A hazard analysis of federal permitting under the national environmental policy act of 1970

Abstract

The National Environmental Policy Act (NEPA) of 1970 requires federal agencies to assess the environmental impact of proposed federal actions. NEPA thus affects delivery of a wide range of infrastructure projects. NEPA requires the completion of an Environmental Impact Statement (EIS) for environmentally impactful federal actions. For infrastructure projects this can entail significant delays. A typical EIS now takes about four and one-half years and is over 600 pages long. Some EIS’s take over a decade to complete. We provide the first detailed analysis of project approval times under NEPA by examining 1269 EIS permitting processes. We analyze empirically the well-defined interval from Notice of Intent to file to Record of Decision (ROD). We use a Cox proportional hazard model to estimate the impact of several factors on EIS duration. Factors include permits featuring major construction, those including private investment, those for projects located in states with restrictive environmental laws, those using the federal permitting “dashboard,” and those publishing a Supplemental EIS prior to the ROD. We find that privately financed projects receive faster permitting, while projects involving major construction, those undertaken in restrictive states, and those utilizing the federal permitting dashboard, face slower permitting times. We also explore links between EIS page counts and permitting time. Greater EIS page counts are associated with longer permitting times. We conclude by examining EIS completion during economic stimulus programs such as the American Recovery and Reinvestment Act (ARRA), as well as the frequency of EIS completion by the federal government.

1 Introduction

The National Environmental Policy Act of 1970 (NEPA)¹ is a procedural statute requiring federal agencies to assess the environmental impact of proposed major federal actions. NEPA requires the federal government to consider the social and environmental impact of a wide range of proposed activities. A partial list of potentially impacted activities includes: (i) the construction of roads, bridges and highways; (ii) airport construction and renovation; (iii), the generation, distribution, and transmission of both conventional and renewable energy; (iv) construction and renovation of water infrastructure; (v), broadband deployment; (vi) plans to manage and develop public lands, (vii) federal authorization of non-federal activities such as licenses and permits; and (viii) management activities on public lands. Management activities include leases and authorization for energy production, mining, and grazing; management of national parks and forests; and environmental restoration projects, among others.

Despite its clear importance for the environment and the administration of the federal government, NEPA has historically received relatively little political attention. However, the law was at the center of congressional debates over permitting reform as recently as 2022. Several pieces of legislation to “streamline” NEPA were proposed in Congress and several reforms have passed into law. The Trump and Biden Administrations have also produced back-and-forth rulemakings regarding agency guidance for NEPA compliance. Yet even with this renewed political interest, there is a clear lack of information and almost no consensus on even the most measurable aspects of NEPA’s administration. Current literature on the topic contains few conclusions regarding the costs of the process and the factors that drive the duration of studies under NEPA. This lack of empirical conclusions regarding NEPA’s costs and benefits has confounded political debate over optimal permitting reforms. We hope that this study will help address that literature gap.

NEPA initially applied to federal—rather than state or private—projects. However, many privately financed projects fall under NEPA if there is significant federal involvement or if the project is on federal land. For example, privately financed pipelines, or timber or mining projects on federal land are subject to NEPA. NEPA thus impacts the cost and speed of delivering many large infrastructure projects.²

NEPA requires public review and comment and is subject to judicial review. Review under NEPA occurs through the Administrative Procedures Act.³ Courts can require additional study or enjoin the federal action in question, which may already be underway. Applicants must post a draft for public comment prior to agency action. Studies under NEPA often undergo multiple rounds of comment from interested citizens and stakeholder groups. Many are now lengthy. The Council on Environmental Quality (CEQ) determined that the average length of an environmental impact statement (EIS) under NEPA is now over 600 pages (excluding appendices), and that the average time for federal agencies to conduct reviews is about four and a half years.⁴ Some reviews take much longer. EIS documents have also grown rapidly over time. Immediately after NEPA’s passage, highway departments produced hundreds of EISs for projects already in development. The majority were less than 15 pages long while less than 2% were longer than 85 pages (CEQ, 1971).

Despite a lack of empirical evidence, the NEPA process has clearly grown more burdensome since the early 1970’s, and the federal government is completing fewer EIS’s as well. For instance, in 1976 the CEQ reported that it completed 1776 EIS, but in 1984 the number of EIS’s completed by federal agencies was just 577 (Kent & Pendergrass, 1986). The data set on which this study is based includes 120–167 EIS’s per year across the federal government. For the recent EIS’s for which we have page count data, the average is a 661-page main body and 1042 pages of appendices. Clearly the general administrative burden of EIS development, in time or pages, has increased substantially over time, and federal agencies have pursued fewer projects that would require an EIS.

Impacted parties can challenge federal decisions in court after the NEPA process is completed. NEPA litigation delays projects in several ways. Directly, courts may enjoin projects or remand permits back for additional study. Indirectly, agencies (or private investors) may delay taking action if ongoing or threatened litigation exist out of fear that a court may intervene once the project has started. Delay is a powerful tool to block projects entirely, reduce local political support, or increase costs due to inflation (Friesema & Culhane, 1976). Moreover, case precedent to date has not clearly delineated the level of analysis required to avoid a court order for additional study. The procedural requirements of NEPA remain the subject of frequent federal litigation. Factors such as judicial ideology and presidential politics continue to have significant impacts on NEPA litigation outcomes (Adelman & Glicksman, 2018). Beyond the direct impacts of litigation, even the threat of a court challenge may lead federal agencies to significantly expand environmental documentation to “cover” any possible topic to “litigation proof” the study (Smillie & Swartz, 1997). Understanding the determinants of permitting time is thus important. As we discuss below, there are several levels of review under NEPA. We focus specifically on the environmental impact statement (EIS) process, which is critical for large infrastructure projects and other actions with major environmental impacts.

Despite NEPA’s importance, research on the costs of completing an EIS, the drivers of costly delays, and on compliance more broadly, is limited. A 2014 Government Accountability Office study found that most federal agencies did not track direct NEPA-related costs in terms of staff time or contractor fees (GAO, 2014). Exceptions do exist, however. The Department of Energy reported that it spent $220 million in contractor costs on 38 EISs between 2003 and 2012 (DoE, 2013). Such direct costs of EIS preparation are likely to be small relative to indirect compliance costs. Those include project delays that—in addition to increasing costs—postpone public enjoyment of project benefits. Indirect costs also include the cost of uncertainty surrounding EIS process completion.

One of the challenges of conducting empirical research on NEPA and major federal actions is that the costs and durations of these actions, as well as the costs and durations of their environmental studies, are impacted by a wide range of political, economic, social and environmental factors. Quantifying the impacts, in either costs or benefits, of a law with NEPA’s scope is thus challenging. Although our study contributes to the growing body of research on NEPA’s impact on federal actions, it is descriptive in nature. Despite this natural limitation, we find an appraisal of NEPA’s costs and benefits to be a vitally important research endeavor.

We proceed as follows. Section 2 provides more detail on the process that agencies must follow to complete the NEPA process from NOI to ROD. Section 3 describes the data collected and our study’s limitations. Section 4 posits several hypotheses regarding the determinants of completion duration, as well as our empirical approach. Section 5 reports our estimates and analyzes model validation used in this specific case. Section 6 provides information on other factors potentially affecting approval duration. Section 7 summarizes and concludes.

2 NEPA: background and process

2.1 The history and structure of NEPA

NEPA was America’s first comprehensive environmental law (Liroff, 1976). Congress spent 10 years prior to NEPA’s enactment discussing how to address growing environmental concerns (Luther, 2008). NEPA was thus the culmination of many years of legislative deliberation and framed the country’s overall environmental policy for decades (Council on Environmental Quality, 2007).⁵

NEPA gives specific mandates to all federal agencies and makes them responsible for its implementation. Each agency is responsible for developing its own NEPA procedures. NEPA also created the Council on Environmental Quality (CEQ) within the Executive Office of the President. Although CEQ has numerous responsibilities, its primary task is overseeing federal agency implementation of NEPA. CEQ must approve each agency’s procedures to ensure compliance. NEPA regulations require that every federal agency comply with Sect. 102 of NEPA, which establishes the EIS requirement.

As interpreted in regulation, NEPA creates three levels of review based on the action’s expected environmental impacts: (i) a categorical exclusion determination (CE); (ii) an environmental assessment/finding of no significant impact (EA/FONSI); and (iii) a required Environmental Impact Statement (EIS). An EIS, the highest level of review, is required for the most impactful federal projects. Agencies prepare approximately 170 EISs, 10,000 EAs, and apply categorical exclusions to approximately 100,000 CEs to federal actions annually.

Courts have determined that privately financed projects requiring federal permits or land leases from a federal agency are also subject to NEPA. In those cases, the federal agency undertaking the action is responsible for completing and certifying the NEPA study, although the private sponsor is naturally involved in the process (Hayes & Hourihan, 1985).

Agencies also complete EAs for federal actions with uncertain environmental impacts. If the agency determines likely environmental impacts are minimal, it issues a FONSI and the action may proceed.⁶ If instead the lead agency determines that impacts may be significant then the agency must complete an EIS. The third and most common level of review is the Categorical Exclusion (CE). A CE applies when no significant effect is expected and thus no further impact assessment is required. Federal agencies are required to list the types of projects that may be eligible for a CE (Luther, 2008).

NEPA includes broad language and, as a procedural statute, relies on citizen-lawsuits for enforcement and the generation of case law. NEPA combines that with CEQ guidance to refine how NEPA is applied in practice. Because of this, NEPA has evolved over time as standards or practices have changed in response to case law or new guidelines.

NEPA was uncontroversial when passed by Congress. Indeed, many legislators viewed their vote as largely symbolic. Although NEPA’s Congressional architects had differing views on how its aspirational language would impact federal agencies’ statutory authority, they would later reflect that they didn’t understand what they were passing (Liroff, 1976).⁷

Despite this, and some of the potential costs we discuss below, NEPA has generated benefits in the United States and even globally. It was the first major procedural environmental law of its kind, and its passage has influenced similar laws and regulations at the state level and globally. NEPA’s requirement for environmental study facilitated development of the environmental science and planning industries, as well as the practice of environmental law. It has in turn improved our understanding of the impacts of human actions on the environment while incorporating the language of environmental impacts into U.S. case law and administrative procedures. It has also promoted the good-governance aims of citizen input into public actions, transparency, and responsive government.

2.2 The NEPA process

The NEPA process begins with the relevant federal agency’s plan to take an action addressing a public need. When multiple federal agencies are involved in an action, a lead agency and cooperating agencies are designated.⁸ In the first stage, the federal agency determines whether the project is subject to a CE or not. It does so based on experience and NEPA procedures applying to that agency. If a categorical exclusion is inappropriate, the federal agency must choose between an EA versus moving directly to an EIS. In an EIS, agencies study the proposed action’s environmental impacts as well as feasible alternatives, including a “no build” alternative in the case of an infrastructure project. The EIS process begins with a Notice of Intent (NOI). The NOI states the intention to undertake an EIS for a certain project. It also explains the project and the federal agency’s alternatives. It further defines the scoping process required by NEPA regulations and how to involve others in that process.

Following the NOI, the federal agency prepares a Draft EIS. The Draft EIS is published and subject to public comment. Public meetings are then held. Once the time for public comments and meetings has elapsed, the federal agency must prepare a Final EIS (FEIS) to address issues arising at the Draft EIS stage and incorporating responses to comments. Both the Draft EIS and the Final EIS are communicated to the public through a Notice of Availability (NOA).

The final step is the Record of Decision (ROD). The ROD generates the final EIS process document. It reviews the factors considered in the EIS process and the actions that must be taken. Projects may also require a Supplemental EIS, which is appropriate if there are substantial changes in the project or new circumstances to be analyzed.

The ROD represents the final decision by the federal agency for an action, allowing the selected alternative for the project to move forward. In practice, however, NEPA timelines could be impacted by other political or economic factors. The ROD could select the “no build” alternative, effectively ending the permitting process with an abandoned project. The NEPA analysis could also be challenged via environmental litigation after the ROD, and later remanded for additional study. In other cases, long delays after a Final EIS is completed but before a ROD is published could represent other changes, including electoral changes that cause a project to lose political support, or the project could lose critical sources of funding. These changes are not necessarily represented by a “no build” decision but could also manifest as a long delay during the permitting timeline. When long delays occur, agencies (or federal courts) may conclude that additional, supplemental environmental studies are needed simply because the original environmental study is now outdated. NEPA is thus best understood as a central component but certainly not the only requirement of infrastructure development in the United States.

There is now a body of literature on the science underlying NEPA reviews. However, there is much less on the processes, duration, and costs of NEPA studies. DeWitt and DeWitt (2008) studied EIS durations, comparing the preparation time for more than 2,000 EIS’s by their lead agency. They report an average duration of 3.4 years between NOI and FEIS, but also found that durations increased markedly over the study period of 1998–2006 (DeWitt & DeWitt, 2008). They conclude that the recommendations from a 2003 CEQ task force on ways to speed up NEPA were either ineffective in doing so or had not yet yielded results.

Little empirical data exists on the costs associated with the development of EIS’s or the NEPA process generally. Very little aggregate data has been identified where the federal government has attempted to quantify NEPA’s costs (GAO, 2014). This is due partially the difficulty of quantifying the transaction costs and other externalities associated with environmental policy (McCann et al., 2005). Although the direct costs associated with the production of an EIS could conceivably be measured and tracked by federal agencies, they rarely are in practice. Those direct costs include environmental staff or consultants required to develop the report and navigate the process in addition to any direct costs in defending litigation. Even this significantly understates the costs associated with the EIS because many of the costs are indirect. Those indirect costs include any scope changes or additional federal actions necessitated by the EIS (or the settlement of EIS litigation) and, importantly, the indirect costs of delaying the federal action in question. This is critical for large infrastructure projects, for which delays and uncertainty both entail large direct and indirect costs.

Processes and requirements under NEPA have evolved in part due to changes in CEQ policies. NEPA has also changed considerably in response to judicial actions from environmental litigation.⁹ The question of whether an EIS, rather than an EA or CE, is required for a federal action has also evolved and is regularly revisited by the courts. Although the requirement is for the anticipated environmental impacts to be “significant,” courts have avoided defining that standard with specific metrics. NEPA litigation is often focused on whether the action will have “significant” environmental impacts (Bear 1989; Luther, 2007).

2.3 Prior literature on NEPA costs

The Congressional Research Service (CRS) and Government Accountability Office (GAO) have each completed multiple reports on the costs and duration of NEPA reviews. In 2007, the CRS published a report summarizing proposals to streamline NEPA. It assessed the viability of policies to better coordinate permitting across NEPA and other environmental regulations, and to designate Categorical Exclusions and Lead Agency responsibilities more clearly, among other possible reforms (Luther, 2007). However, a 2014 GAO study is tellingly entitled “Little Information Exists on NEPA Analyses” (GAO, 2014).

Such reports also highlight the challenges of assessing the drivers of costs and duration of NEPA permitting. Agencies have estimated the direct costs, in staff time and consulting fees, spent on developing an EIS, although those are likely a small part of the overall costs of an EIS for a major federal action. If the action involves a large infrastructure project, the costs of delays and the uncertainty of environmental litigation are likely much greater but remain difficult to measure. Likewise, the time between the start of an EIS and the Record of Decision is not a function of the NEPA process alone. Planning or permitting delays under other environmental regulations may be the more proximate cause of delay. Projects could be delayed for budgeting or political reasons during NEPA, separate from the permitting process. Moreover, the end of the EIS process, as marked by a Record of Decision, does not automatically mean that the action may proceed. Environmental litigation or other issues after a Record of Decision may delay the action further.

3 Data sources and limitations

3.1 Data collection and description

Our primary data source is a novel set of EIS permits collected by the CEQ. It includes 1269 federal actions, all of which have a completed FEIS at minimum. Only 1161 of the federal actions had received a ROD and thus completed permitting when data were collected. Our main data set thus contains right-censored observations. All ROD’s were completed between the beginning of January 15, 2010 and June 7, 2018. All permits were completed during the Obama administration and the beginning of the Trump administration, though permitting started (for some outliers), as early as the George H.W. Bush administration.

The CEQ data includes the department and agency primarily responsible for the EIS, in addition to the state (or states) in which the project or federal action is located. The CEQ also collected four dates delineating phases of the EIS process based on the federal register. Those are: (i) the date of the NOI for the start of the permitting process; (ii) the date of the NOA for the Draft EIS; (iii) the date of the NOA for the Final EIS; and (iv) the date of the ROD for the end of the permitting process.

The CEQ published a report containing summary statistics as well as comparisons between EIS completion timelines in 2018 (CEQ, 2018a, 2018b). For those federal actions with a completed ROD, the mean completion time was 4.5 years, and the median was 3.6 years. The standard deviation was 3.1 years for that subset of actions. For all projects in the data set and the completed-ROD subset, the average time to complete a Draft EIS was 2.6 years, the average time between Draft and Final EIS was 1.5 years, and the average time from Final EIS to ROD was 0.43 years.

We supplemented CEQ data with information on project conditions or federal action. We coded projects based on whether they were infrastructure or other large construction projects, which comprises about half of the full-permit data set. Other permits include area management plans, power-plant license renewals, or other major environmental actions that are important environmentally but do not entail considerable capital investment. Projects were also coded by whether the permit or federal action was for a project backed by private investors rather than a pure public work. Those constitute about 15% of the data set, mostly for energy-sector projects or license renewals.

We used data on Supplemental EIS’s to determine which projects eventually required completion of a Supplemental EIS. Supplemental EIS’s are used to study other environmental factors as well as alternatives to the action in question. Federal agencies complete them after the original EIS, based on the agency’s determination (or a court order) requiring additional study. We also coded projects based on inclusion in the Federal Permitting Dashboard. The federal government created the dashboard in the Fixing America’s Surface Transportation Act (FAST-41) of 2015 to provide a permitting resource for large and/or complex infrastructure projects. Its goal is enhancing coordination and accountability by making the environmental review process more efficient.

We also coded projects by geography. We include an indicator for projects or actions crossing multiple states to determine if interstate projects take longer to permit. For single-state permits, we added an indicator for projects in states with restrictive state environmental laws. We here borrow from Brooks and Liscow (2019), who classify the restrictiveness of state environmental laws based on four factors: (i) the degree to which uninjured parties can challenge a project in court; (ii) the threshold for triggering state environmental review; (iii) the degree to which private actions subject to government permitting also require environmental review; and (iv) the procedural opportunities for citizen involvement under state environmental laws. Restrictive states under those criteria are California, Massachusetts, Minnesota, New York, and Washington.

We also created a data subset using 556 EIS’s based on the availability of page-count data for those studies. We included the number of pages of the Draft and Final EIS for each project, distinguished by the study’s main body and the appendices. We used that data set to examine the correlation between the duration of each phase (or the total NEPA process) and the length of the environmental statement itself. The average number of pages of FEIS’s in this data set is 673 pages for the main body and 1693 in total (main body plus appendices), with a large standard deviation for each.

Our variable selection was based on a combination of data availability and our hypotheses regarding the key drivers of environmental study durations. We were particularly interested in variables related to stakeholders with a direct or indirect interest in the outcome of the federal action. Survey-based studies of federal NEPA permits indicate that stakeholder factors such as the “likelihood of litigation” or “political attention” may impact the permitting process even more than “likely environmental impacts” (Mortimer et al., 2011). We therefore included categorical variables for construction and private sponsorship of the project being assessed, as well as variables for projects crossing multiple states or projects in restrictive states. We further included indicators of a Supplemental EIS and whether the project is in the Federal Permitting Dashboard to examine if these administrative decisions impact permit durations.

We utilize several ancillary data sources. Under the American Recovery and Reinvestment Act (ARRA), the CEQ collected and aggregated quarterly data on federal permitting for all projects receiving stimulus funding from May 2009 through November 2011. Due to the timing of EIS’s in our main data set, there is negligible overlap between those two data sources. From 2001 to 2013, the CEQ also collected data on NEPA litigation against federal agencies.¹⁰

3.2 Limitations of our data

Our data set is limited to EIS’s or permits for federal actions having major environmental impacts. Those represent less than 1% of federal permitting actions. We do not examine the drivers of permit durations for CEs or EAs. Our data set also consists of permits completed in a limited window between 2010 and 2018. We are thus unable to study broader trends in EIS duration over time beyond that window or between presidential administrations. Importantly, we define permits in our data as “complete” when the sponsoring agency publishes a ROD. However, this does not mean that the federal action in question will be taken immediately or, for an infrastructure project, that construction will begin. In some cases, for political, budgetary or other reasons a federal action may not be taken even after ROD publication. Environmental litigation may also be filed after a ROD is published, further delaying the action in question (Table 1).

Table 1 Summary statistics for NEPA permitting decisions (in years)

As shown in Table 2, DoT exhibits longer durations because almost all its EISs are for major public works. The other departments (including DoD, which completes public works through USACE) have a mix of major construction permits and other non-construction permits.¹¹

Table 2 Summary statistics for NEPA permitting decisions by project type (in years, all permits)

4 Hypotheses development

We analyze permitting time for projects entailing heavy construction, projects that benefit (or are on behalf of), private investors or corporations, projects executed in states with restrictive environmental laws, projects that are included in the federal permitting dashboard, and projects for which a supplemental EIS was prepared. We test the following five hypotheses:

H1: Permits for projects that entail heavy construction (for public works or otherwise), will take longer to complete relative to major environmental permits that do not entail construction.

H2: Permits that benefit (or are on behalf of) private investors or corporations will be completed faster than permits benefitting the public only.

H3: Permits in states with restrictive state environmental laws will take longer to complete than permits for projects in other states.

H4: Permits for projects included in the federal permitting dashboard will be completed faster than other permits.

H5: Permits that include a Supplemental EIS after the primary EIS is completed will take longer to complete relative to projects not completing a supplemental EIS.

We test these hypotheses among others based on our understanding of the EIS process in general. Although all EIS processes are likely to be environmentally important and attract participation from environmental groups, projects with heavy construction are likely to attract more citizen participation and engagement because the projects have direct, immediate impacts. Such projects are also likely to attract participation from citizens or groups with an interest outside of the project’s environmental impacts, such as their appraisal of a project’s community impacts or economic impacts. We include projects with a private-investment interest (H2) in order to test whether a concentrated economic interest in a private party has an impact on permit durations as compared to permits for public projects, in which the cost increases associated with delays are borne primarily by taxpayers. We include designations for restrictive states (H3) and the federal permitting dashboard (H4) to determine whether state laws impact federal permitting timelines and whether the dashboard correlates with shorter permits to improve bureaucratic coordination in permitting, as intended. Finally, we include a designation for Supplemental EISs as this is an indicator of a conflict, a court order, or another issue causing an agency to revisit its original environmental analysis.

5 Empirical approach and estimation

We use Cox proportional hazard to estimate the impact on EIS duration for permits involving major construction, projects supporting a private investment relative to a public work, permits for projects located in restrictive states, permits that are included in the federal permitting dashboard, and permits that published a Supplemental EIS after completing a Final EIS but before a ROD.¹²

We estimate the effects of those variables on permitting-process duration, from NOI to ROD. We estimate additional Cox regressions for the duration of each permitting phase to examine how those factors impact duration during a particular phase. The data sets for the full permitting process and the duration from final EIS to ROD are right-hand censored, as 1161 of the 1269 had published ROD’s at the time the data set was created, while the remainder did not. All permits in the data set have completed Draft and Final EIS’s, so the data sets for those regressions are not censored.

We report summary data regarding the average duration of permits based on the above factors in Table 3. We report hazard coefficients and confidence intervals for variables in each regression in Table 4. The dependent variable is the hazard rate with the duration of the permitting process (or phase of the permitting process for those regressions) in years. The hazard rate is the conditional probability of completing the permit or phase of the permitting process. In our duration analysis, the coefficients listed in Table 4 measure the effect of that categorical variable on the expected log of the conditional probability of a permit or phase being completed. A negative coefficient indicates that, for permits including that categorical variable, the hazard rate decreases or alternatively the duration of that permit or phase increases.

Table 3 Average permit duration by phase and covariates

Table 4 Cox regression results for permit characteristics

For regressions on total permit duration, the categorical variables for construction, private projects, projects in restrictive states, and projects in the federal permitting dashboard are all significant. The coefficient for private projects is positive and large with a parameter of 0.689, or a relative hazard ratio of 199% for private projects. Therefore, permits for a privately financed project are likely to be completed faster than public-sector permits. As expected, the coefficients for construction and restrictive states are negative, indicating longer permit durations. We report coefficients of -0.33 for construction and -0.14 for restrictive states, or relative hazard ratios of 72% and 87%, respectively. The coefficient for permits in the federal permitting dashboard is also negative, revealing longer durations for those projects as well, although they constitute a small portion of the data set. Figure 1 includes Cox adjusted hazard curves for construction, permits for private investors, and permits for projects in restrictive states to illustrate the effect of each of those variables on permit duration.

Fig. 1

Cox Proportional hazards model – adjusted survival curves for select categorical variables

Figure 2 includes histograms of permit durations segmented by three of our categorical variables: construction, private sponsorship and restrictive states as these more intuitively illustrate the average permit durations between the groups in our data set. Due to censorship, hazard ratios do not translate directly into differences in average permit duration. For instance, our hazard ratio of 199% for private projects means that those projects have almost twice the probability of completing an EIS in a given time period relative to the other permits in our dataset. In our dataset, private projects have an average permit duration of 3.66 years compared to the average of 4.47 years across all of the completed permits. For construction projects, our hazard ratio of 72% is the likelihood of those projects completing their permit in a given time period, relative to the other permits in our dataset. Construction projects naturally also have a longer average duration of 5.21 years.

Fig. 2

Histograms of permit durations for select categorical variables

Cox’s regressions for each phase of the permitting process estimate the impact of a categorical variable on permit actions. Only construction was significant, with a negative coefficient, for each phase of the permitting process in addition to the process in its entirety. The coefficient for private-sector permits is positive and significant in publishing a Draft and then a FEIS.¹³ Although significant for overall permit duration, permits for projects in restrictive states are not significant in either of the first two phases. They instead are negative and significant only between the completion of the final EIS and the ROD. Although not statistically significant in the overall permitting process regression, the categorical variable indicating that a supplemental EIS for the permit was published is significant in several sub-phases. It is significant with a positive coefficient (or shorter phase duration), during the first phase in developing a Draft EIS. It is also negative and significant during the final phase.

6 Estimation, discussion and model validation

6.1 Discussion

Identification regarding the outcomes of large infrastructure projects or other public actions is important. The durations of environmental permitting for large public works and other federal actions are obviously influenced by numerous variables, and the results of projects or permits are idiosyncratic. Two seemingly similar federal projects may have widely varying environmental impacts, stakeholder groups, political situations, macroeconomic environments and otherwise. Our variables relate to the incentives of project stakeholders during the EIS development process, but other factors could also impact EIS duration.

We expect that permits for projects or assets backed by private investors will be completed faster because investors have a concentrated financial incentive to expedite permitting. However, even for privately financed projects, the lead federal agency remains responsible for completing and certifying the EIS and even for selecting the preferred alternative, with information from the private party. Many private-sector permits are for projects in the energy sector such as power plants, transmission lines, or pipelines. We expect permits in these sectors to take longer to complete because they are for particularly challenging or environmentally impactful projects. Our estimates are inconsistent with that view, however. Based on permit-phase estimates, it appears that private permits impact durations primarily in publishing a Draft and then a Final EIS, since that coefficient loses significance during the final phase. Durations are thus shorter for developing the environmental documentation and studies required for the EIS, but not necessarily shorter in achieving a final approval. This indicates that private project sponsors may be able to hasten permit development during the drafting phase, but that private project sponsorship is not necessarily associated with shorter durations during the public consultation phases of the permitting process.

Permits for major construction projects or public works take significantly longer than other environmental permits. As noted, this is not necessarily due to greater environmental impact. Although large public works or construction projects are likely to have major environmental impacts, all large permits are for federal actions that generate significant environmental impacts. Moreover, we have no measure of the cumulative environmental implications of each permit to facilitate comparison of permit durations. Projects involving construction or public works are likely to attract more stakeholder engagement, including popular opposition or environmental litigation. This may drive longer durations during the second and third phases of the permitting process. Greater stakeholder involvement can cause longer durations during the EIS drafting phase. That may happen if sponsoring agencies increase outreach or increase the environmental factors or alternatives studied in anticipation of challenges to the EIS in later phases. Additional research on the frequency of litigation or volume of stakeholder engagement for environmental permits based on construction may help determine the specific causes of longer durations in such cases.

Permits for projects in restrictive states have longer durations. Although this is not surprising, technically standing and other procedures for federal projects are governed by NEPA, rather than state law, so the impact of restrictive state environmental laws on permit durations must be indirect. Restrictive states were not designated because their state environmental laws are particularly onerous (though they may be). Restrictive states were instead selected because they have state environmental laws with lower standing requirements for legal challenges, lower thresholds for triggering state environmental reviews, and more procedural opportunities for citizen involvement in the permit approval process. The phase-specific Cox regressions do not indicate that durations are longer in restrictive states for the Draft or Final EIS phases. Only the phase between the Final EIS and the ROD is statistically significant. Our projects in restrictive states also have a higher rate of right-censorship than the full data set, which is not reflected in our histograms of completed permits (see Fig. 2).

There are alternative explanations for longer durations in restrictive states, and longer durations during the final phase of the permitting process. Restrictive states may have larger communities of stakeholders or a larger environmental litigation industry that results in greater participation, or a heightened threat of litigation during the permitting process, even when those permits are governed by federal rather than state environmental law. If federal permits receive higher opposition or more litigation relative to other states, it may lead to increased “litigation proofing” of EISs and thus longer permit durations. Environmental interest groups and other stakeholders may also be better organized and more active in states with restrictive state environmental laws because of there are many more opportunities to intervene in the environmental permitting process in those states.¹⁴ Finally, NEPA delays could occur in restrictive states if portions of a project are also governed by state environmental law.

We ran several additional tests on the influence of states on federal permitting results. We originally hypothesized that projects spanning or involving multiple states would have longer permit durations as these projects would have more local stakeholders and more coordination challenges. However, we did not find a relationship between mutli-state projects and longer permit durations. This is another indication that permit timelines may be less driven by the complexity of the project or stakeholder network involved in it, and more by conflict between various stakeholder groups.

In order to further investigate our finding that federal permits in Restrictive States take longer, we ran an additional regression to account for state political leanings. We ran this regression to investigate whether state environmental laws impact permit durations when controlling for state political leanings, because all of the Restrictive States in our regression (CA, WA, MA, NY, MN) are also Democrat leaning states. We approximate state political lean with a categorical variable indicating whether the state voted Democrat in the 2016 presidential election. Our regression results are shown in Table 8. The Restrictive State variable is still significant and correlated with longer permit durations. Our variable for state political lean is correlated with shorter permit durations, though it is only significant at the p < 0.1 threshold. We interpret this result as showing that Democrat-leaning states actually have shorter permit durations when controlling for the Restrictive States. This indicates that state-level environmental laws, rather than state-level political leanings, control for federal environmental permit durations.

Despite being a small fraction of our data, projects in the federal permitting dashboard have longer durations relative to non-dashboard projects. This is counterintuitive because the permitting dashboard was created to improve or streamline the project approval process. Importantly, the permitting dashboard includes many different federal permits in addition to EIS’s. We assess only a small portion of the dashboard. Our finding may also be due to selection bias if projects included in the federal permitting dashboard are selected because they are very large, impact multiple stakeholders or agencies, or entail significant environmental impacts.

The publishing of a Supplemental EIS’s did not have a statistically significant impact on overall permit duration, although it was significant in several phases. The coefficient is statistically significant and associated with longer durations during the final phase of the process and shorter durations during the first phase of the process. The longer duration in receiving a ROD is understandable since this final phase is when the need for a Supplemental EIS is identified and the additional environmental study and report are completed. The finding that the need for a supplemental EIS is correlated with shorter durations during the Draft phase of the EIS is unsurprising. However, this variable is associated with a very small portion of the data set. We next discuss model validation.

Our study focuses on the duration of the NEPA process, rather than the costs of NEPA compliance. Although longer permit timelines may increase costs and delay project benefits, it is not possible to directly correlate longer permit durations to a specific rate of cost increase for several reasons. While the direct costs of permits (for environmental consultants, staff time, or legal fees) increases as permits grow in duration and pages of study, this is likely an extremely small fraction of the total costs of delayed federal actions, most of which are indirect. For large infrastructure projects, during a period of high inflation, the costs of even relatively short delays can be vast. In aggregate, there are strong indicators of Baumol’s cost disease in the U.S. construction sector since 1970.¹⁵ Both labor and total factor productivity data for the construction sector indicate that the sector has become significantly less productive since 1970 despite productivity gains in prior decades.¹⁶

6.2 Model validation

Cox Proportional Hazard incorporates several assumptions that must be validated to ensure the model describes the data accurately. The proportional hazards assumption is key. It means that the ratio of hazards for data subsets is constant over time. We test this assumption in our models using Schoenfeld residuals and by examining survival curves graphically. The second major assumption is that independent variables enter the model linearly. Since our model consists of only categorical variables, this assumption need not be tested. We also test for the possibility of particularly influential observations, or outliers, primarily by examining deviance residuals.

In order to test the assumption of the hazard ratio being proportional with time, we calculate Schoenfeld residuals for both the primary Cox PH model (p = 0.317 χ2 = 5.89 on 5 degrees of freedom) and the model incorporating page counts that will be discussed below (p = 0.292 χ2 = 7.32 on 6 degrees of freedom). The Schoenfeld residuals for neither model is statistically significant, which indicates that the hazard ratio is proportional over time. We thus accept the proportional-hazard assumption for each model.

To examine the effect of outliers, we produced plots of the impact on coefficients of the removal of individual observations. We produced a plot of the deviance residuals, which is symmetric around zero. The largest magnitude coefficient impact for most independent variables was small in comparison to the coefficients themselves. In both Cox models, it is large and negative for a few observations of both the Supplemental EIS and the Permitting Dashboard, reaching − 0.1 in two cases in the primary Cox PH model. This is still relatively small compared to the Permitting Dashboard coefficient, but large in comparison to the Supplemental EIS coefficient. This variable is not statistically significant over the full permit duration timeline in our regression, however. The relatively higher impact of outliers for those variables may be because each represents a small portion of the total data set.

7 Other factors influencing permit duration

7.1 Page count and permit duration analysis

7.1.1 Rationale to study the relationship between page count and permit duration

We next examine EIS page counts. Like permit durations, the number of pages in an EIS has increased over time. Like permit durations, page counts have also been a concern of policy makers (as well as the CEQ), for a long time. In fact, the CEQ included a main body page limit of 150 pages for an EIS in its 1978 guidance to agencies. EIS page counts are a proxy for federal action’s direct permitting costs.¹⁷

Finally, it is useful to explore the relationship between EIS page counts and duration by permit phase. The number of pages in a Draft EIS or FEIS is likely correlated positively with the duration of that phase. Indeed, Smillie and Swartz (1997) suggest that EIS page counts are increased intentionally to provide a defense against expected litigation. We explore the effect of increased page counts in one phase of the permitting process on durations in future phases.

We report summary information on the subset of EIS’s with page count data is included in Table 5. Final EIS’s had an average page count of 1693, with the average length of the body of the report at 673 pages and an average length of appendices at 1020 pages. Most subsets corresponding to our categorical variables have higher page counts, on average, than the data set as a whole. For example, Final EIS’s in restrictive states have an average total page count of 2273. Figure 3 shows a graph of the data set by total page count of the FEIS and the date of the record of decision. The number of pages of the EIS’s in the study have increased over time for the federal actions in our data set.

Table 5 Permit duration and page count summary data

Fig. 3

Final EIS total page count and date of record of decision

7.1.2 Discussion of estimates on page count and permit duration

We estimate how the number of pages in an EIS affects the permitting process. We use the same variables as above but focus on the time between the publication of the FEIS and the ROD. We posit that federal actions with more voluminous studies take less time during the post-EIS phase. Variables from the full data set include: (i) whether the action involves considerable construction; (ii) whether it benefits a private entity; (iii) whether it is in the permitting database; and (iv) whether the action is in a restrictive state.

We report Cox Proportional Hazard estimates in Table 6. The total Final EIS page count is statistically significant and associated with a lower hazard rate for both the total permit duration and the final phase of the process between the Final EIS and the ROD. The dependent variable is measured in years and the coefficient represents the impact of an increase of only one page in the Final EIS, which may explain the relatively small coefficient.

Table 6 Cox regression results including page count data

Higher page counts in an EIS are associated with a lower hazard rate, or longer permitting times. That may be unsurprising since larger EIS’s likely take longer to complete. However, page length is also associated with longer durations of the final phase of the permitting process, from FEIS to ROD. The document itself is already completed at this stage, but durations are still longer for high-page-count permits. This suggests that “defensive” EIS’s that have very high page counts are not effective in reducing permitting times in the later phases of the process.¹⁸ Higher page counts may also be indicative of the general amount of conflict or opposition to a proposed federal action. Permitting times are thus longer during all phases and not simply those phases in which the study is drafted.

Several categorical variables lose significance when total page counts are included. That is unsurprising since those categorical variables are correlated with higher page counts. Federal actions involving a major public work, or those in a restrictive state, those in the permitting dashboard, all have higher average EIS page counts. Permits benefiting private projects are positive and significant in the regression on total permitting time. Private projects have shorter permit durations irrespective of the number of pages of the EIS.

7.2 Frequency of EIS permit completion

We next examine the frequency with which the federal government completes permitting for EIS actions. We seek to determine whether the federal government completes EIS’s more frequently during certain time periods or during stimulus programs when more federal actions are undertaken. We use our primary data set, which includes 1,161 RODs. We also incorporate data from quarterly NEPA reports published by the CEQ during the American Recovery and Reinvestment Act (ARRA) that track the number of EIS’s completed over time.¹⁹

Figure 4 shows a bar chart of the number of RODs published by the federal government during each of the weeks included in the data set. The federal government published 2.94 RODs per week on average. The number of RODs per week appear to be relatively consistent over time, with a few notable exceptions. The early period does not reveal an increase, although this is likely because most of the EIS’s completed under ARRA were complete before 2010.

Fig. 4

Number of RODs published per week

There are several clear changes associated with the transition from the Obama to Trump administrations. There are two weeks that are significant outliers in which 12 or more RODs were completed. Both occurred during the last week of September in 2012 and 2015. The federal fiscal year ends in September 30th. Spikes may be due to year-end funding requirements for their respective federal actions. In late 2015, the Bureau of Land Management also published eight RODs on the same day that were all various land-use-plan amendments associated with the greater sage-grouse.

There is a large spike in ROD completion during the presidential transition. It occurs between the election of President Trump and his inauguration. In fact, more unconnected RODs were published on the day prior to the inauguration than on any day during our study period.²⁰ After the inauguration, the rate of RODs decreases.

Table 7 includes estimates of the number of RODs completed per week. We used a negative binomial regression due to over-dispersed data. The weeks during the presidential transition period display a positive coefficient of 0.602 and weeks during the Trump administration have a negative coefficient of − 0.365. On average, the weeks during the transition period show an increase in the expected number of RODs of approximately 83%, while the weeks following the inauguration of the Trump administration include a decrease in the expected number of RODs by 31%.

Table 7 Negative Binomial Model Regression for ROD’s per Week

An initial decrease in EIS completion rates between presidential administrations is unsurprising. Agency leadership turns over and the new administration begins to implement its priorities. The CEQ data includes less than 1.5 years of the Trump Administration’s tenure. We are thus unable to conclude that the Trump Administration overall completed EIS’s at a lower rate compared to the Obama Administration.²¹

The spike in RODs between the election and inauguration of President Trump may be due to increased uncertainty around federal budgets or approval processes across administrations. If federal funding for a project or other action is delayed until EIS completion, then federal agencies may quickly issue a ROD if there is a concern that funding will be removed by the next administration. Local project sponsors (such as state or city agencies) may also push federal agencies to issue RODs due to uncertainties surrounding new leadership. Uncertainty between administrations may also cause federal agencies to make decisions on RODs when there is a conflict over an EIS. If an EIS is delayed due to conflict between stakeholders over the preferred alternative for the project, the prospect of withdrawal of federal funds post-inauguration may lead some parties to compromise.

Examining the federal actions that led to completed permits during the lame-duck period generates added insights. The 55 projects that received RODs between the election and inauguration of President Trump had an average permit time of nearly 5.8 years—more than a year longer than the average duration for the full data set.²²

7.3 EIS completion during ARRA

Due to the Obama Administration’s focus on fiscal stimulus, a key focus of its program was to get infrastructure projects “shovel ready” and under construction as soon as possible. Between late 2009 and the end of 2011, the CEQ published a quarterly report of the federal permits that were completed and still in process under the ARRA. The final report, in November 2011, included 841 complete EISs and 18 still pending. Including CEs and EAs, 275,636 Federal permitting actions were tracked by the CEQ for ARRA-funded projects.

Given the duration times from our (mostly post-ARRA) data set, the completion of 859 EISs over a 2 year federal stimulus program seems challenging. There is evidence that very few EISs were actually completed under ARRA. Figure 5 shows the reported completed and pending EISs for ARRA projects by the CEQ over a 2 year period (CEQ, 2018a, 2018b). There are initially under 200 completed EISs in the report. After one quarter the number jumps to 842 completed EISs and 97 EISs pending. The number fluctuates further the following quarter, presumably as some projects are removed from ARRA and others are added. For large projects requiring an EIS, rather than complete permitting faster, projects were selected that already had an EIS completed.

Fig. 5

Reported complete and pending EIS under ARRA (CEQ)

Because the administration wanted to stimulate the economy as fast as possible, it was clear that only “shovel ready” projects would be prioritized under ARRA. For projects requiring an EIS, “shovel ready” essentially meant “NEPA compliant.” The CEQ NEPA reports do, however, illustrate the challenges as well as a possible adverse-selection problem in implementing a Keynesian stimulus program for projects that require an arduous or long-duration environmental permitting process. Projects that already had a completed EIS and were also selected for funding under ARRA were not funded via other sources. That is, the federal, state, and local governments declined to prioritize those projects for funding prior to ARRA, despite such projects already holding a completed EIS. It is therefore more likely that those projects had a lower forecasted economic return on investment. Depending on the difficulty of the permitting process, the “shovel ready” projects selected are likely to be those that elected officials or administrators believe have low economic value.

8 Summary and conclusions

This study examines durations and frequencies of federal permits for actions with significant environmental impacts under NEPA. Our primary data source is EIS permit durations collected by the CEQ, which is the largest government-wide aggregation of EIS data to date.

EISs for infrastructure projects take longer than those for federal actions that do not entail large capital investments or significant construction activity. Large public works attract more stakeholder interest or opposition (environmental and otherwise) relative to other federal actions with similar environmental impacts. Other correlated factors may also cause EIS durations to be longer for infrastructure projects relative to other federal actions.

Projects benefiting private investors or corporations complete the EIS process faster relative to other EISs; this factor was particularly impactful during the drafting phase of the EIS rather than the final ROD phase. That may be because such concentrated economic interests are able to navigate the permitting process more quickly, which leads to shorter permit durations even when private beneficiaries are reliant on federal agency input and comment on EIS documentation. The high concentration of EISs in the energy sector may also be a factor driving shorter durations for those permits.

Permits in restrictive states take longer than those in other states. This is not due to state law directly however, as NEPA is a federal law. There may be additional stakeholder opposition or threats of litigation against permits in restrictive states even for projects permitted under NEPA rather than state environmental laws. Permits in the federal permitting dashboard, although a small percentage of our data set, take longer to complete. This is counterintuitive as the dashboard was created to improve coordination and speed up the permitting process. Projects that are large public works with anticipated permitting challenges are more likely to be included in the dashboard, which is one explanation for longer observed permit durations for those EISs.

For those EISs for which page counts are available, higher page counts are associated with longer permit durations. This is consistent with predictions, although higher page counts were also associated with longer durations between the permit’s FEIS and its ROD. Durations increase with page count even after the Final EIS is written. Although higher page counts are not associated with reduced durations during the final phase of the EIS process, they may reduce the likelihood of litigation after completion of the EIS. Furthermore—because it is a counterfactual – it is not clear that a lower page count reduces permitting times for a given project, all else equal.

The frequency of EIS completion (measured in RODs per week), was stable with a few exceptions. There are several weeks where many RODs were published. Both occur in September near the end of the federal fiscal year. There is also a significant and sustained increase in RODs published in the between President Trump’s election and inauguration. Uncertainty around funding availability and investment decisions between presidential administrations may lead agencies and state or local project proponents to complete permitting and allocate funding during lame duck periods. As the transition between the Obama and Trump administrations is the only transition in the data set for this study, we cannot determine if similar permit frequency spikes have occurred during other presidential transitions.

We also examined the implications of EIS permitting during economic stimulus programs that fund infrastructure projects using ARRA data. The data collected for permitting under ARRA highlight a potential adverse selection problem when stimulus funding is allocated to projects that must undergo an arduous review and permitting process such as an EIS. Elected leaders and administrators seeking immediate impact may direct stimulus funds to “shovel ready” projects that have completed the EIS process, but which were never funded by those same leaders and administrators prior to the stimulus. This may lead to the allocation of stimulus funds to projects with lower expected returns on investment.

NEPA is the foundational law of U.S. environmental regulation. For federal actions with significant environmental impacts, it requires a detailed study of those impacts, mitigation strategies, and alternatives to the proposed action. Numerous initiatives and policy changes have been proposed to streamline the NEPA administrative process for significantly impactful federal actions. NEPA is, however, more than an administrative process for federal agencies. It supplements the administrative process with opportunities for citizen and stakeholder input and oversight through judicial review. NEPA has, through case law, administrative experience, CEQ guidance changes and presidential administrations evolved significantly since it was passed more than 50 years ago. Additional data collection efforts on NEPA permit durations, costs and outcomes will lead to more informative research and better-informed policy as NEPA continues to evolve in the future (Table 8).

Table 8 Cox regression results including variable for state political leaning

Appendix 1

Durations in survival analysis are represented by a hazard function, which is the rate at which duration will end at time t conditional upon duration not ending before time t. Durations can be unknown, or censored, in two possible ways. Duration is considered left-censored if the event being observed has already occurred before observation begins for that case. Duration is considered right-censored if observation ends before the event occurs. This study includes right-censored data as some of the permits in our data set remain incomplete at the time of this publication. The hazard function λ(t) is thus represented as:

$$\lambda \left(t\right)=\frac{f\left(t\right)}{S\left(t\right)}=\underset{\Delta t\to 0}{\mathrm{lim}}\frac{P(t\le T\le t+\Delta t|T\ge t)}{\Delta t}$$

Here, T represents the duration, and f(t) is a probability density function corresponding to cumulative density function F(t), which is the probability that a variable will have a duration less than t, while S(t) represents the survival function and corresponds to F(t) in that

$$S\left(t\right)=1-F\left(t\right)=P(T\ge t)$$

For this analysis, a proportional hazard model is used to represent the hazard function as we are investigating the influence of multiple covariates on duration. Cox’s Proportional Hazards Model represents hazard as a baseline component and a component that is dependent on multiple individual covariates, or:

$$\lambda \left(t\right)={e}^{-\beta Xt}{\lambda }_{0}(t)$$

where X represents our vector of covariates and β our vector of parameters while λ₀(t) represents our baseline hazard function. The baseline hazard function can be left unspecified for Cox’s Proportional Hazards model. Other parametric models give a specific form to the baseline hazard function. An exponential distribution, for instance, assumes that the hazard function is constant and thus that hazard does not increase over time, and is thus likely inadequate for the present study. The Weibull distribution is a two-parameter variation of the exponential distribution in which hazard increases with time. The distribution specifies the baseline hazard as:

$${\lambda }_{0}\left(t\right)=\gamma \alpha {t}^{\alpha -1}$$

As demonstrated by Cox, however, the Cox Proportional Hazard Model is useful in that it is semi-parametric and estimates the partial likelihood of covariates impact on hazard without requiring an assumption as to the shape of the baseline hazard function (Cox, 1972).