2). Unlike party polarisation in general ideology, energy policy polarisation is characterised by the inter-chamber differences. Therefore, we consider policy-specific factors as well as general political contexts to fully account for the party politics in this particular issue area in the next section.

Major events in energy policy

The PET suggests that we need to look for policy-specific factors to explain shifts in party behaviour in specific policy areas. Specifically, the PET argues that “American political institutions were conservatively designed to resist many efforts at change” but also that occasionally, often precipitated by crises or major shifts in issue definition, “some issues catch fire, dominate the agenda, and result in changes” (Baumgartner et al. Reference Baumgartner, Jones, Mortensen, Sabatier and Weible2014: 62; see also Baumgartner and Jones Reference Baumgartner and Jones1993). Other scholars have also ascribed an important role to sudden disturbances. For instance, Birkland’s study of natural disasters describes the importance to policy change of such events that are “sudden, relatively rare, can be reasonably defined as harmful or revealing the possibility of potentially greater future harms” and “is known to policy makers and the public virtually simultaneously” (Birkland Reference Birkland1997: 22).

We build on these notions of the importance of shocks to policy systems by testing the influence of energy policy-specific events on the degree of energy policy polarisation. Specifically, we argue that events described earlier such as the oil crises of the early 1970s and the emergence of the climate change debate sowed the seeds for the partisan alignment on energy policies that has existed ever since. The oil crises of the 1970s were seminal events in American energy politics. The real price of a barrel of imported crude oil jumped from an average of $19.6 in 1971 and 1972 to an average of $43.6 in 1973 and 1974 and continued to grow by 1981.Footnote ⁴ Many scholars assert that the oil crisis events had immense, long-term consequences for American politics in general but especially for energy issues (Graetz Reference Graetz2011; Fiege Reference Fiege2012; Lowry and Joslyn Reference Lowry and Joslyn2014). Thus, we use the real price of a barrel of imported crude oil (logged) to tap the influence of rising oil prices.

We also examine how the climate change debate in the US has affected the polarisation over energy policy. Although scientists had warned of global warming before then, many analysts describe Hansen’s testimony in June 1988 as the catalyst for debate among American policymakers (Yergin Reference Yergin2011: 454; Titlow and Tinger Reference Titlow and Tinger2016: 250). The United Nations also created the Intergovernmental Panel on Climate Change in 1988. Further, Hansen’s testimony prompted efforts to contradict his arguments from the Global Climate Coalition formed by fossil fuel companies the following year. While the debate raised awareness in the US, creating the potential for changes in energy policies, it had yet to produce a consensus among the public or members of Congress on causes, consequences or responses. Given this variation, we measure the influence of the climate change debate by the percentage of respondents who expressed that they are personally worried about global warming or the greenhouse effect in the Gallup surveys since 1989.Footnote ⁵ We lag it by one year to account for the possible delay in response to the climate change debate and also to reduce the possibility of reverse causality.

In addition to these two major shocks to the policy system, we need to take into account the increased salience of energy policy due to other events, such as the Three Mile Island nuclear accident and oil spills (e.g. the EXXON Valdez oil spill). Although one can try an indicator variable for each of these events, the small number of Congress-level observations and the difficulty in measuring their lasting influence make such an approach less effective.Footnote ⁶ Therefore, to measure the salience of energy policy in Congress effectively, we use the number of Government Accountability Office (GAO, formerly General Accounting Office) reports on energy. The GAO is an independent, nonpartisan office that is designed to provide objective, reliable information to Congress. The GAO produces reports in response to requests from congressional committees or subcommittees or that are statutorily required by certain laws. Upon receiving a request, the GAO studies the issue and responds with a report typically 10–18 months later. Conveniently, the GAO categorises their reports for each year by issue area, including one specifically for energy. We thus count the number of GAO reports on energy, lag it by one year to allow for publication time, and consider it as a percentage of all reports as the GAO is more prolific in some years than others. This is a nice measure for salience for several reasons. First, the assumption that congressional interest reflects public interest in an issue is realistic. Second, unlike many public opinion polls used to reflect salience, the numbers are consistent over time. Third, the GAO has consistently maintained high standards for reliability (Kingsbury Reference Kingsbury2010). Fourth, as mentioned, the reports are focused on specific policy areas. Surprisingly, this measure has rarely been used in other analyses that try to capture issue salience over time although some studies have found it useful (Kingsbury Reference Kingsbury2010; Lowry Reference Lowry2008). We expect that this measure of energy policy salience will be positively associated with the degree of energy policy polarisation.Footnote ⁷ Thus, we summarise the hypotheses as follows:

Changing nature of energy policies

In addition to these events, we examine a hypothesis based on the unique nature of energy policymaking in the US. A study by Lowry (Reference Lowry2008) explains the increased partisanship on energy policymaking in Congress by the change in the nature of energy policy from distributive to regulatory. According to Lowry, as energy policy has become more regulatory in its nature, the issue has turned into a more conflictual one, similar to environmental policy. We expect the politics surrounding energy policy to become more intertwined with the politics surrounding environmental policy over this time period for several reasons. First, Americans have become increasingly aware of the environmental consequences of energy use and consumption. Examples abound – but here is just one illustration: a substantial portion (as much as half by some estimates) of unexploited fossil fuels in the US is in federal lands and waters, areas that are precious to environmental advocates (Adler Reference Adler2015: 2). Second, and related, decisions on energy policies are required by law to take into account environmental consequences since the passage of the National Environmental Policy Act (NEPA) of 1970. The NEPA requires assessment of the environmental impacts of all federal decisions (and others using federal money), including energy policies such as the siting of power plants. Third, starting in the 1970s, energy policies often created winners and losers, as do regulatory policies. As just one example, the increased CAFÉ standards first established in 1975 rewarded makers of high-mileage cars while creating significant challenges for makers of low-mileage vehicles. Fourth, environmental interest groups, although not always in agreement on specific energy policies such as proposals for carbon taxes, have coalesced around certain energy policies since the 1970s such as preventing oil exploration in the ANWR and reducing the use of fossil fuels that contribute to climate change. As the politics of energy and the environment converge, so too will the policies become more similar in tone.

If the convergence of energy policy and environmental policy has contributed to party polarisation over energy policy, we expect to see the convergence of the energy policy position and the environmental policy position of a member. That is, if the nature of energy policy has changed into something similar to environmental policy, a member’s position on environmental policy will predict the member’s position on energy policy better over time.

Thus, we capture the effect of the convergence by examining the association between individual members’ environmental policy extremity and their energy policy extremity. For this, we need a measure of a member’s environmental policy position. We apply the dynamic ideal point estimation technique to all environmental policy votes in Congress to measure a member’s environmental policy position.Footnote ⁸ Once we have a measure of a member’s environmental policy position, we measure the member’s environmental policy extremity in each Congress using her appropriately signed distance from the floor mean. This environmental extremism score is signed positively if the member’s deviation is in the same direction as that of her party mean, and signed negatively if in the opposite direction. This way, a moderate member will be coded as contributing to the moderation of party differences, whereas an extreme member will be coded as contributing to the polarisation of party differences.

Partisan electoral rivalry

In addition to factors specific to energy policy, we also consider factors related to party polarisation in general. First, we hypothesise that polarisation on energy policies has been driven partly by the competitiveness of partisan electoral rivalry. While political parties are driven by policy goals, they are also driven by electoral goals. Since policy success of the opposition party is likely to have a negative impact on a party’s electoral goals, political parties’ choice of a policy position is likely to be partially affected by the prospect of elections (Lee Reference Lee2016). We expect that this effect should be stronger when party control of the respective chamber is closely contested (Dion Reference Dion1997; see also Patty Reference Patty2008). Expecting a close election, the minority party’s incentives to oppose the policy success of the majority party will increase. In their study of the role of partisanship in party polarisation, Lowry and Shipan (Reference Lowry and Shipan2002) found that when the majority party has a slim seat margin, the parties are more polarised on domestic policy. We would expect the same pattern to emerge on energy policy.

Therefore, to test the effects of partisan electoral incentives, we use the majority party’s seat margin (%) and the status of divided government.Footnote ⁹ As we discussed, Lowry and Shipan (Reference Lowry and Shipan2002) found that a minority party is more prone to differentiate itself from the majority party when the difference between their numbers of seats is small. We also expect energy policy polarisation to depend on whether party control of government is unified or divided. Under divided government the congressional majority may have a stronger electoral incentive to oppose the president, which would tend to increase polarisation. Thus, we include a measure of divided government – an indicator variable that takes a value of 1 when the majority party of the chamber is different from the president’s party.

Party regional realignment

The literature on party polarisation suggests that the regional realignment of the party base has caused party polarisation in Congress. In fact, regional realignment is one of the most widely recognised sources of party polarisation (Aldrich Reference Aldrich1995; Jacobson Reference Jacobson, Bond and Fleisher2000). By making the affected states’ partisan and ideological tendencies mutually reinforcing, realignment has increased the number of states that reliably elect conservative Republicans in the South or liberal Democrats in the Northeast.

Therefore, to examine the influence of partisan realignment, we focus on the role of Southern members. If the Southern realignment has contributed to the polarisation of energy policy as well, Southern members must have had conservative positions on energy policy as well. In other words, Southern Democrats must have contributed to the moderation of energy policy by making Democrats less extreme, whereas Southern Republicans must have contributed to the polarisation of energy policy by pulling Republicans towards a more conservative energy policy position. Then, the increasing number of Southern Republicans and the decreasing number of Southern Democrats will contribute to energy policy polarisation.

Member-fixed-effects models

Combining the member fixed effects and environmental policy extremism at the member-level and Congress-level factors, we thus account for energy policy polarisation as follows:

$$Energy - Policy - Extremit{y_{ij}}{\rm{}}^{\sim}{\rm{ }}N({\mu _{ij}},{\sigma ^2}),\;{\rm{for}}\,\,I{\rm{}} = \;1,{\rm{ }} \ldots ,N$$ (1)

$${\mu _{ij}}{\rm{}} = {\rm{ }} + {\rm{}}{\alpha _{l[i]}}{\rm{}} + {\rm{}}{\beta _{j[i]}}Environmental - Policy - Extremit{y_i}$$ (2)

$${\rm{}}^{\sim}{\rm{ }}N(\mu _j^{Congress},\sigma _{Congress}^2),\;{\rm{for}}\,\,j\; = \;1,\; \ldots ,M$$ (3)

$$\mu _j^{Congress}{\rm{}} = {\rm{ }}{\gamma _0}{\rm{}} + {\rm{}}{\gamma _1}Divided \,\,Governmen{t_j}{\rm{}} + {\rm{}}{\gamma _2}Majority - Seat - Margi{n_j}{\rm{}} + {\rm{}}{\gamma _3}Oil\,\, Pric{e_j}{\rm{}} + {\rm{}}{\gamma _4}Climate\,\, Chang{e_j}{\rm{}} + {\rm{}}{\gamma _5}GAO \,\,Report{s_j}$$ (4)

where N is the total number of members; M is the total number of Congresses from the 91st through 114th; α _l[i] represents member fixed effects; Environmental-Policy-Extremity_i is the measure of environmental policy extremity, as we discussed above. Note that the coefficient for this variable is modelled as a varying slope because the effects of environmental policy extremism on energy policy polarisation are expected to change over time.

The estimate of this varying slope and varying intercept model is presented in Table 2 and Figure 4. First, Table 2 shows that energy policy event variables are, in general, significantly associated with the increased polarisation over energy policy, providing support for Hypotheses 1a, 1b and 1c. First, the rising oil price is found to be significantly related to polarisation over energy policy in the Senate. Oil price is not found to be significantly related to polarisation in the House. To gauge the substantive effect of the Oil Shocks in the 1970s, we used the estimated model to predict the average level of energy policy extremism in the 92nd Congress (before the Oil Shocks) and the 94th Congress (after the Oil Shocks). We held the value of all other variables at their level in the 92nd Congress, while allowing the price of oil to change from its level in the 92nd Congress (i.e. $19.6) to its level in the 94th Congress (i.e. $43.6). This resulted in a 63% increase in the average energy policy extremism in the Senate and a 5% increase in the House.

Table 2.

Congress-level variables and energy policy polarisation (member-fixed-effects models)

Note: Member-level coefficient estimates are reported in Figure 4. ***, ** and * indicate that 99%, 95% and 90% credible intervals do not include 0, respectively. DIC, Deviation Information Criterion.

Figure 4

Varying relationships between environmental policy extremism and energy policy extremism.

Note: The coefficients represent varying slopes of environmental extremism (i.e. β in Equation 2). The bar around a dot represents 95% credible interval. Member fixed effects are included in the model, but not reported here.

Second, the salience of climate change debate is found to be significantly associated with energy policy polarisation in both chambers, albeit at the 90% level in the House. Although the observational date does not allow us to make a strong claim about the causality, the positive coefficient indicates that political parties tend to be more polarised over energy policy when the issue of climate change was salient. To gauge the substantive effect of the climate change debate, we used the estimated model to predict the average level of energy policy extremism in the 100th Congress (before the Hansen’s testimony) and the 101st Congress (after the testimony). We held the value of all other variables at their level in the 100th Congress while allowing the salience of climate change debate to change from its level in the 100th Congress (i.e. 0%) to its level in the 101st Congress (i.e. 32.5%). This resulted in an 80% increase in the average energy policy extremism in the Senate and 10% in the House.

Third, the salience of energy policy in Congress – measured by the percentage of GAO reports on energy – is positively and significantly related to polarisation over energy policy in both chambers, albeit at the 90% level in the Senate. The positive coefficient indicates that political parties tend to be more polarised over energy policy when Congress perceived the issue of energy policy as salient and thus requested more reports on energy policy from the GAO. To gauge the substantive effect of the salience of energy policy in Congress, we used the estimated model to predict the average level of energy policy extremism in the 93rd Congress (the lowest salience of energy policy) and the 97th Congress (the highest salience of energy policy). We held the value of all other variables at their level in the 93rd Congress while allowing the percentage of GAO reports on energy to change from its level in the 93rd Congress (i.e. 0.7%) to its level in the 97th Congress (i.e. 10%). This resulted in a 51% increase in the average energy policy extremism in the Senate and 49% in the House.

Also, presented in Figure 4, the varying slopes of environmental policy extremity (β _j ) are consistent with the expectation that energy policymaking has become more like environmental policymaking over the years (Hypothesis 2). In the early 1970s (House) and entire 1970s (Senate), the connection between environmental policymaking and energy policymaking was substantively low (the House) and statistically insignificant (the Senate). However, as time went by, the coefficient for environmental policy extremism has become significantly distinctive from zero. The positive slopes indicate that members with more extreme environmental policy position were more extreme in their energy policy positions. Moreover, the increasing size of these coefficients indicates that the association between environmental policy extremity and energy policy polarisation has increased from earlier Congresses to later ones. This is consistent with Lowry’s (Reference Lowry2008) argument that energy policy has become more like environmental policy and that this trend has contributed to partisan divergence on energy policymaking in Congress.

Finally, Table 2 shows that there is partial support for the partisan electoral rivalry theory. Although the majority party’s seat margin (Hypothesis 1a) was not significantly related to energy policy polarisation in both chambers, the divided government variable (Hypothesis 1b) is positively and significantly associated with increased party polarisation over energy policy in the Senate. When the majority party of the Senate was different from the party of the president, senators were more likely to take more extreme positions on energy policy. We do not find the same effect in the House. In the House, whether the majority party was different from that of the president had no distinctive effects on party polarisation over energy policy. Since energy policy had converged to environmental policy early in the 1970s in the House (see Figure 4), we suspect that political parties were polarised over energy policy regardless of the status of the government.

Senate model with no member fixed effects

We also want to examine a key argument for party polarisation that is not specific to energy policy as well as an argument that suggests long-term potential constituent influence. Since the member-fixed-effects model does not allow us to test the effect of partisan realignment in the South and the influence of members representing fossil fuel producer states, we test these theories by including several member-level variables in lieu of member fixed effects. We do this for the Senate data only because fossil fuel production data are not available at the House district level. Thus, the Equations 1 and 2 are replaced by the following:

$$Energy - Policy - Extremit{y_{ij}}{\rm{}}^{\sim}{\rm{ }}N({\mu _{ij}},{\sigma ^2}),\;{\rm{for}}\,\,I\; = \;1,{\rm{ }} \ldots ,N$$ (5)

$${\hskip-1.2pc\mu _{ij}} = + {\beta _1}Democrat{s_i} + {\beta _2}Sout{h_i} + {\beta _3}Fossil{\rm{ }}\,\,Fue{l_i} + {\beta _4}Interio{r_i}\\ \quad + {\beta _5}SEP{C_i} + {\beta _6}Democrat{s_i}\,^{*}\,Sout{h_i} + {\beta _7}Democrat{s_i}\,^{*}\,Fossil \,\,{\rm{ }}Fue{l_i}\hskip-1.4pc \\ \quad + {\beta _8}Democrat{s_i}\,^{*}\,Interio{r_i} + {\beta _9}Democrat{s_i}\,^{*}\,SEP{C_i} \\ \quad + {\beta _j}_{[i]}Environmental - Policy - Extremit{y_i}$$ (6)

where Interior represents the membership on the Senate Committee on Interior and Insulars Affair (Senate Committee on Energy and Natural Resources since 1977). SEPC represents the membership on the Senate Committee on Environment and Public Works. Fossil Fuel represents the total amount of fossil fuel production (coal, crude oil and natural gas in billion btu) per capita in each state.Footnote ¹⁰

The results are reported in Table 3.Footnote ¹¹ The results at the Congress level are very similar to those with the member fixed effects. The salience of climate change debate, the salience of energy policy in Congress, the rising oil price and the divided government are positively and significantly associated with polarisation over energy policy.

Table 3.

Energy policy polarisation in the senate (without member fixed effects)

Note: ***, ** and * indicate that 99%, 95% and 90% credible intervals do not include 0, respectively. DIC, Deviation Information Criterion.

At the member level, the results are consistent with the expectations. First, Southern Democrats (South × Democrats) are negatively and significantly associated with energy policy polarisation. By being more conservative within the Democratic Party (even after holding the state level of fossil fuel production constant), Southern Democrats contributed to the moderation of energy policymaking. This suggests that political parties’ realignment in the South has contributed to the polarisation of energy policy in the Senate, providing support for Hypothesis 4. In order to gauge the degree and influence of Southern realignment in both chambers, we plot the median energy policy positions of Southern and non-Southern members in each party in Figure 5. It shows that Southern Democrats have contributed to the moderation of energy policy by being more conservative on energy policy than non-Southern Democrats in both chambers. On the contrary, Southern Republicans have contributed to the polarisation of energy policy by being more conservative within the Republican Party. Also, the bottom panel of the figure shows the depth of this realignment. It shows that the number of Southern Democrats has been decreasing and that of Southern Republicans increasing, contributing to the increasing polarisation over energy policy.

Figure 5

Median positions and the percentages of Southern and non-Southern members.

Note: The bar around a dot in the top two panels represents 95% credible interval. The South includes Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas and Virginia.

Second, we find that Democrats who represent fossil fuel producer states (Fossil Fuel × Democrats) are negatively and significantly associated with energy policy polarisation. This confirms that Senators like Byrd and Long contributed to the moderation of energy policymaking by being more conservative within the Democratic Party. In contrast, the positive coefficient for Fossil Fuel indicates that Republicans who represent fossil fuel producer states contributed to the polarisation of energy policy by being more conservative than other Republicans.

Finally, the members on the Senate Committee on Interior and Insular Affairs (later Energy and Natural Resources) are significantly related to polarisation on energy policy. The members of this committee are likely to have self-selected into this committee due to their state interests. Thus, Democrats in the committee (Interior × Democrats) tended to have a more conservative position on energy policy than other Democrats, contributing to the moderation of energy policy. In contrast, Republicans in the same committee (Interior) contributed to the polarisation of energy policy by being more conservative than other Republicans. We do not find such an effect in the membership on the Senate Committee on Environment and Public Works.