Evidence-Based Recommendations for Improving National Environmental Policy Act Implementation

The National Environmental Policy Act requires federal agencies to consider environmental impacts before acting. NEPA is the Magna Carta of U.S. environmental law, a topic of intense debate, and the subject of ongoing rulemaking efforts. Prior NEPA scholarship focuses almost exclusively on Environmental Impact Statements, which account for just 1% of all NEPA decisions. Little is known about the length of time required to complete the other 99% of agency decisions, which involve a more streamlined review. This is a critical gap in the literature because NEPA compliance involves an estimated 50,000 federal decisions annually. NEPA reform, we believe, should begin with a careful understanding of NEPA practice at all levels of review.

any other environmental laws." 7 If NEPA were repealed, compliance with other environmental laws would still be required. 8 Even though NEPA is not the source of the obligation-and some delays attributed to NEPA may originate from sources external to the law itself-NEPA is often blamed for the perceived delay associated with compliance. 9 The time and effort required to comply with NEPA has engendered heated debate. 10 Efforts to "streamline NEPA" abound, and sustained calls for reforms to the Act and its implementing regulations reverberate from both sides of the aisle. NEPA's detractors malign it as the source of delays, job losses, and failures to update infrastructure. 11 Other critics characterize NEPA as "bureaucratic red-tape," 12 and as "the weapon of choice for opponents seeking to stop or delay an activity requiring federal action." . 9 CRS, NEPA: BACKGROUND AND IMPLEMENTATION, supra note 7, at 26 ("The perception that NEPA results in extensive delays and additional costs . . . can be magnified when compliance with multiple environmental laws and regulations is required. . . . The sometimes extensive reviews, documentation, and analysis required by agencies, such as the Army Corps of Engineers, the U.S. Fish and Wildlife Service, the Coast Guard, and the EPA, as well as various state regulatory and review agencies, add further to the perception that extensive delays are related to the NEPA process. Such 'delays' may actually stem from an agency's need to complete a permit process or analyses required under separate statutory authority (e.g., the Clean Water Act or Endangered Species Act), over which the lead agency has no authority."). See also id. at 27-28 (reporting the results of a survey of the Department of Defense, the Department of the Interior, and the Forest Service in which respondents identified "factors 'outside the NEPA process'" "as the cause of delay between 68% to 84% of the time"). NEPA's admirers are no less passionate, heralding it as the Magna Carta of environmental law. 14 They believe that "public involvement and careful consideration of alternatives has produced better outcomes-for the agencies themselves, for the nation, and for the human environment." 15 Anecdotes, rather than data, however, drive these characterizations. 16 When asked to review various NEPA compliance issues, including (1) the number and type of NEPA analyses; (2) costs and benefits of completing the analyses; and (3) the frequency and outcomes of litigation, the Government Accountability Office concluded that very little information exists regarding these issues. 17 Absent information, most recommendations for NEPA reform have historically been loosely moored to empirical data. The research that does exist generally focuses on one aspect of the law-Environmental Impact Statements (EISs)-which constitute a very small percentage of the law's application. 18 We endeavor to advance the debate by providing empirical evidence of how NEPA functions at all levels of analysis, studying more than 41,000 U.S. Forest Service NEPA decisions from 2004 through 2020. We describe Forest Service practice implementing the law, and we seek to identify sources of delay within the process by using a regression model that analyzes the year a project was initiated, the level of analysis applied, 19 the activities involved in the action, and the region conducting the analysis. We also explore indications that some sources of delay are external to the NEPA process. We then use those observations to provide recommendations for improving NEPA efficacy.
(pertaining to the Full Committee oversight hearing titled, "The Weaponization of the National Environmental Policy Act and the Implications of Environmental Lawfare"). 14 MANDELKER ET AL., NEPA LAW AND LITIGATION, supra note 10, § 1:1. 15 Russell E. Train, Foreword to ENV'T L. INST., NEPA SUCCESS STORIES: CELEBRATING 40 YEARS OF TRANSPARENCY AND OPEN GOVERNMENT 3, 4 (2010). 16 GAO, NEPA: LITTLE INFORMATION EXISTS, supra note 3, at 7 ("Governmentwide data on the number and type of most National Environmental Policy Act (NEPA) analyses are not readily available, as data collection efforts vary by agency."). 17 Id. at GAO Highlights (sidebar describing "Why GAO Did This Study"). 18 See generally NAT'L ASS'N OF ENV'T PRO., 2019 ANNUAL NEPA REPORT OF THE NATIONAL ENVIRONMENTAL POLICY ACT (NEPA), https://naep.memberclicks.net/assets/annual-report/2019_NEPA_Annual_Report/NEPA_Annual_Report_2019.pdf [https://perma.cc/C9G4-57HD] (providing statistics on preparation times and other information for EISs filed in 2019 and providing link to archived reports from previous years). 19 As described in more detail in Section II.A., NEPA requires different levels of analysis depending on the significance of environmental effects: (1) an Environmental Impact Statement (EIS), which is the most searching level of analysis preserved for actions with significant environmental impacts; (2) an Environmental Assessment (EA), a lower level of analysis for activities with less significant or uncertain environmental impacts; and (3) Categorical Exclusions (CE), the lowest level of review for activities that have been categorically excluded from detailed analysis through a regulatory or statutory determination that the effects of the action are unlikely to be significant.
Our analysis focuses on decision-making times; however, we embrace this framework with caution. Time is a convenient metric, but it is not the only metric for evaluating NEPA's effectiveness. The most important metric for regulatory reforms is how well proposed changes advance statutory objectives. The U.S. Supreme Court summarized these principles as first, "to consider every significant aspect of the environmental impact of a proposed action;" and second, to "inform the public that it has indeed considered environmental concerns in its decision-making process." 20 Regulatory reforms that do not advance these statutory aims will not help "fulfill the responsibilities of each generation as trustee of the environment for succeeding generations." 21 While we believe that reducing the burden of NEPA compliance is an important objective, that goal should not displace statutory objectives.
Our research is presented as follows. After this introduction, Section II provides background information, summarizing NEPA's statutory and regulatory structure and the Forest Service's data collection system. To its credit, the Forest Service is one of the few agencies with a comprehensive database gathering information about the NEPA process at every level of review. This dataset provides a unique opportunity to observe NEPA's functionality in more detail than has been done in the past. Using this database, we describe the Forest Service's NEPA practice, including the number of documents completed annually, the level of analysis conducted, 22 the time required to complete the analysis, and trends over time. Section III briefly describes a multi-variate regression model developed for this paper in order to test the influence of NEPA-specific factors on decision-making times. 23 It also describes quality control measures used in developing the model.
Section IV provides the regression model results. To our surprise, we discovered that the individual factors included in the regression model (level of analysis, activities involved in the action, geographic region, and year initiated) could only explain 25% of the variability in decision-making times. To understand this result, we carefully analyzed each individual factor within the regression model. Section IV.A explores the effect of level of analysis on decision-making times. Specifically, we sought to understand whether there is a predictable increase in time when a project moves from a Categorical Exclusion (CE)the least searching level of analysis-to an Environmental Assessment (EA),  (1). 22 Whether the action was analyzed in an EIS, EA, or a CE. See Section II.A. for background on these levels of analysis. 23 The NEPA-specific factors are: (1) level of analysis; (2) year of initiation; (3) activities involved in a project; (4) region. [Vol. 46:S and then to an Environmental Impact Statement (EIS)-the most searching level of analysis. Predictably, we found that an EIS generally takes longer to complete than an EA, which generally takes longer than a CE, and that this relationship remained stable over the course of the study. We also found that level of analysis is an imperfect predictor of decision-making times-a result contrary to common assumptions. A surprising number of CEs take longer to complete than the median completion time for an EA, and a sizeable number of EAs also take longer than the median completion time for an EIS. Simply moving an activity into a more expedited level of review may therefore not result in faster decisions. Thus, common assumptions about "streamlining NEPA" by avoiding EISs or expanding the use of CEs may target the wrong problem.
Section IV.B probes whether the activities involved in a project influence decision-making times. To understand what might cause delay, we focused on the top three activities that the regression model associated with longer completion times. To understand the wide variability in completion times that we observed, we reviewed the statutory and regulatory structure governing each activity, reports from the Government Accountability Office (GAO) and the Congressional Research Service (CRS), industry analysis, and other scholarship, which provided further insight into the implementation of these three activities. Our research revealed that staff availability, a lack of expertise, inconsistent funding, market conditions, and compliance with other statutory and regulatory obligations are all common sources of delay in implementing projects for each activity. We conclude that these external factors are reflected in the NEPA process even though the delays are not necessarily caused by NEPA's regulatory structure. If NEPA were the sole source of delay, we would have expected to see more consistency in decision-making times for similar activities.
Section IV.C. describes the effect of Forest Service Region on decision-making times. The regression model revealed that the Forest Service Region where the analysis was conducted had an unexpected effect on decisionmaking times at each level of analysis. Because each Region implements the same laws, subject to the same regulations, and guided by the same policies, this regional variation cannot be attributed to the statutory or regulatory structure of NEPA.
Section IV.D. examines additional factors that likely affect the variability in decision-making times observed in our research. These factors may impact decision-making times for specific activities or Regions, but they are not captured by the Forest Service data.
Section V provides specific recommendations for regulatory and administrative reforms that are grounded in the results of our empirical research.
Although our observations are based on Forest Service practice, we believe that the observations and conclusions are applicable to other agencies.

II. BACKGROUND
The National Environmental Policy Act (NEPA) 24 was signed into law on January 1,1970. Americans began to see the environment differently, and NEPA marked a sea change in federal environmental policy, declaring that it is our national policy to "encourage productive and enjoyable harmony between man and his [or her] environment; [and] to promote efforts which will prevent or eliminate damage to the environment and biosphere and stimulate the health and welfare of man. . . ." 25 Broad in scope and procedural in nature, 26 NEPA can be described as the hub from which the spokes of U.S. environmental law emanate. 27 Unlike other environmental laws that apply to specific resources like air, water, or wildlife, NEPA focuses less on the "what" and more on the "how." 28 NEPA mandates that federal agencies engage with the public, thoroughly consider the environmental impacts of their actions, and evaluate a range of alternatives before undertaking federal actions. 29 NEPA, however, "does not mandate particular results," nor does it require agencies to choose the least 24 42 U.S.C. § § 4321-347. 25 42 U.S.C. § 4321. 26 While often described as procedural in nature, Congress intended NEPA to produce substantively beneficial environmental effects. Indeed, NEPA's preamble makes this intent explicit, announcing a federal policy to "foster and promote the general welfare, to create and maintain conditions under which man and nature can exist in productive harmony, and fulfill the social, economic, and other requirements of present and future generations of Americans." 42 U.S.C. § 4331(a). 27 MANDELKER ET AL., NEPA LAW AND LITIGATION supra note 10, § 1:1 (describing NEPA as an "environmental Magna Carta that has profoundly influenced decisionmaking by federal agencies"). See also Or. Nat. Desert Ass'n v. Bureau of Land Mgmt., 625 F.3d 1092, 1100 (9th Cir. 2010) (citing Calvert Cliffs' Coordinating Comm. v. U.S. Atomic Energy Comm'n, 449 F.2d 1109, 1111 (D.C. Cir. 1971) (describing NEPA as the "broadest and perhaps most important" of environmental laws)). 28 MANDELKER ET AL., NEPA LAW AND LITIGATION, supra note 10, § 1.2; Calvert Cliffs' Coordinating Comm., 449 F.2d at 1112 ("NEPA, first of all, makes environmental protection a part of the mandate of every federal agency and department. . . . Perhaps the greatest importance of NEPA is to require . . . agencies to consider environmental issues just as they consider other matters within their mandates."). 29 Robertson v. Methow Valley Citizens Council, 490 U.S. 332, 349 (1989) ("The statutory requirement that a federal agency contemplating a major action prepare such an environmental impact statement serves NEPA's 'action-forcing purpose in two important respects. It ensures that the agency, in reaching its decision, will have available, and will carefully consider, detailed information concerning significant environmental impacts; it also guarantees that the relevant information will be made available to the larger audience that may also play a role in both the decisionmaking process and the implementation of that decision." (cleaned up)); Balt. Gas & Elec. v. Nat. Res. Def. Council, 462 U.S. 87, 97 (1983) ("NEPA has twin aims. First it places upon an agency the obligation to consider every significant aspect of the environmental impact of a proposed action. Second it ensures that the agency will inform the public that it has indeed considered environmental concerns in its decisionmaking process." (cleaned up)). [Vol. 46:S environmentally damaging alternative. 30 NEPA, in short, requires that agencies look before they leap, but it does not bar them from leaping. In addition to its environmental purpose, NEPA's procedures necessitate government transparency. In the words of Russell Train, the second Administrator of the Environmental Protection Agency, NEPA's procedures were "an experiment in governance" that brought about "a revolutionary change in governmental decisionmaking" and "opened up the federal [decisionmaking] process." 31 As the Congressional Research Service summarized, "one of the primary goals of NEPA is to give the public a meaningful opportunity to learn about and comment on the proposed actions of the federal government before decisions are made and actions are taken." 32

A. NEPA's Regulatory Structure
NEPA's crosscutting approach imposes procedural requirements on all federal actions that potentially affect the environment. Before acting, agencies must undertake a "searching and careful" 33 inquiry into potential environmental impacts, a standard that is often referred to as a "hard look." 34 Furthermore, under NEPA, agencies are obligated to inform the public of major pending actions, provide the public an opportunity to offer input, and consider carefully any input received before making a decision. 35  Under NEPA and its implementing regulations, all "major Federal actions significantly affecting the quality of the human environment" must undergo an environmental review before those actions can proceed. 46 This includes decisions authorizing projects on federal land, such as logging, mining, or livestock grazing. 47 Whether a project's impacts would be "significant" is not always clear. 48 Where a project's impacts are likely to fall below the significance threshold, an expedited review may be conducted to confirm that assumption. 49 The result is a tiered system of review where routine and environmentally benign projects undergo a truncated analysis, while larger and more complex projects can require in-depth review.
When a project's impacts are known to be significant in nature, the lead agency must complete an Environmental Impact Statement (EIS). 50 EISs represent the most searching level of review, and as discussed below, can take years to complete. 51 When an EIS is required, it is prepared in stages. The EIS preparation process begins with publication of a Notice of Intent to Prepare an EIS (NOI) in the Federal Register. 52 The NOI describes the 45 The 2020 publication of the Code of Federal Regulations contained both versions of the regulations. To distinguish between the two sets of regulations, we are silent as to date or cite to the 2019 Code of Federal Regulations when referring to the 1978 version. When referring to the revised regulations, we cite to the 2020 Code of Federal Regulations. 46 42 U.S.C. § 4332(C). 47 See e.g., Stand Up for California! v. Dep't of the Interior, 959 F.3d 1154, 1163 (9th Cir. 2020) (agencies are required to comply with NEPA for "all 'major Federal actions significantly affecting the quality of the human environment' so long as the agency has some control over preventing the environmental effects," which may include permit issuance. (citations omitted)). 48 The meaning of the term "significantly" within the NEPA context is complex. The 1978 version of the CEQ regulations (in force until September 14, 2020), defined the term in relation to "context" and "intensity," with ten factors to assess the intensity of an action. 40 C.F.R. § 1508.27 (2019). The 2020 regulatory revisions omitted the definition of "significantly" in section 1508.27 and revised section 1501.3 to include less detailed direction on the meaning of significance. See 85 Fed. Reg. 43,321-22 (Jul.16, 2020) (describing changes); 40 C.F.R. § 1501.3 (2020). On October 7, 2021, the CEQ published a Notice of Proposed Rulemaking signaling a two-phase rulemaking process to reconsider the 2020 regulatory revisions, suggesting that further changes may be imminent. 86 Fed. Reg. 55,757, 55,759 (Oct. 7, 2021). Meanwhile, practitioners strive to understand the implications of these changes. See, e.g., JAMES MCELFISH, JR., ENV'T L. INST., PRACTITIONER'S GUIDE TO THE PROPOSED NEPA REGULATIONS, (2020), https://www.eli.org/sites/default/files/eli-pubs/practioners-guide-proposed-nepa-regulations-2020.pdf [https://perma.cc/J43Y-CTDT]. 49 The expedited review could take the form of an Environmental Assessment (EA), 40 C.F.R. § 1508.9 (2018) (defining "environmental assessment" under the 1978 regulations) and 40 C.F.R. § 1501.3 (2020) (describing "when to prepare an environmental assessment" under the revised regulations). Actions that "normally do not have a significant effect" on the environment may undergo an even more truncated analysis through a Categorical Exclusion (CE actions that are contemplated, as well as the reasons for taking those actions. The NOI then invites the public (including other federal, tribal, and state agencies) to comment on issues or concerns associated with the proposed action, and to suggested alternate means of achieving project objectives. 53 After considering public comments, the lead agency then prepares a Draft EIS analyzing the impacts of both the proposed action and one or more alternative means of achieving the desired end. 54 The Draft EIS compares the impacts projected to result from each alternative against a "no action alternative" (the impacts that would result from a continuation of the status quo). 55 After another public comment period and any appropriate revisions, a Final EIS and Record of Decision (ROD) are issued. 56 If significant deficiencies are identified in a Draft or Final EIS, the lead agency may prepare a Revised or Supplemental EIS. 57 Most federal actions do not involve obviously significant environmental impacts and therefore do not require an EIS. 58 If questions exist as to the significance of likely environmental impacts, the agency will prepare an Environmental Assessment (EA) to determine whether the proposed action would cause significant impacts. 59 If projected impacts fall below the significance threshold, the agency issues a Finding of No Significant Impact (FONSI) and the NEPA review process is complete. 60 Alternatively, the agency may issue a "mitigated FONSI," which includes measures to reduce impacts to below the level of significance. 61 If an EA results in a determination that a proposed action is likely to have a significant effect, then an EIS is required.
Finally, there are numerous federal actions that are categorically excluded from the preparation of an environmental assessment or an environmental impact statement. The CEQ's NEPA regulations authorize agencies to identify categories of actions that do not normally have a significant impact on the human environment. 62 Actions that fall within one of these "Categorical Exclusions" (CEs) can be approved without an EIS or EA, provided that the 53 Id. 54 Id. § 1502.9(b) (2020). 55 Id. § 1502.14(c) (2020). 56 Id. § § 1502.9 regulations. 72 The CEQ's regulations were first issued in 1978 and remained largely unchanged until 2020. 73 Shortly before leaving office, the Trump Administration finalized wholesale revisions to the CEQ's NEPA regulations. 74 The new rules were intended to "modernize and clarify the regulations to facilitate more efficient, effective, and timely NEPA reviews by Federal agencies." 75 Efficiencies under the new rule were achieved by imposing page limits, aggressive deadlines, and modifying the requirement to consider the cumulative effects of a project. 76 The CEQ regulations required all federal agencies to revise their regulations in accordance with the CEQ's far-reaching changes. 77 Because the Forest Service's revisions had been initiated before the CEQ's revisions, the "new" Forest Service regulations did not incorporate CEQ's new regulatory changes and will again require updating-assuming that the CEQ's 2020 regulations remain in effect. The MYTR database contains a wealth of information, including (but not limited to) the project name, the Forest Service region where the project occurred, the level of analysis (CE, EA, or EIS) conducted, the date the project was initiated, the date that the decision was signed, and the elapsed time for decision-making (initiation to decision signature). The database also classifies each project based on one or more of eighteen identified project purposes; and one or more of almost fifty distinct activities. The database was designed as a tracking system to facilitate compliance with public disclosure duties. As a result, the information that it contains is specific to NEPA decision documents. Decisions are distinct from the time required to implement a project following its approval and MYTR does not track the time to implement projects. MYTR also was not designed to support statistical analysis of the Forest Service's NEPA activities. We therefore undertook the following quality control review at the outset of our analysis.
First, we excluded incomplete projects because they lacked a reviewable decision. 84 Second, we excluded projects completed before January 1, 2004 or after December 31, 2020 because data outside this window appeared incomplete. Third, we excluded thirty-five decisions documented in a "PAD" 85 because the number of decisions evaluated in a PAD was too small to evaluate statistically. shows these trends, with two caveats. First, the sharp increase between 2003-2004 likely reflects initial efforts to utilize the database rather than an increase in NEPA document production. Second, the decrease in the number of cases from 2016 onwards is amplified (particularly for EAs and EISs) because it only includes cases that were completed more quickly than the average case. We discuss this in more detail below. 84 The database identifies project status as "complete," "canceled," "in progress" or "on hold." We selected decisions that were "complete." Additionally, the database provides a date that the final decision for a project was signed. Projects without a final decision were excluded. 85 PADs are used to document that a project was previously analyzed adequately in another NEPA documents and are therefore better characterized as a determination of NEPA adequacy rather than as a NEPA decision.

E. Decision-Making Times at Each Level of Analysis
Despite impassioned critiques that NEPA causes delay, there is very little published information available regarding the length of time it takes federal agencies to make decisions at each level of review. In 2020, the CEQ issued a report concluding that across all Federal agencies, the average (i.e., mean) EIS completion time was 4.5 years, and the median completion time was 3.5 years. 88 The CEQ report also provided the number of EISs completed during the period of study (2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)

F. Striking Difference Between Mean and Median Completion Times Shows Skewing by Anomalous, Lengthy Decisions
The striking difference between the mean and median values is important. Mean values reflect the average time to complete the NEPA analysis, while the median value reflects the middle value of the distribution of completion times. Half the cases will take longer than the median, and half the cases will be resolved more quickly than the median. that difference has not been developed fully. It is also important to note that most statistical models (including the regression model developed for this paper) utilize mean values when measuring central tendency. 96 The direction of skewing is also important and informative. The mean consistently exceeds the median, indicating that outliers are long projects, rather than short ones. This observation squares with other research, suggesting that mean completion time is skewed by extreme events. 97 For example, the mean completion time for a CE is 86% longer than the median. The difference between median and mean is smaller for EAs and EISs than for CEs (39% and 23% respectively). 98 However, in all cases the mean exceeds the median, indicating that the distribution of completion times is heavily skewed by lengthy projects.
Graphing the distribution of project completion times illustrates the degree to which the distribution is skewed by outlying values. As the graphs below demonstrate, the distribution of project completion times is heavily skewed with a long tail extending to the right. This distribution of data points is consistent with the observation by Fleischman et al. that a proportionately small number of projects take a long time, even though most projects are completed quickly. 99 completion time is prone to overstating normally occurring completion times). See also, Fleischman et al., supra note 58, at 412. 96 The implications of models that rely on mean values is discussed in more detail in Section II.D. Our study addressed these concerns by modeling in a logarithmic scale. See Section III.A. 97 Fleischman et al., supra note 58, at 413. 98 The comparatively large percentage difference for CEs is not surprising because CEs completion time is often measured in weeks rather than months or years, and even small departures from the norm will represent large percentage changes. 99 Fleischman et al., supra note 58, at 412 (justifying decision to report median rather than mean completion times).
Although the point at which the curve flattens differs between levels of analysis, the shape of the curve is generally consistent. This trend is not surprising when considering EISs, which can involve complex and controversial projects requiring careful analysis as well as extensive public involvement. However, it is surprising that CEs, which are designed to expedite decision-making times for routine projects, also sometimes experience extreme delay. Table 1 displays this same trend with more granularity.  5th  10th  25th  50th  75th  90th  95th  CE  19  30  54  112  245  481  714  EA  91  133  235  445  779  1,306  1,765  EIS  294  395  595 1,007 1,585 2,415 3,020 The long tail at every level of analysis caused us to investigate whether there are factors that can be used to identify projects that are at heightened risk of long NEPA review periods. First, we wondered whether decisionmaking time had changed over the course of the study, and whether those trends were consistent across levels of analysis. Second, we questioned whether certain activities were associated with longer decision-making times. Finally, we questioned whether there were regional differences in decision-making times.

III. REGRESSION ANALYSIS OF TRENDS IN DECISION-MAKING TIMES
We developed a regression model to analyze whether decision-making time could be predicted based on information about individual projects identified using the MYTR database. We utilized a regression analysis in order to isolate the influence of each of these factors. As the Harvard Business Review explains, a Regression analysis is a way of mathematically sorting out which of those variables does indeed have an impact. It answers the questions: Which factors matter most? Which can we ignore? How do those factors interact with each other? And, perhaps most importantly, how certain are we about all of these factors? 100 The regression model enabled us to isolate the influence of the following variables: (1) three levels of analysis; (2) the year when the project was initiated; (3) any combination of forty-three separate activities involved in project implementation; 101 and (4) the nine Forest Service Regions exercising responsibility over the NEPA analysis for the project. The model also indirectly measured project complexity. Several projects involved multiple activities, and the model tested each activity independently. 102 This enabled consideration of the complexity of the activity where multiple activities were included in a single project. 103 The regression model allowed us to compare how each of the variables identified in our model affected elapsed time while controlling for the influence of all the other variables. Appendix 1 provides a more detailed description of the weighted least squares regression model predicting elapsed time on a log scale that we developed for this paper.

A. Quality Assurance and Model Data
When dealing with highly skewed data, it is common to look at the data on a log scale. 104 The log scale reduces the influence of extreme values, thereby helping to satisfy the assumptions necessary for the regression analysis. The log scale also affects the meaning given to the regression coefficients, showing the rate of change as opposed to actual changes in values. For clarity and ease of understanding, this article reports results in terms of percent or relative changes in elapsed time rather than actual change in days for elapsed time. Thus, throughout our discussion of the regression model results, all references to the "average" are on a log scale.
For the regression model, we eliminated projects initiated after 2016 to minimize the risk of selection bias. The risk of selection bias is best explained by example. The median completion time for an EIS is 1,007 days, and approximately seventy percent of all EISs are completed within 1,460 days, which is the maximum amount of time available to complete an EIS initiated in 2017 and completed in 2020 (365 days x 4 years). Including in our model EISs initiated in 2017 and completed in 2020 would have excluded the roughly thirty percent of EISs initiated in 2017 that we estimate would not have been completed within the time available. The problem would have been more severe for EISs initiated in 2018 and 2019. Including recently initiated and completed projects, while potentially ignoring recently initiated projects that remained pending at the time of our analysis, could skew model results and inaccurately indicate a reduction in elapsed time during the most recent years in the data set. That potential for selection bias was visible when graphing both the average elapsed time and median elapsed time for each level of analysis, which shows a marked requiring elimination of either purposes or activities. We chose to analyze activities for three reasons. First, feedback from the Forest Service indicated that the "activities" category provided more accurate data than the "purposes" category. Second, with 43 possible categories, "activities" supported a more granular analysis. Third, the model produced a higher R squared value when using "activities" rather than "purposes," further indicating a higher level of reliability. 103 In other words, the model "expects" that the addition of activities to a single project would take additional time and recognizes divergence from this expectation. downward trend after 2016. 105 In addition, the number of completed EISs decreased rapidly after 2016. We therefore excluded from our regression analysis all projects with an initiation date of 2017 or later. Consistent with standard practice in regression analysis, we checked for high leverage data points, which are individual decisions having an unusual combination of values for the independent variables resulting in a disproportionate effect on the dependent variable. We also looked for highly influential data points that could skew model results through a combination of unusual values for the dependent variable and an unusual combination of values for the independent variables.
Using standard regression diagnostics DFFITS, Cook's D, hat values, and standardized residuals, a total of 341 high leverage and highly influential observations were identified and removed from our dataset. The regression analysis was performed with and without the 341 observations, and the change in the results was negligible. The results in this report are for the regression model with the 341 high leverage observations removed.
Having applied these quality control measures, our final data set contained 33,283 observations (27,134 CEs; 5,605 EAs; and 544 EISs). With this final dataset, we were prepared to run the model and analyze the results.

IV. MODEL RESULTS
Contrary to our expectation, the regression analysis revealed that the level of analysis, date of project initiation, Forest Service Region, and activities involved in each project could only explain 25% of the variability in the elapsed time required to complete the NEPA review (R 2 = 0.25). Threequarters of the variation in NEPA completion time is attributable to factors not controlled for in our model.
This result is consistent with observations made by the GAO and the Congressional Research Service that many delays associated with NEPA compliance are caused by external forces, including permitting or legal compliance with other statues, unstable funding, and inadequate staffing. 106 If delays are caused by factors independent of the NEPA process, as these and other sources suggest, it follows that these delays would not be predictable by 105 See supra Section III.E. 106 GAO, NEPA: LITTLE INFORMATION EXISTS, supra note 3, at 1, 15 (noting that for non-federal projects requiring a federal permit, delays in obtaining project funding, changes to the proposal that occur during the NEPA process, and non-federal approvals may all delay a project). The Congressional Research Service also notes that NEPA may run concurrently with other permitting efforts, and delays obtaining other permits may indirectly delay the NEPA process. LINDA LUTHER, CONG. RSCH. SERV., RL33267, THE NATIONAL ENVIRONMENTAL POLICY ACT: STREAMLINING NEPA 9 (2007) [hereinafter CRS, STREAMLINING NEPA] ("In fact, the NEPA process may be extended as a result of the need to complete a permitting process or other analysis required under separate statutory authority (e.g. the Clean Water Act or Endangered Species Act), over which the lead agency may have no authority.").
measuring factors within the NEPA process. This finding has significant implications for regulatory reform and lawmaking, which we explore in Section V.
Despite the model's muted ability to predict decision-making time on a case-by-case basis, each of the factors that we measured demonstrably influence the length of time required to complete the NEPA analysis. Those observations are also discussed below.

A. The Level of Analysis is an Imperfect Predictor of the Length of Time to Complete a Document
NEPA reforms frequently focus on reducing the level of analysis as a method for expediting decision-making. 107 The regression model allowed us to test the validity of this assumption by evaluating whether a predictable increase in decision-making time was associated with a higher level of analysis and whether that relationship had changed over time.
As expected, level of analysis is the strongest predictor of the elapsed time required to complete a NEPA decision. The full regression model (which contains predictor variables: level of analysis, year, activities, and region) can explain 25% of all the variation in elapsed time for a NEPA decision. By itself, level of analysis can explain 20% of the variability in our response variable. Shifting an otherwise identical project to a more rigorous level of analysis increases the average time required to complete the review. In 2004, if a project evaluated in a CE shifted to an EA (with the same activities and in the same region), the model predicted the duration of the analysis would have increased by an average of 226%. If a project evaluated in an EA shifted to an EIS, the model predicted the duration of the case to increase by approximately 21%. These comparisons are made after adjusting for activities and region.
This relationship remained generally stable through the course of the study, though each level of analysis followed its own unique trend. By 2010 the predicted increase in the duration of an analysis that shifted from CE to EA was 242% (up from 226% in 2004). The increase from an EA to an EIS was 117% (up from 21% in 2004). By the end of the study in 2016, the predicted increase from CE to EA was 338% (up from 226% in 2010). The predicted increase from an EA to an EIS would be 27% (down from 117% in 2010). The figure below shows the fitted quadratic trend lines for average elapsed time on a log scale versus the actual average elapsed time on a log scale. 108 The rate of change for all three levels of analysis showed mild quadratic trends. The average length of time required to complete a CE decreased slightly between 2004 and 2011, and then increased slightly between 2011 and 2012. EAs followed a similar pattern, but they increased between 2008 and 2009. Interestingly, EIS cases followed almost the opposite pattern of CEs. The average rate of change for EISs increased until 2011 or 2012, then decreased slightly.
These results lead to the unsurprising observation that over the course of the study, on average, an EIS took longer to complete than an EA, which in turn generally took longer to complete than a CE. This is as expected because EISs involve the most searching level of review and public comment, while CEs are reserved for projects that do not require a deep analysis. However, the regression analysis also reveals that level of analysis is an imperfect predictor of decision-making times. While the statement, "an EIS takes longer than an EA, which takes longer than a CE," seems to belabor the obvious, it does not always hold true. Some CEs take longer to complete than some EAs and even some EISs. Similarly, some EAs take longer to complete than some EISs. In other words, there is important variability in decision-making times across levels of analysis. This variability cautions against moving projects into a less rigorous category of analysis without first considering other factors. The graph below illustrates this point by displaying the overlapping completion time curves of each level of analysis.
At the extreme, the 95th percentile of CEs took 714 days or longer, which is almost double the median time to complete an EA. 109 Similarly, the 95th percentile of EAs took 1,765 days or longer, which is roughly two years longer than the median time to complete an EIS. 110 However, these extreme cases are not the only reason for the variability. Anomalous NEPA decisions with long completion times, like the ones just discussed, are so infrequent, that they would not produce the high percentage of variability generated by the regression model.
Independent of the extreme outliers discussed above, the timelines of different levels of analysis overlap with sufficient frequency that the level of analysis does not reliably predict decision-making time. As our observational data revealed in Section II.F, impact of including one or more of the forty-three potential activities in a project while holding all other variables constant. We therefore used the regression model to test the relationship between the time required to complete the NEPA analysis and the activities involved in the project. This helps determine whether certain activities are associated with delays, and conversely whether certain activities tend to expedite decision-making. Appendix 3 contains the fitted regression coefficients for each activity (in descending order), the 95% upper and lower confidence bounds for the estimated coefficient, and the estimated change in elapsed time if the activity is present in an individual NEPA analysis. The regression coefficients quantify the effect of individual activities on elapsed time after controlling for level of analysis, region, and year. Some coefficients for activities are negative, indicating that they are associated with faster than average completion times. Activities in bold text are statistically significant at the 0.05 level.
The definitions for each of the activities are defined in the Forest Service's PALS User guide v.5.12 and are included as Appendix 2. The width of the confidence intervals indicates the degree of confidence that the presence of this activity will be associated with higher (or lower) elapsed times. A wider confidence interval indicates less certainty as to the actual magnitude of the increase (or decrease) in elapsed time for a specific activity. 118 The top three activities associated with longer decision-making times are: 119 (1) Forest Plan Creation/Revision, which is associated with a predicted 97.2% increase in the elapsed time; (2) Oil, with a predicted 87.9% increase in elapsed time; and (3) Land Exchanges, with a predicted 75.5% increase in elapsed time. We found that even though the regression model indicated that these activities are statistically significant predictors of delay, there were also projects with the same activity where the NEPA analysis was completed more quickly than the median time, indicating that quick completion was possible despite modeled predictions of delay. Additionally, the variability in completion times had a different profile for each activity. For example, almost every NEPA analysis with oil as an activity was completed more quickly than average, but a small number of CEs lingered over 1,000 days. In contrast, the distribution of times required to complete the NEPA analysis on land exchanges appeared random, while forest plan revisions consistently took longer than the Forest Service's median completion times during the course of this study.
To understand these results, we turned to investigations conducted by the Government Accountability Office (GAO) and the Congressional Research 118 The number of NEPA decisions that involve a given activity has a direct effect on the width of the confidence interval, with more records resulting in more robust predictive ability and therefore narrower confidence intervals (all else being equal). Here, wider confidence intervals are often associated with activities that were not addressed in very many NEPA analyses. 119 For ease of reading, we refer to the name and definition as defined by the PALS user guide rather than using the exact title in the table. [Vol. 46:S Service (CRS), legal analysis, and industry commentary. These sources suggested possible explanations for the wide variability in completion times, many of which were external to the NEPA process. We discuss the activities in reverse order from least to most significant predictors of delay.

Land Exchanges: Sources of delay identified in the land exchange process apply equally to the NEPA decision-making process
Land exchanges are transactions where the Forest Service conveys away National Forest System lands in return for non-system lands that better advance Forest Service objectives. The Forest Service, for example, may give up an isolated forest parcel that is difficult to manage in return for private inholdings within a National Forest or lands along the border of a National Forest that improve public access to the forest. Land exchanges demonstrate the way in which administrative issues within an agency, such as a lack of experienced staff, uncertain funding, and alternative priorities can delay the decision-making process. These challenges also affect the NEPA decision-making process.
The regression model evaluated 236 projects listing Land Exchanges as an activity. The shortest NEPA review of a land exchange took two days, 120 and the longest took almost ten years (3,642 days). 121 Of these, 93 (39%) land exchanges were evaluated in CEs, 134 were evaluated in EAs (57%), and 9 (4%) were evaluated in EISs. Roughly one-third of the projects (80) were completed in less than a year.
Comparing these results to the Forest Service's median time to complete decisions at each level of review, five of the nine EISs took longer than the median time of 1,006 days, and three took at least twice as long. 122 In contrast, two EISs were completed more quickly than the median time for an EA (445 days). 123 Turning to EA completion times, 27.6% took longer than the median time to complete an EIS (37 projects out of 134 took longer than 1,006 days). An almost equal number were completed more quickly than the median EA (42 projects took less than 445 days). Additionally, 3% of EAs were completed in less than the median time to complete a CE (5 projects took less than 112 days). Finally, looking at the 93 CEs, just 16% were completed more quickly than the median time to complete a CE (15 projects out of 93 took less than 112 days). However, 37% took longer than the median EA (35 out of 93 took longer than 445 days), and 10.6% took longer than the median time for an EIS (10 projects out of 93 took longer than 1,007 days). In summary, the NEPA analysis required for land exchanges varied widely in terms of completion times. Although some projects were completed quickly, many projects exceeded the median time to complete a higher level of review. Notably, this applied to CEs, which do not require detailed analysis. A relatively high proportion of CEs experienced delay beyond the median times for an EA or an EIS. To understand why, we investigated the legal process for accomplishing land exchanges, as well as GAO investigations identifying delays within the land exchange process.
Most land exchanges go through a similar process: 124 "receiving or making a proposal for an exchange, conducting a feasibility analysis, signing a nonbinding agreement to initiate, obtaining appraisals of the land, conducting resource and environmental analyses, deciding on whether to complete the exchange, and preparing title and closing documents." 125 The Forest Service's handbook provides implementation schedules for various types of exchanges that include a range of 56 to 71 action items. 126 According to the GAO, lack of qualified staff, inadequate funding, and lower prioritization of land exchanges compared to other activities were identified as sources of delay for land exchanges involving the Forest Service. 127 When reviewing 250 land exchanges conducted from October 1, 2004 through June 30, 2008 by the BLM and the Forest Service, the GAO reported that in almost every reviewed case, the agencies took longer than estimated to complete the exchange. 128 In explaining these delays, both agencies cited "the lack of staff" and the "lack of qualified appraisers." 129 Both agencies reported that "owing to an increasing number of retirements and the need to work on higher priority activities-such as processing energy rights-ofway-staff may not be available to process exchanges." 130 These same cross-cutting challenges plague the NEPA process. In 2018, the Forest Service collaborated with the National Forest Foundation to conduct a series of regional roundtables focused on Environmental Analysis and Decision Making (EADM). 131 The regional results were synthesized into a The report recognizes an increase in average decisionmaking times between 2005 and 2016, and also notes that the "non-fire workforce is at its lowest capacity in years." 133 Emphasizing the point, the report indicated that in 1998, non-fire personnel exceeded 17,500, while fire personnel sat closer to 5,000 employees. By 2015, non-fire personnel had been reduced to around 10,000 employees, while fire personnel had grown to over 11,000. 134 The transition to a fire dominant staff affects the availability of qualified personnel to conduct environmental planning, monitoring, and analysis, which includes NEPA reviews.
The EADM Roundtable Synthesis Report described how the reduction in qualified staff, inadequate or uncertain funding, and lower prioritization of environmental planning had affected NEPA decision-making times. "Turnover, detail assignments and fire response often reduce productivity due to interruptions in project momentum and changes in project direction." 135 Inadequate funding further affects decision-making times. "Budget shortfalls and statutory mandates on funding for fire response, combined with a shortage of trained employees in areas other than fire and/or a frequent diversion of staff to emergency response or shifting priorities, hamper the ability of the Agency to make progress on other important forest and grassland management efforts." 136 Finally, effective and efficient environmental decision-making requires qualified staff. "[T]he complexity of landscapescale (e.g., climate, fuels, insects, and disease) demands a high level of expertise and deep knowledge of forest conditions at multiple levels of the agency." 137 Despite this need, "training in project and personnel management, resource specializations, and EADM itself remains an unaddressed need throughout the USFS." 138 and the Endangered Species Act. According to the Forest Service website, "EADM is a change effort that intends to reduce the time and cost of our environmental analysis and decisionmaking processes to produce efficient, effective, and high-quality land management decisions to accomplish more work on the ground and be more responsive to the public we serve." Im- In addition to delays within the land exchange process, the GAO also remarked on a significant reduction in the number of land exchanges accomplished annually. In summary, the activity "land exchanges" reveals the degree to which internal management issues including a lack of experienced staff, an insufficient number of staff, insufficient funding, and competing agency priorities create delays. These same challenges affect NEPA implementation. These are serious problems that must be addressed, but they are problems that grow from an under-resourced agency struggling to adapt to a rapidly evolving mission. They are not problems rooted in agency NEPA regulations or practice. Providing the Forest Service and other agencies the resources they need to fulfill their NEPA obligations should be the starting point for improving NEPA efficacy.

Oil: Abnormally long completion times for a small number of projects may be caused by external factors including operator priorities, market dynamics, and lease suspensions
Activities involving oil demonstrate the variability in completion times across levels of analysis discussed in Section III(A). A small number of CEs took extremely long, creating an impression of delay for this activity. The regression model identified 64 projects involving oil as an activity. The fastest project took 20 days. 148 The longest took almost 8 years (2,910 days). 149 Of the 64 projects, 75% were completed in less than a year. Lengthy CEs were common when oil was included as an activity. Specifically, just 36% of the CEs (18 out of 50) were completed more quickly than the median for all CEs, 112 days. Ten percent of CEs (5 out of 50) took more than 1,000 days to complete, which is close to the median time for an EIS (1,006 days), 150 and 16% of CEs took longer than the median time for an EA (8 out of 50 took longer than 445 days). CEs involving oil are, in short, more likely to result in delays than CEs for other activities. In contrast to longer CE completion times, 75% of EAs involving oil were completed faster than the median time of 445 days. 151 There were only two EISs: one was 146 Id. at 14. 147 See EADM ROUNDTABLES NATIONAL SYNTHESIS REPORT, supra note 131 at 15 ("Turnover, detail assignments, and fire response often reduce productivity due to interruptions in project momentum and changes in project direction."). 148  completed in 679 days, which is much faster than the median time of 1,006 days for EISs. The other required 2,910 days, which is far longer than the median.
In summary, most of the delays associated with oil were due to the slow processing of CEs. 152 This is unusual because all of the CEs involved an Application for a Permit to Drill (APD), 153 which comes only after multiple prior environmental reviews. 154 Thus, these outliers provide an opportunity to explore why the lowest level of analysis did not result in an expedited decision.
To understand why some CEs took so long when most of the other NEPA documents for the same activity were processed within normal to fast timeframes, we looked to the regulatory structure and GAO reports investigating the oil and gas permitting process. 155 The results of those investigations provide insight that may help explain the wide variability in decisionmaking time for CEs within the Forest Service process. Specifically, sources of delay within the BLM permitting process include waiting for information from the operator, market dynamics, and operator priorities.
152 The PALs database distinguishes between "Oil" and "Natural Gas" as activities, but the definitions seem similar. See infra Appendix 2 (providing definitions of activities for PALs database). Of the 64 projects involving "Oil" as an activity, 44 also included "Natural Gas." To understand the different results for "Oil" and "Natural Gas," we focused on the 33 "Natural Gas" projects that did not include "Oil" as an activity. Of these, there were 2 EISs; 7 EAs; and 24 CEs. Each level of analysis had wide variability in decision making times; however, there were no long CEs. The longest CE took 273 days. That appears to be why these functionally similar activities received such different results in the regression analysis. 153 Theoretically, most of the information to make a decision on an APD should have already been considered either at the land use planning stage and again at the leasing stage. However, these early analyses often attempt to delay gathering environmental information until later in the leasing stage. A GAO report from 1990 concluded that "inadequate land use plans and/or environmental studies have resulted in leasing being suspended, primarily on Forest Service Lands" and that the foregone revenues from delayed oil and gas leases "far exceed any reasonable estimated cost to develop such information for resource areas and forests with high oil and gas potential. and uncertainty as partially attributable to "inexperienced and/or unempowered team leaders" who "may be preparing his/her first EIS" and specifying that the "lack of training results in unnecessary wasted time" including "failing to tier to earlier documents" and also citing concerns that the process "grinds to a halt" when the team leader is "out of the office for vacation, illness, training, or other priorities"); id. at n.5 and accompanying text ("It has been the author's frequent experience that the BLM and the Forest Service delay decision-making in order to prepare more and lengthier NEPA documents in an effort to bulletproof their decisions from appeal."). For a more thorough discussion of delays caused by litigation aversion, see infra Section IV(D)(2). relevant, they would not explain why most of the EAs in this category were processed expeditiously, while an unexpectedly high proportion of CEs took much longer than average. All of the CEs that took longer than average involved an APD approval, which is the final stage for development of an oil or gas well. 160 After a lease has been issued, the lessee has ten years to drill a well and commence production. 161 By the time the Forest Service and BLM act on an APD, the development proposal has already undergone at least two NEPA reviews (at the Forest Planning phase where the Forest Service determines whether oil development is an appropriate use of National Forest System lands, and at the leasing stage where the Forest Service determines whether a specific parcel is appropriate for development and what surface use stipulations are needed to protect other resources). Each analysis considers more detail as site-specific analysis becomes more focused. With appropriate tiering, and barring unforeseen complications, approval of an APD should be simple and capable of expedited review.

b. Some Delays in APD Approval Are Attributable to the Operator
One source of delay identified by the BLM is time spent waiting for information from an operator. 162 The BLM depends on information from the operator when processing an APD. If the operator responds slowly, decisionmaking time increases, skewing data reported in MYTR, even though the delay is not caused by the Forest Service or the BLM. The BLM quantifies this phenomenon. The BLM maintains ongoing data on the time required to process an APD that distinguishes between time the BLM spent waiting for an operator to provide information and time the BLM spent analyzing an APD. For nine out of ten published years (2012 to 2020), the BLM spent more time waiting for an operator to provide information than it spent Step 1, the operator submits the APD, and the adjudicator verifies that the lease is valid and the payment has been received. In Step 2, the adjudicator identifies potential deficiencies in the application and provides the operator 45 days to correct. At this stage, the 30-day public notification process begins.
Step 3 involves the environmental analysis and NEPA compliance. If additional information is required during step 3, the BLM defers its decision and the operator has up to 2 years to provide information. GAO, ACTIONS NEEDED TO IMPROVE BLM'S DATA SYSTEM supra note 155 at 8 (Figure 2: BLM's APD Review Process). [Vol. 46:S reviewing the APD. In some years, the BLM spent almost twice as much time waiting for an operator as it spent analyzing the APD. 163 These delays, which appear to reflect slow NEPA analysis, are not attributable to federal agency action and cannot be resolved by changes to agency regulations or practice. In crafting regulatory reforms, this source of delay should be distinguished from delays caused by agency inefficiencies.
Federal oil lessees may have operational reasons for delaying their responses, or they may need additional time to respond to changing circumstances. Substantial time may pass between leasing and the submission of an APD. During that time, development of adjacent parcels, identification of a nearby cultural resource or sensitive species, improved technology, or a communitization agreement or unitization orders may necessitate changes to an operator's Surface Use Plan of Operations. 164 The site-specific analysis required at the APD phase, or amendments to an existing APD, may require additional planning and analysis to address these developments. These delays are reflected in the NEPA decision-making process, but they are caused by the operational uncertainties of oil exploration in a complex, regulated industry.
Additionally, APDs bridge the divide between aspiration and implementation. According to the GAO, "the three primary factors influencing operators' decisions to apply for or use APDs were economic factors, infrastructure availability, and lease terms." 165 The primary economic and infrastructure-related factors influencing operators were: (1) the price of oil and natural gas; (2) drilling success and geological attributes; (3) technological changes; (4) access to infrastructure, including pipelines; and (5) drilling rig schedules. 166 In addition to these physical factors, market influences came into play. "Some operators may obtain APDs to increase the value of the company without using the APD to drill." 167 Other operators confirmed that they like to keep approved but unused APDs on hand to ensure drill rigs could be kept busy. 168 The number of APDs that get approved but go unused demonstrates the influence of these external factors.  166 Id. at 16-17. 167 Id. at 19. 168 Id. at 20. 169 Id. at 11 (reporting that 9,991 APDs had been approved and put to use, while 9,950 had been approved, but were not being used).
Lease suspensions may also affect the lengthy decision-making times for CEs reflected in the MYTR database. Federal oil and gas leases expire at the end of their 10-year primary term unless oil or gas is produced in paying quantities or the lease otherwise qualifies for an extension. 170 A lessee can avoid expiration of a lease term without producing oil in paying quantities by applying for a lease suspension, tolling the running of the lease term and, in some cases, suspending the lessee's obligation to pay rent while the lease is suspended. 171 As of 2016, there were 2,750 BLM oil and gas leases identified as suspended. 172 Lease suspension may be granted because of market conditions, logistical challenges, weather-related issues, or administrative delay (including waiting for approval of an APD days between project initiation and a decision would increase even though no NEPA action was being taken. This would appear as a long NEPA project in MYTR and the delay would likely be misattributed to the NEPA process even though the delay was caused by the lease suspension. It is also possible that environmental conditions discovered during the NEPA process may make a project less attractive, inducing an operator to apply for a lease suspension pending completion of the required environmental studies. 176 Once the suspension was issued, the operator may not have an incentive to continue pursuing the NEPA analysis if the economics of the well were marginal. For example, when the GAO investigated lease suspensions at the BLM, it identified multiple lease suspensions in Montana that had been in place for more than 30 years. 177 Several of these had been subject to a court order requiring additional consideration of environmental impacts. Documents from the Forest Service indicated that there was little interest at the time in conducting those analyses because of their expense and because the operators had minimal interest in developing the lands for oil and gas production. 178 With the suspension in place, the operators could avoid the expense of additional environmental review without losing the lease.
There is no way of knowing whether the slow-moving CEs in our study were delayed due to a lease suspension. In addition to limited oversight, monitoring, and documentation of lease suspensions by the BLM, 179 the MYTR database numbering system does not interface with the BLM lease suspension database. Nevertheless, it is important to recognize that outside influences may affect NEPA decision-making times in unexpected ways.
In summary, operator priorities, market forces, technological developments, and lease suspensions could extend decision-making times. Focusing solely on decision-making times to assess NEPA efficiency fails to capture these relevant nuances. In these circumstances, "streamlining" procedures that focus on creating new and more expansive CEs or compulsory deadlines would not address the underlying cause of delay, but they would reduce transparency, consideration of alternatives, and opportunities for environmental mitigation. between delays that are caused by industry dynamics and those that are caused by the NEPA process itself.

Forest Plan Creation and Revision: Delays Caused by Compliance with Other Laws May Skew NEPA Compliance Time Data
Like a zoning ordinance, a Forest Plan establishes a vision intended to guide management of a large landscape for fifteen to twenty years, identifying portions of a forest where certain activities are generally appropriate. That the NEPA analysis for Forest Planning takes longer than the analysis for other activities is unsurprising given the geographic scope of these decisions, the often-controversial nature of allocating resources for years into the future, and the potential impacts that are likely to result from those decisions. 180 Additionally, planning itself requires information gathering, analysis, and deliberation, which takes time.
Forest Plan Creation/Revision was associated with the highest rate of longer than average NEPA completion times and, indeed, these activities took longer than most decisions. This is also an activity with multiple and overlapping legal requirements, making it difficult to distinguish between delays caused by NEPA compliance and those attributable to compliance with other laws. 181 The regression database identified 86 projects involving Forest Plan Revisions. The fastest took 45 days and the longest took 5,695 days. 182 Only 16 (19%) took less than a year. Fifty-two of the 84 projects (60%) were analyzed in an EIS, 21 (24%) were analyzed in EAs, and 13 (15%) were analyzed in CEs. Eighty-four percent of the EISs took longer than the median time for EISs (44 out of 52 took longer than 1,006 days). Half of these took at least 2,012 days, which is double the median time for an EIS. Of the 13 CEs, almost half took longer than the median time for a CE (6 out of 13 took longer than 112 days), and most of these took almost double that amount of 180  Forest planning provides a specific example of the CRS observation that NEPA often functions as an "umbrella statute-that is, a framework to coordinate or demonstrate compliance with any studies, reviews, or consultations required by any other laws." 184 Forest Planning occurs within the context of legal requirements imposed by a host of laws that operate independently of NEPA, 185 including the MLA, the Taylor Grazing Act, the Endangered Species Act, the Wilderness Act, the National Historic Preservation Act, and many more.
Principal among these laws, the National Forest Management Act (NFMA) requires the Forest Service to use "a systematic interdisciplinary approach to achieve integrated consideration of physical, biological, economic, and other sciences" while preparing "standards and guidelines" for the management of each national forest. 186 Planning must consider that actions taken on adjacent non-forest system land can impact forest resources, and viceversa. NFMA also demands robust public participation, including making the plans available to the public for at least three months, soliciting comments, and holding public meetings prior to adoption. 187 While much of this can be done concurrently with NEPA, the long decision-making times associated with forest planning may reflect forest management laws and regulations other than NEPA. If NEPA alone were the source of delay, we would not expect to see the disparity in completion times that distinguishes forest planning from other activities. In addition to the legal complexity of forest planning, controversy can also cause delay by generating a large volume of comments on projects that must be resolved before planning can conclude. Avoiding conflicts necessitates communication and coordination with other federal agencies; state, local, and tribal governments; and other interested stakeholders and organizations-all of which takes time. As one Forest Service study notes, "Additional private landowners adjacent to national forests and grasslands means more neighbors with whom the Forest Service needs to coordinate in arranging access for fire management and recreation, managing ecosystems jointly across the landscape, and other management issues." 192 Despite the complexity of this undertaking, forest planning is not well funded compared to other programs. In fiscal year 2019, the "hazardous fuels" and "forest products" programs received almost twice as much funding as "land management, planning, assessment and monitoring. " 193 Additionally, forest planning is rarely triggered by an individual request from a permitted entity. Without supplemental funding from a permit applicant (who may pay the cost of hiring a third-party contractor to prepare the NEPA analysis on behalf of the Forest Service) and without the motivating influence of a project proponent who is eager to commence development, it is possible that forest planning either takes a back seat in the priority queue, or that staff needed to complete planning work are routinely reassigned to other projects. Moreover, forest planning is underfunded and understaffed. In recent testimony before the Senate Energy and Natural Resources Committee, the Deputy Chief of the Forest Service testified that more than half of the 154 forest plans are at least 15 years old, and that the Forest Service "doesn't have enough staff or money to catch up." 194 He added that the Forest Service has seen a decline of about 40 percent in natural resource professionals who work on the management plans because, "we just can't pay for those positions anymore. The length of delay associated with forest planning decisions deserves attention. As with the previously discussed activities, it is possible that a lack of funding and inexperienced or rotating staff exacerbate delays. 200 Moreover, changing circumstances related to climate change and urbanization ever more visible along with conflicts between the participants-personified by intense controversies over motorized use, wilderness designation, mountain biking, and hunting. These growing problems, though commonly linked to individual choice in recreational preferences are also coupled to powerful economic and political forces that are driving what some now regard as an 'industrial scale' recreation problem."); Andrew  200 These issues, particularly the frequent rotation of staff, were often identified as sources of delay during the regional roundtables. See, e.g., REGION 1 ROUNDTABLE REPORT, supra note 192 at 8-9 (identifying "high turnover of permanent staff positions within all levels of agency" and "'move on, move up'" practice of relocating staff for career advancement as sources of knowledge voids and delays within the NEPA and planning process); REGION 2 ROUNDTABLE REPORT, supra note 192 at 8 (identifying "leadership change and staff transitions" and "acting positions" as sources of delay in the NEPA and planning process); NAT'L FOREST FOUND., SOUTHWESTERN REGIONAL EADM PARTNER ROUNDTABLE 9 (2018) [hereinafter REGION 3 ROUNDTABLE REPORT] ("Staff transitions are too frequent. . . . NEPA delays caused by staff turnover"); REGION 4 ROUNDTABLE REPORT, supra note 192 at 8 (identifying staff turnover, hiring freezes, lengthy hiring process, temporary workforce, and staff without local institutional knowledge as sources of delay in the NEPA and planning process); REGION 5 ROUNDTABLE REPORT, supra note 192 at 8 ("rapid turnover undermines productivity of partner relationships, especially at the local level" and "short tenure of leadership staff limits their ability to apply local knowledge"); REGION 6 ROUNDTABLE REPORT, supra note 192 at 7 (identifying "lack of continuity fostered by 'mobility policy' both in terms of USFS staff often having short tenure in their positions and also leaving for details" as a source of delay); REGION 8 ROUNDTABLE REPORT, supra note 192 at 8 ("Lack of staff continuity negatively affects EADM. Loss of knowledge between staff due to lack of overlap."); REGION 9 ROUNDTABLE REPORT, supra note 192 at 7 (identifying "turnover of both leadership and staff in the course of one project" as a source of delay in NEPA and planning decisions); NAT'L FOREST FOUND., ALASKA REGIONAL EADM PARTNER ROUNDTABLE 7 (2018) [hereinafter REGION 10 ROUNDTABLE REPORT] (identifying "rapid loss of NEPA team leadership as well as other NEPA expertise" as a source of delay).
require deliberation and provoke controversy, which takes time to resolve. Creative recommendations suggest ways in which NEPA could facilitaterather than hinder-more efficient forest planning. 201 Pilot projects within the Department of Transportation demonstrate NEPA's ability to advance coordinated efforts, as discussed in more detail in Section V. Just as NEPA may not be the sole cause for delay within the Forest Planning process, it also cannot serve as the sole remedy. Finding solutions to facilitate faster forest planning decisions involves complexities and nuances that are worthy of discussion but beyond the scope of this article.

C. Geographic Region Has a Significant Influence on Decision-making Time
The regression model revealed that the Forest Service administrative region responsible for overseeing a NEPA analysis has a significant influence on decision-making times. The relationship between region and NEPA completion time varied with each level of analysis. Despite this variation, Region 1 (the Northern Region) consistently took longer to complete NEPA decisions at all levels of analysis, and Region 8 (the Southern Region) and Region 9 (the Eastern Region) consistently boasted the fastest decisionmaking times at all levels of analysis.
Initially, this finding surprised us. Each Forest Service region is implementing the same laws, subject to the same regulations, pursuant to the same administrative guidance, involving the same activities, and (presumably) subject to similar financial and staffing challenges. We therefore did not expect to see a large variation in elapsed times across regions. The regional variation in completion times suggests that factors external to the NEPA process affect completion times. If the delays were caused solely by the NEPA process, we would expect similar mean completion times across regions, after controlling for the year of project initiation, level of analysis, and activities.
It is possible that ecological differences between the regions affect the variation in completion times. Cultural differences may also cause varying completion times. Although we explore some of these potential influences below, regional differences in completion time justify further research. Understanding why some regions complete the NEPA process more quickly than other regions may reveal administrative and management efficiencies that could be replicated. 201 See, e.g., Mark Squillace, Rethinking Public Land Use Planning, 43 HARV. ENV'T L. REV. 415, 437-52 (2019) (recommending elimination of "standards and guidelines" in forest plans and a shift toward landscape level planning with a robust system of monitoring and adaptation leading to informed and thorough activity-level planning tiered to larger scale documents).

Regression Model Results Regarding Forest Service Regions
The model computed regression coefficients for all Forest Service Regions relative to Region 5. 202 Although any region could serve as a baseline, Region 5 was used because it was involved in NEPA decisions with every level of analysis and all activities, establishing a uniform baseline for comparison.
Given that the regression model is predicting elapsed time on a log scale, it is easiest to discuss regional impacts in terms of percentage change in elapsed time. The table below sets forth the results and lists the estimated percent change in elapsed time if a NEPA decision is in a specified region other than Region 5. These differences exist after controlling for the level of analysis, year of project initiation, and the activities involved in the project. Following the model results, we provide a brief discussion of ecological characteristics of each region, and then turn to a discussion of budgetary challenges caused by wildfire suppression that could have regional effects. For projects undergoing review in a CE, Regions 1, 2, 4, and 6 are associated with NEPA completion times that are 20% to 30% longer than Region 5. Regions 8, 9, and 10 are associated with elapsed times roughly 10% to 15% below Region 5. For projects undergoing review in an EA, Regions 1 and 6 are associated with the longest elapsed times. Regions 2, 3, and 4 are associated with elapsed times within 5% of Region 5. Regions 8, 9, and 10 are associated with the longest elapsed times. For projects subject to review in an EIS, Regions 1, 3, and 4 are associated with the longest elapsed times. Completion times in these regions are more than 50% longer than those in Region 5. Why EISs completed in Regions 1, 3 and 4 should take longer may warrant more careful review. EISs completed by Region 9 also deserve careful consideration, as Region 9 may have found an opportunity to maximize analytical or procedural efficiencies. Region 8 is listed as NA because there were too few EISs completed during the study to accurately estimate the effect of Region 8 on EIS cases. by a forest supervisor, and the units are arranged into nine administrative regions, each headed by a regional forester. 204 Most Forest Service lands are concentrated in the West (87%); however, the Forest Service administers more federal land in the East than all other federal agencies combined. 205 The national forests in the eastern states have smaller contiguous landscapes and are peppered with inholdings. 206 Glancing at a map of Forest Service lands and regions demonstrates the wide variability in scale and contiguous landscapes between the different regions. 207 Differences in patterns of regional development may affect the scale and intensity of NEPA decisions in different regions. Regions 8 and 9 complete the most NEPA analyses at the fastest rate. These regions are also in areas with established urban areas, smaller national forests, lower wildfire risk, and more established patterns of landscape use. 208 Region 9 characterizes the national forests in its region as "islands of green in a sea of people," which is appropriate because Region 9 encompasses twenty states with over forty-three percent of the national's population and nine of the largest twenty metropolitan areas in the U.S. 209 Even though Regions 8 and 9 cover a vast territory, they have the smallest amount of federal land within their regions, partially due to inholdings. 210 Though national forests in Region 8 include over 25 million acres of land, only 13.4 million acres are National Forest System land, while 12 million acres are non-federal inholdings. 211 Similarly, Region 9 encompasses over 22 million acres of National Forest System lands of which almost half are non-federal inholdings. 212 In contrast, Regions 4, 1, and 3 are each associated with some of the longest decision-making times. These regions, located within the Intermountain West, are all grappling with drought, wildfire, and potentially a faster rate of climate change affecting the landscape. 213 These regions also have larger national forests, broader swaths of public land, and a higher concentration of areas with very high wildfire hazard potential. 214

Wildfires May Have Disparate Fiscal Effects Across Regions
Unequal regional burdens associated with wildfire management may contribute to differences in NEPA decision-making times. A 2006 Office of Inspector General Report found that wildland urban interface (WUI) protection "was the major driver of [Forest Service] suppression costs, with some staff estimating that between 50 to 95 percent of large wildfire suppression expenditures were directly related to protecting private property and homes." 215 Where Forest Service protection responsibilities are directly adjacent to housing developments, Forest Service Line Officers often feel compelled to aggressively suppress wildfires, even if the fires pose no threat to National Forest resources. 216 The Office of Inspector General reported that Regions 1, 5, and 6 bore "an inequitable wildfire protection burden" because wildland fire protection agreements between the Forest Service and other agencies in Oregon, Washington, California, Montana and Idaho had not been renegotiated to reflect appropriate WUI protection responsibilities." 217 While updates may have partially addressed these concerns, fire related responsibilities continue to increase and dated or inadequate agreements would have impacted the decisions reviewed in this analysis. The Wildfire Hazard Potential map, 218 produced by the Forest Service, demonstrates that Regions 1, 3, 4, 5 and 6 have the highest concentration of wildfire hazard potential. With regards to NEPA decision-making times, nation in population growth" and that recent warming in that region is "among the most rapid in the nation, significantly more than the global average in some areas"). Regions 1, 2, 3, and 4 are associated with the longest regional decision-making times.
Other than the correlation identified above, it would be difficult to link the budgetary shortfalls caused by wildfire suppression to regional differences in decision-making times. According to the GAO, the Forest Service does not systematically track such impacts at a national level. 219 Understanding the cause of regional differences in decision-making times is an important aspect of NEPA reform. If regional differences in decisionmaking times are caused by ecological differences, evolving demographics, or disparate budgetary challenges, those underlying management challenges should be recognized and addressed. If budgetary shortfalls cause delay, then fiscal, rather than NEPA reforms, should be considered.

D. Background Factors Affecting NEPA Decision-making Timeframes
Through our research, two issues arose consistently: budgetary uncertainty caused by wildfire borrowing and a culture of litigation aversion within the Forest Service. These two dynamics likely influence decisionmaking times, even though the effect cannot be specifically identified through the MYTR database or our regression modeling. We discuss each issue in turn.

Budgetary Uncertainty Caused by Wildfire Borrowing Affects Program Efficacy, Including Planning and Environmental Analysis
Wildfire suppression costs exceeded appropriations in most years since 1990. 220 When firefighting expenses exceed funds appropriated for wildfire suppression, Congress allows the Forest Service to transfer funds from other programs to cover those costs in a practice referred to as "fire borrowing." 221 Congress typically reimburses the Forest Service for unanticipated firefighting expenses, but the reimbursement is often incomplete or delayed. 222 For example, the Forest Service, beginning in the mid-1980s,  Concerned about the viability of that fund, in 2001, the Forest Service began transferring funds from other management programs and activities. While this practice ensured that bills were paid, it left other obligations, likely including NEPA, with less discretionary funding. 225 Congress recognized this problem and its implications for National Forest System management, 226 and in fiscal year 2018, Congress enacted legislation to stabilize funding. 227 While this legislation will likely go a long way towards stabilizing funding, fire borrowing continued through the course of this study. 228 Determining whether funding reforms result in improved NEPA efficacy is a question that cannot be answered yet based on the MYTR database and that will require further research.
Throughout our study period, fire borrowing affected the staff and resources available to complete NEPA projects and thereby increased NEPA compliance times. 229 A 2004 GAO report investigating fire borrowing concluded that the Forest Service "canceled or delayed numerous projects, failed to fulfill certain commitments to partners, and faced difficulties in managing their programs when funds were transferred for fire suppression." 230 In some cases, this practice "increased the costs and time needed to complete projects." 231 Additionally, transfers "disrupted agency efforts to effectively manage programs, causing planned activities to go unfunded and, in some cases, causing programs to be depleted or overspent." 232 Further, "officials often had to duplicate their efforts because of [budgetary] transfers, which prolonged delays and added costs." 233 The stop-start funding also caused "a domino effect: deferring one year's projects displaces the next year's projects, which must in turn be deferred to the following year." 234 A 2019 Report from the CRS confirmed that the practice of wildfire borrowing continued to affect other Forest Service programs, including activities that are central to NEPA compliance. 235 "Fire expenditures continue to climb, affecting the implementation of other programs . . . through personnel and funds transferred to fire control." 236 Additionally, "stakeholders identify other administrative barriers-such as inadequate program funding levels and training-as preventing FS from implementing planning requirements in a more efficient manner." 237 The uncertainty caused by wildfire suppression activities was identified as a cause of delay complicating NEPA compliance during the 2018 EADM roundtables. 238 "Budget shortfalls and statutory mandates on funding for fire response, combined with a shortage of trained employees in areas other than fire and/or a frequent diversion of staff to emergency response or shifting priorities, hamper the ability of the Agency to make progress on other important forest and grassland resource management efforts." 239 As an example of how fire borrowing affected resource management projects, consider a project from the Bitterroot National Forest (Region 1). In that case, a project to stabilize nine miles of dirt road was delayed when 230 GAO, WILDFIRE SUPPRESSION FUNDING TRANSFERS CAUSE DELAYS, supra note 219, at 14. 231 Id. 232 Id. 233 Id. at 15 ("For example, officials had to revise budgets and construction plans, update cost estimates and rewrite land acquisition documents when delays caused them to be outdated, all of which further compounded project delays. . . . In addition, when delays were prolonged, supply costs increased, land prices rose, and impacts to natural resources spread, which also increased the projects' costs."). 234 Id. at 31. The report went on to project that "the agencies and the Congress will repeatedly confront difficult decisions in determining how much funding to transfer from which programs and how much to reimburse." Id. 235 HOOVER & RIDDLE, supra note 144, at 22 ("Congress has expressed concern about the impact of fire borrowing on other NFS management activities and about the increasing portion of FS budget going toward suppression funding."). 236 Id. at 24. 237 Id. 238 See supra note 131 and accompanying text. 239 EADM ROUNDTABLES NATIONAL SYNTHESIS REPORT, supra note 131, at 18; see also id. at 15 ("Turnover, detail assignments and fire response often reduce productivity due to interruptions in project momentum and changes in project direction."). $1.2 million was transferred to wildfire suppression in 2002. 240 The road was collapsing, causing sediment to run into a stream and jeopardizing fish habitat (including a threatened species). 241 Two years after the transfer, only $430,000 was reimbursed to the project. 242 With reduced funding, the project shrank to two of the original nine miles, but in the interim, additional sediment had accumulated in the stream, exacerbating the problem and making restoration even more complex. 243 Although the GAO report describing this project did not discuss the NEPA decision-making process required for this project, one can imagine how it could be affected. Delayed implementation and deteriorating environmental conditions could result in new or more significant issues, requiring supplemental environmental analysis. 244 These changes could extend NEPA decision-making times, even though the cause of delay was budgetary uncertainty. NEPA decision-making times could also be extended by staff reductions or shifting personnel from project management to wildfire duties. 245 Personnel temporarily assigned to a fire, for example, would be unavailable to work on NEPA projects. Temporary reassignments could also impact the availability to complete fieldwork required for the NEPA analysis. A hypothetical project involving impacts to a sensitive plant species may require botanical surveys coinciding with the period when the plant flowers. Temporarily reassigning a botanist to a fire may only last a few weeks, but if that reassignment overlaps with the botanical survey window, that brief reassignment could delay the analysis by a year. In such cases, delays would be captured by the time that lapsed between project initiation and a final decision, but attributing those delays to NEPA would obscure the true problem and increase the risk that reforms would not produce the desired results.
risk aversion with comments suggesting that Forest Service staff avoid making controversial decisions for fear of affecting opportunities for promotion. 247 Litigation aversion leads to unwieldy, bulky, time-consuming documents. The EADM Roundtables National Synthesis Report summarized the problem as follows: "Minimal litigation or objection is viewed as a positive outcome in terms of a project moving to implementation, but the negative costs of defensive over-analysis, unwieldy documentation, and narrowing the scope of projects in order to 'fly under the radar' of litigants are not usually considered." 248 The concern resurfaced later in the report when discussing lengthy documents as a barrier to efficient decision-making. "Risk aversion and a history of legal challenges to USFS decisions have led to the 'bulletproofing' of environmental analysis documents and specialist reports." 249 The report continued, noting that "the complexity and size of analysis is often inconsistent with the complexity and size of the project." 250 The report explicitly distinguished between this dynamic, which it identified as a cultural barrier within the Forest Service and the NEPA process itself. "NEPA is often blamed for these problems, when really it is not the law itself but the Agency's process that is the cause [of lengthy documents]." 251 This observation is consistent with external research on Forest Service NEPA practice. In 2010, Mortimer et al., found that the threat of litigation had more influence than the degree of environmental impacts on Forest Service decisions whether to prepare an EA or an EIS for recreation and as "risk averse," fearful of "backlash," "not feeling supported in making risky decisions," "perceived risk of being litigated and fear of losing in court" and feeling criticized for taking a risk where "success [is] defined as lack of objections or litigation"); REGION 4 ROUNDTABLE REPORT, supra note 195, at 8 (identifying Forest Service staff as "risk averse" and hemmed by a "sue and settle" reality); REGION 5 ROUNDTABLE RESULTS, supra note 192, at 6, 20, 28 (identifying "risk averse USFS staff" with "fear of making decisions based on imperfect data" and stating that "fear of litigation results in excessive time spent and detail in EADM documents" where EADM documents are "'padded' to mitigate risk of litigation" and "litigation threat undermines opportunities to conduct large landscape EADM"); REGION 6 ROUNDTABLE REPORT, supra note 192, at 6 (identifying "risk aversion" as a barrier with line officers "not wanting to 'rock the boat'"); REGION 8 ROUNDTABLE REPORT, supra note 192, at 6, 8 (identifying "fear of litigation and defensive NEPA stance" as well as reluctance toward "taking on large projects for fear of objection to one small part," suggesting that District Rangers resist a project for political reasons "until they change jobs"); REGION 9 ROUNDTABLE REPORT, supra note 192, at 6 (characterizing a "risk averse USFS culture at all levels" that produces "excessive documentation"); REGION 10 ROUNDTABLE REPORT, supra note 200, at 6 (describing "risk aversion" as a barrier with Forest Service "litigation-proofing documents" based on a "perception that all NEPA documents are challenged when only a small percent are challenged" Those outside of the Forest Service also recognize the problem. As one practitioner remarked, "[i]t has been the author's frequent experience that BLM and the Forest Service delay decision-making in order to prepare more and lengthier documents in an effort to bulletproof their decisions from appeal. As a result, the diversion of agency resources and attention to the preparation of up-front disclosures documents under NEPA means less attention and resources are devoted to on the ground efforts such as monitoring the effects of agency decisions." 254

V. RECOMMENDATIONS
Changes to NEPA practice and to NEPA's implementing regulations should be driven by data on all NEPA decisions rather than anecdotal information about outliers. Analyzing over 41,000 Forest Service NEPA determinations at every level of review taught us unexpected lessons about potential causes of delay within the NEPA process.
We learned that the level of analysis is an imperfect predictor of the time required to comply with NEPA. Forcing a project that merits analysis in an EIS into an EA may not result in a faster decision, and CEs are not synonymous with swift decisions. Reforms should focus on identifying efficient strategies for analyzing complex and controversial projects rather than forcing analyses into a lower level of review.
We observed that reduced agency capacity, inadequate funding, and low prioritization of NEPA-related activities like planning and monitoring cause delays. Without stabilizing agency capacity and providing secure agency funding for NEPA-related activities, even the most elegantly drafted NEPA reforms will falter.
We found that some delays attributable to the NEPA process may be external, including market forces and compliance with other laws. Truncating NEPA compliance will not affect these external forces, but it will reduce transparency and may compromise agencies' capacity to comply with other legal duties.
Finally, we learned that cultural influences, including litigation aversion, cause delay. These cultural influences can be addressed without regulatory reform and enable more prompt, creative, and transparent agency decisions. As regulatory changes to NEPA are contemplated, these cultural, fiscal, and practice-oriented reforms should also be considered.

A. Potentially Useful Changes to NEPA Practice
Our recommendations flow from, and were sometimes included in, preceding sections. In this section, we sought to pair recommendations with realworld examples to demonstrate the practicality, effectiveness, and feasibility of each suggestion.

Ground Change in Good Information, Measure Changes, and Adapt as Needed
There is ample information on the time required to complete an EIS, 255 but the amount of time required to complete the analysis does not tell us why some projects lag. Available data also focuses almost exclusively on EISs, which account for just 1% of all NEPA decisions. It is impossible to design meaningful reform without understanding how NEPA operates for 99% of decisions.
More importantly, and as noted at the outset, NEPA's twin goals involve meaningful public engagement and careful consideration of environmental impacts. Faster does not necessarily mean better progress towards advancing these objectives. It is also impossible to test whether reforms succeed without better data. Databases must also allow for tracking of projects through revisions and litigation.
The Forest Service should be commended for developing the MYTR database and the detailed information captured within it. We are unaware of any other federal agency that maintains comparable data. 256 Analyzing information in the MYTR database provided an opportunity to identify nuances in NEPA practice that were unexpected and sometimes counterintuitive. We strongly encourage federal agencies to compile statistical information on NEPA decisions that would enable similar future insights. Such information could benefit individual agencies and could facilitate 255 See CEQ, EIS TIMELINES 2010-2018 supra note 88. 256 Both the Bureau of Land Management and the Department of Energy compile information on NEPA analysis, and both agencies deserve commendation for these efforts, but neither dataset contains the level of information found in MYTR.
comparison of NEPA practice across agencies, highlighting successful practices that could be beneficial if adopted elsewhere.
There are at least three pieces of information that are not captured in MYTR but would be helpful in identifying future NEPA reforms. First, MYTR does not compile information regarding the source of authority relied upon for a CE decision memo. An investigative report by Wild Earth Guardians reviewed the Forest Service's use of specific CE authorities from January through March of 2020 based on projects found on the agency's Schedule of Proposed Action (SOPAs). 257 Wild Earth Guardians reviewed the SOPAs for 75 national forests across 11 states in Regions 1 through 6, 258 concluding that the SOPAs often failed to identify the specific CE authority for projects. 259 Of 175 fuel management projects across 58 forests, 43% failed to disclose the CE authorities in the scoping document. 260 Only 41 projects issued decision memos that identified the CE authority used. 261 Failing to provide the source of authority for a CE forecloses opportunities to assess whether CEs were applied appropriately. Gathering information regarding the source of CE authority would also allow the Forest Service to assess the frequency with which certain CEs are used and analyze whether some CEs are disproportionately associated with litigation or delay.
Second, MYTR does not indicate whether a decision was initiated as a CE and elevated to an EA due to the existence of extraordinary circumstances. Gathering this data would be helpful in identifying areas or CEs that regularly require more thorough analysis due to extraordinary circumstances. Third, MYTR does not indicate how many alternatives were considered in an EA or EIS. The number of alternatives considered may be useful in considering the extent to which agencies achieve NEPA's twin aims of taking a hard look at the environmental impacts of an action and engaging the public. 262 Other research found a relationship between the number of alternatives considered and achievement of NEPA's goal to reduce environmental impacts-a larger number of alternatives resulted in fewer environmental impacts. 263 Agencies should not be shy about sharing NEPA data. Transparency regarding the NEPA process has proven to increase efficiency. For example, the Federal Infrastructure Projects Dashboard was created in an effort to increase the efficiency of infrastructure development. 264 The Dashboard enables federal agencies to publicly track schedules and status information on pending federal infrastructure projects. 265 Publishing the schedule facilitates interagency cooperation by creating an incentive for agencies to resolve issues in a timely manner in order to meet the agreed upon schedule. 266 According to one participant, "The increased level of accountability helps to ensure that federal agencies are not unnecessarily sidetracked in their NEPA review process." 267 The benefits of this simple transparency device are evident. Since its creation, over thirty high-priority federal infrastructure projects have completed the environmental review and permitting process more quickly than pre-Dashboard projects. 268 As we noted earlier, it is hard to fix something without first understanding how it works. It is also hard to tell whether reforms have delivered the intended outcome without a performance metric. Reforms should include gathering data, analyzing the data, and incorporating the lessons learned in future actions to ensure that reforms function as intended and are corrected if they fall short of that goal. are completed more quickly than the slowest 25% of EAs. CEs also do not guarantee fast decision-making. The slowest 25% of CEs were completed almost as quickly as the fastest 25% of EAs, and 11% of CEs took longer than the median time to complete an EA.

Focus on Improving Capacity, Not Downscaling Analysis
Rather than forcing decisions into a less rigorous analysis, agencies should promote a strategically-sized analysis for long-term efficiency. Although this approach may require additional work on the front-end, it can result in long-lasting efficiencies. Below we discuss three real-world examples where this approach yielded demonstrably improved decision-making times and efficient project implementation over the long-term.
First, programmatic NEPA documents can leverage long-term efficiency by facilitating tiering and accelerating subsequent decisions that require a lower level of analysis. This can be achieved through programmatic analyses to which implementation decisions may be tiered, and through monitoring programs that provide real-time, accurate data to which implementation decisions can be tiered. For example, the Government Accountability Office analyzed the average time to review an Application for Permit to Drill in selected BLM field offices between 2016 and 2019. 273 Where decisionmaking times for other field offices ranged from 106 to 220 days, the Pinedale Office averaged 49 days to make a decision. 274 Rather than avoiding environmental review, Pinedale's efficiency was attributable to careful up-front analysis and effective tiering. Pinedale had conducted thorough programmatic EISs for each of the three oil and gas fields it managed. With the potential environmental impacts of oil and gas drilling to work from, agency officials could efficiently expedite review by tiering a CE to the relevant programmatic analysis. 275 This efficiency was achieved without sacrificing the transparent, deliberative process required by NEPA.
Along these lines, several commentators have recommended implementing post-decisional monitoring processes to simplify future decisions by eliminating the need to repetitively gather data or hypothesize about the effects of a project or a mitigation measure. 276  appropriateness of assumptions in a plan, as well as the adequacy of the NEPA analysis supporting it. 278 Producing a supplemental EIS to respond to these changed circumstances has proven time-consuming and burdensome. 279 In contrast, a monitoring program would enable the incorporation of new information obtained through monitoring in future decisions more seamlessly. For planning agencies, like the Forest Service, this approach would shift the emphasis from periodic large-scale forest plans to a more regular and continuous incremental decision-making process. 280 Where this approach has been adopted, the monitoring process reduced conflict by generating evidence that could be used to develop mutual understanding. 281 For example, in eastern Oregon and Washington, monitoring led to broad consensus among stakeholders for treatments in dry forests. 282 Thus, postdecisional monitoring can simplify the NEPA process, increase agency credibility, and facilitate the improved environmental decision-making intended by NEPA's authors.
Second, using the NEPA process as a framework for structured inter-agency collaboration on large projects can facilitate decision-making and implementation through the life of the project. In a pilot project selected by the CEQ for developing best practices for NEPA implementation, the Federal Railroad Administration (FRA) initiated a two-stage EIS for improving intercity passenger rail service in the Northeast Corridor. 283 Multi-state transportation projects of this scale often encounter delays attributed to conflicting jurisdictions, overlapping authorities, and interagency conflicts.
To avoid these delays, the FRA used the NEPA process to engage stakeholders early. 284 For example, to overcome the challenge of inter-agency variance in decision-making, formal points of contact were established for each federal and state resource and regulatory agency. 285 This early effort enabled agencies to speak to the FRA with "one voice." Engaging stakeholders as collaborative partners in NEPA compliance (for example, developing a purpose and need statement, formulating alternatives, and developing impact assessment methodology) facilitated coordination. Partner agencies could provide timely information that the technical team utilized, avoiding conflict down the road. 286 The communication protocols also enabled the creation of an interactive dataset encompassing multiple local and state jurisdictions, transportation authorities, and watersheds that could be used for other environmental analyses. 287 Though this collaborative process imposed demands on agencies' time that were uncommon on the front-end, it avoided conflict on the backend. 288 Moreover, the communication protocols, data-sharing, and decision-making procedures developed during the NEPA process created a framework for interagency collaboration that would foster continued efficiencies beyond project implementation because future projects can utilize the established inter-jurisdictional database and communication protocols.
Third, utilizing the NEPA process to develop consensus can avoid delays caused by conflict and expand agency resources through partnerships. This has been demonstrated in several pilot collaborative forest planning initiatives. For example, in 2012, the Forest Service completed the 4FRI EIS, which analyzed the largest number of acres in Forest Service history for restoration-based mechanical treatments. 289 The project goal was to restore the ponderosa pine forest stretching across northern Arizona (incorporating four different national forests), while reducing the threat of destructive wildfire to communities, rehabilitating ecosystems, and sustaining forest industries that strengthen local economies. 290 It was "the largest collaborative landscape-scale restoration initiative in the country, the largest initiative of its kind ever endeavored." 291 Despite its ambitious scale, the EIS was completed more quickly than the average (mean) timeframe for EISs completed that year. 292 Although not specifically included in the reports, it is likely that adequate funding and high prioritization of the planning effort helped speed completion. When it came to implementation, the Forest Service was not delayed by litigation. 293 This result was possible because the collaborative process increased stakeholder support for the 286 Id. at 3 (noting particularly that agencies expressed appreciation for being engaged before project alternatives were developed as opposed to a "post-decisional" consultation). 287  These are some examples of how focusing on public engagement and informed decision-making, rather than analytical downsizing, can produce long-lasting efficiencies. An annual inter-agency, inter-governmental training hosted by the CEQ highlighting "lessons learned" from the past year would help propagate best practices. Further research is warranted to explore additional best practices for conducting thorough, transparent, and efficient NEPA analyses at each level of review. These future studies should focus on best practices for effectively scaling lower levels of analysis, 294

Increase and Stabilize Agency Capacity
Inadequate staffing, a lack of experienced staff, unpredictable staff availability, temporary reassignments, and inadequate or unstable funding were frequently identified as sources of delay. This theme arose in GAO reports identifying delays associated with specific activities. It surfaced again in each of the EADM Roundtables. And it was echoed in industry comments regarding sources of delay in the NEPA permitting process. Problems associated with inexperienced staff plague multiple agencies. In a 2004 Rocky Mountain Mineral Law Institute Article, Laura Lindley emphasized "inexperienced and/or unempowered team leaders" as a major source of delay in the oil and gas permitting process. Specifically, she noted that the interdisciplinary team leader "may be preparing his/her first EIS." 303 The "lack of training results in unnecessary wasted time" including "failing to tier to earlier documents, focusing on formatting or other non-substantive details, re-creating the EIS format or layout each time [and] failing to focus on the proposed action and reasonable alternatives." 304 Where the document is written too narrowly, project changes require a new analysis. For example, where other drilling occurs while the NEPA document is being produced, an applicant may revise its plan with respect to spacing or anticipated number of wells. "The result can be the need to commence an additional NEPA document as soon as the current one is completed." 305 In other words, inexperience causes delay.
The importance of agency capacity in avoiding NEPA delays was also emphasized by Helen Serassio, who spent fourteen years working at the Department of Transportation. "Insufficient staff and resources are two of the biggest hurdles federal agencies face when working to meet their NEPA requirements in a timely manner. 2011 found that a lack of Forest Service staff trained in NEPA had led to a backlog of more than 3,500 expired special use authorizations that were awaiting NEPA review. 307 Even Congress recognizes that funding increases efficiency. For example, the first legislative infrastructure bill devoted to increasing the efficiency of the permitting process for infrastructure projects included a funding mechanism to help agencies achieve established timelines. 308 Several years later, Congress explicitly recognized the connection between prompt environmental review and financial resources by directing that "adequate resources," devoted to ensuring that expeditious environmental reviews are implemented, be made available. 309 That language was retained in later legislation and remains in effect. 310 Increasing and stabilizing funding for staff with expertise in environmental planning and decision-making would improve NEPA efficacy. Funding to develop and train interdisciplinary team leaders, resource specialists, and avoiding staff reassignments during a project would reduce delays. Providing funding to support landscape scale environmental analyses to which project-level decisions can be tiered would enable agencies to realize efficiency gains. Stabilizing funding for environmental planning and monitoring would help agencies develop interagency databases, collaborative protocols, and landscape scale analyses that could produce long-lasting efficiencies across agencies. Without addressing these common-sense sources of inefficiency, efforts to systemically improve the NEPA process will falter.
Litigation aversion was repeatedly identified as a source of delay, even though only a small percentage of decisions are litigated. 311 Governmentwide, only about two-tenths of one percent of more than 50,000 NEPA decisions that are documented each year result in litigation. 312 Litigation rates are higher for the Forest Service than for the government as a whole. 313 An investigation by the GAO regarding Forest Service fuel reduction projects from fiscal years 2006 through 2008 revealed that only 29 out of 1,415 decisions were litigated, and litigation impacted about 1% of lands slated for fuel reduction projects. 314 Rather than attempting to avoid litigation by developing overly expansive and detailed documents, the Forest Service could acknowledge litigation as part of the transparency function of NEPA. This shift in focus would enable agencies like the Forest Service to encourage field officers to act promptly. Indeed, the CEQ encourages agencies to focus the analysis on significant issues and refine the breadth of issues to address through a scoping process. 315 Selecting the issues of importance is an exercise of discretion, which is subject to judicial deference. 316 Public participation helps justify the exercise of that discretion. For example, although it is not required by the regulations, providing a public scoping process and publishing a draft EA provides an opportunity for the agency to document and justify the reasons for distinguishing between significant and non-significant issues and limiting the scope of the EA. It may seem counter-intuitive to achieve efficiency by inviting public comments on an EA, however, this approach enhances efficiency in five ways. First, it facilitates compliance with other statutory obligations that require public participation. 317 Second, it provides an opportunity for the agency to ensure that it has focused on the significant issues. "If agency staff truly understand the public's concerns at the beginnings, they can avoid spending time and money on issues in which the public has no interest." 318 Third, the response to comments provides a public forum for explaining the agency's decision for focusing the scope of the analysis, which builds a record enhancing the likelihood of success in litigation. 319 Fourth, providing an opportunity for public comment narrows the range of claims that can be litigated and ensures that an agency is not surprised by an issue raised for the first time in litigation. 320 NEPA litigants must generally raise their objections during the administrative process to preserve their right to litigate. 321 Litigants are also generally barred from raising issues not aired during the administrative process. 322 No such limits exist where agencies forgo public engagement. Finally, public participation provides an opportunity to identify controversial issues and may help diffuse tensions surrounding controversy. 323 It is also helpful to remember that litigation may serve a positive function. As Robert Dreher, a professor at Georgetown testified, "[c]ritics overlook the essential role that the independent federal judiciary plays under NEPA. When Federal agencies fall short, citizen suits are the only mechanism that enforce the act's commands for environmental review and public consultation." 324 There may be some projects that simply should not move forward without additional consideration or mitigation. Litigation provides this procedural backstop. Even though litigation is rare, 325 it often has merit.
Just 0.22% of NEPA decisions result in litigation, 326 and a recent study of NEPA litigation observed that environmental plaintiffs won more often at both the district court and appellate level than other litigants. 327 The authors concluded that low rates of challenge and high rates of success provide "strong evidence that NEPA litigation is grounded on legitimate claims," rather than strategic efforts to delay government projects. 328 These studies affirm Professor Dreher's observation. When federal agencies fall short, citizen suits enforce agencies' statutory duties. In practice, accepting the risk of litigation requires experienced and knowledgeable staff who are capable of utilizing the discretion afforded to agencies, and who feel supported by their superiors. That demands expertise and an investment in personnel. Promoting a culture of action, rather than incentivizing avoidance, may help avoid NEPA decision-making times that are elongated by fears about blame and job security.

B. Changes to Avoid
Our research confirmed the observation made by the Congressional Research Service that many delays blamed on NEPA actually arise elsewhere. Common external sources of delay identified in our research were inadequate staff and funding, operator decisions and market influences, coordination with other entities, and compliance with other legal or regulatory requirements. Many of the "changes to avoid" discussed below fail to recognize these common causes of delay. It is also important to remember that NEPA's charge is to make transparent and informed decisions, and while efficient decisionmaking is important, speed may not be the best measure of efficacy.

Treating the Wrong Problem
The regulatory changes introduced by the CEQ in 2020 were intended to "facilitate more efficient, effective, and timely NEPA reviews by Federal agencies." 329 To achieve this result, the new regulations impose page limits, eliminate the requirement to consider the cumulative effects of a project, and mandate aggressive deadlines. 330 These reforms treat the symptom not the rate at which agencies prepare EISs, and that the rate of NEPA litigation is declining while general civil litigation against the federal government is on the rise). 326  the cause and leave agencies vulnerable to violating NEPA's statutory mandate of transparency and deliberation. First, page limits stand in contradiction to NEPA's mandate of fulsome disclosure. 331 Imposing page limits on a disclosure document is like imposing page limits on a telephone book. The only way to meet the page limits is either to remove relevant information or reduce the scope of the disclosure. Neither of these two approaches meet NEPA's aims of transparency and public engagement.
Second, attempting to streamline NEPA by eliminating the scope of required disclosure is like treating a water leak by turning off the water-it ends the problem, but it does so at the expense of the entire program. In an era of compounding challenges (like climate change, drought, urbanization, and wildfires) a myopic analysis of effects will not facilitate agencies' abilities to achieve NEPA's mandate of deliberate and informed decision-making. Third, arbitrary page limits and deadlines may have unintended consequences. Indirectly encouraging agencies to cut projects into bite-sized analyses that meet the regulatory page limit standard could result in legally impermissible segmentation. 332 Furthermore, during judicial review of NEPA compliance, courts evaluate compliance with NEPA's statutory procedures and assess whether the agency took a hard look at environmental consequences and shared that information with the public. 333 Previous research observed that there is an inverse relationship between the amount of time spent preparing an EIS and the likelihood that an EIS will be challenged in court. 334 Other research suggests that rushed EISs may be more likely to require supplementation, which causes unintended delay. 335 331 42 U.S.C. § 4332(C). 332 "'Impermissible segmentation' occurs when parts of an otherwise 'major' federal action have not been evaluated together in the same NEPA document-'segmented'-in order to avoid conducting the NEPA analysis that would be required if the segmented actions had been evaluated together." Oak Ridge Env't Peace All. Arbitrary deadlines and page limits may, in short, make it more difficult for agencies to demonstrate that they met their statutory obligations.
Finally, aggressive deadlines may undermine NEPA's function as an umbrella statute coordinating compliance with other statutory and permitting requirements. For example, a commercial logging project may require road building across a wetland and through sensitive wildlife habitat contiguous to tribal lands. In addition to requiring a NEPA analysis, this project would also likely trigger permitting requirements with the U.S. Army Corps of Engineers for a fill and dredge permit under the Clean Water Act, consultation with the Fish and Wildlife Service under the ESA, and consultation obligations with the Tribe under the NHPA. All of these statutory obligations are independent of NEPA's obligations and are not subject to its regulatory deadlines. Requiring the NEPA process to be completed independent of these interrelated statutory procedures would be inefficient, time-consuming and confusing.
These reforms treat the wrong problem, are unlikely to produce beneficial results, and may have unintended consequences that result in project delays.

Avoid Inviting Unintended Consequences
Some proposed reforms invite unintended consequences that may decrease long term efficiency by increasing NEPA's complexity and inviting litigation. Three examples illustrate this possibility. First, in an effort to avoid perceived delays from the NEPA process, Congress has revised the NEPA process by creating legislative CEs for specific federal actions, mandating streamlining processes, limiting participating agency input, imposing unique administrative review requirements, and limiting public participation. 336 This ad hoc approach creates a complex and confusing compliance matrix with varying legal standards depending on the proposed action and the agency or agencies involved. Having different NEPA requirements for various federal agencies makes a combined analysis difficult and could also lead to unpredictable judicial determinations. 337 Inconsistent requirements also create challenges for stakeholders and cooperating agencies who may need to respond to multiple and inconsistent agency requirements. The network of shortcuts may therefore be less efficient than a clear and consistent path forward. Second, multiple "streamlining" bills introduced in Congress establish mandatory deadlines with financial penalties for agencies that miss a deadline and de facto approvals if the NEPA analysis is not completed within the 336  deadlines established for the act. 338 Imposing financial penalties on agencies with limited funding will only exacerbate delays caused by limited funding. Mandatory approvals if arbitrary deadlines are missed creates an incentive to game the system and foster delays in the hope of receiving a permit by default. And prioritizing speed over deliberation leaves society vulnerable to projects with unjustified and unmitigated environmental effects.
Third, the temptation to fast-track politically favorable projects through vast categorical exclusions subverts the ability to consider environmental consequences. For example, within the Forest Service, the desire for speedy action has led to a proliferation of regulatory categorical exclusions authorizing large scale vegetation management, timber sales, logging, thinning, and prescribed burning. 339 Additionally, wildfire risk has led to statutory categorical exclusions authorizing massive operations in the name of hazardous fuel management. 340 Fast tracking projects in large CEs results in limited deliberation, truncated consideration of alternatives (if any), and scant assessment of the indirect and cumulative impacts. While extraordinary circumstances can limit the availability of a CE, cumulative effects are not included in the list of extraordinary circumstances. 341 Thus, a forest could endure a thousand cuts authorized in CEs without undertaking NEPA's requisite "hard look" or meaningfully engaging with those who will most likely suffer injury.
According to investigative research by WildEarth Guardians, during the first quarter of 2020, Regions 1 through 6 used CEs to authorize hazardous fuel support life on earth." 347 Almost 30 years later, Dinah Bear, who served as General Counsel for CEQ for a total of twenty-two years, characterized NEPA as a process "grounded on certain basic beliefs about the relationship between citizens and their government." 348 Those beliefs include "an assumption that citizens should actively participate in their government, that information matters, that the environmental impact assessment process should be implemented with both common sense and imagination, . . . that there is much about the world that we do not yet understand. . .
[and] that the social and economic welfare of human beings is intimately connected with the environment." 349 These complex and multi-faceted goals cannot be achieved by implementing every proposed federal action exactly as it was originally envisioned or by boring holes through substantive and procedural requirements. A fully functioning NEPA will allow simple projects to pass through its review process quickly, while more complex projects will take time. Projects with unacceptable environmental effects may require mitigation. Within this process, a slow decision is not necessarily a bad decision.

VI. CONCLUSION
When considering strategies for streamlining or reforming NEPA, it is important to remain focused on NEPA's objectives. Fifty-one years ago, Congress recognized "the profound impact of man's activity on the interrelations of all components of the natural environment, particularly the profound influences of population growth, high-density urbanization, industrial expansion, resource exploitation, and new and expanding technological advances." 350 In response, Congress directed agencies to "utilize a systematic, interdisciplinary approach which will insure the integrated use of the natural and social sciences and the environmental design arts in planning and decision-making." 351 NEPA's twin goals are to foster public engagement in agency decisions, and to facilitate informed agency decisionmaking. Congress believed that a "hard look" coupled with public engagement would produce less impactful and more sustainable decisions. 352 These lofty ambitions can be achieved without compromising efficiency.
Reviewing over 41,000 NEPA decisions made by the Forest Service over a 16-year period, we observed that reports on average decision-making times across agencies are skewed by outlying decisions with extended timeframes. Focusing on the median decision-making times reveals that the majority of decisions adhere to a more predictable timeframe that is shorter than reported averages. Moreover, level of analysis does not dictate decision-making times. The fastest 25% of EISs are completed more quickly than the slowest 25% of EAs, and the fastest 25% of EAs are completed more quickly than the slowest 25% of CEs. This overlap demonstrates that efficiencies can be achieved at each level of analysis without foregoing the "hard look" required by NEPA. Focusing on activities associated with delay revealed that many sources of delay attributed to NEPA are caused by external factors. Some of these delay factors, like inadequate staffing, insufficient funding, time spent on inter-agency coordination, and litigation aversion can be addressed through fiscal and cultural reforms. Other sources of delay, like delays obtaining information from permittees, are not caused by NEPA and should not drive NEPA reforms. Finally, when used properly, NEPA's function as an umbrella statute and can mitigate or avoid delays caused by compliance with other statutory and regulatory requirements. We hope that our work, focusing on real-world problems causing delay within NEPA implementation, will provide a springboard to reforms that improve NEPA efficacy and advance the twin goals of public engagement and informed decision-making.

APPENDIX 1: THE REGRESSION MODEL
We used a weighted least squares regression model to predict elapsed time on a log scale. 353 A plot of elapsed time after the log transformation is below.
The equation for the model is provided below. 1. Model Equation is a categorical variable with three levels: CE, EA, and 353 We used a weighted least squares model because residual plots from the ordinary least squares model also showed unequal variances from one level of analysis to the other. Essentially the magnitude of the "miss" for our predictions varied by level of analysis. This is referred to more formally as "heteroscedasticity" and requires a weighted least squares regression model. [Vol. 46:S EIS • is a numeric variable representing the year in which the Forest Service initiated the NEPA analysis for a project, and is scaled so that year = 0 is 2004.
• ! is a numeric variable representing the potential quadratic trend over time.

•
. × is an "interaction" term between level of analysis and year. It allows CE, EA, and EIS cases to all have separate linear trends over time.
• . × ! is an "interaction" term between level of analysis and year squared. It allows CE, EA, and EIS cases to all have separate quadratic trends over time.
is an "interaction" between region and level of analysis. It allows the effect of each region on duration to change from one level of analysis to another.
• is an indicator variable for each activity found in the data. This model tested independently for each activity. Several projects included multiple activities. The model considers the type and number of activities included in each project-a dynamic we referred to as the "complexity" of the project.

Model Efficacy--R squared Results
R-squared is a statistical measure of the proportion of the variance for a dependent variable (in our mode time to complete the NEPA analysis) that is explained by the variables in a regression model. The R-squared value for the weighted least squares regression model was 0.248 and the adjusted Rsquared value was 0.246. The proximity of these values indicates the absence of unnecessary or redundant independent variables in the model. The R-squared value of 0.248 indicates that of all the variability in elapsed time on a log scale across all cases, 25% is explained by knowing the level of analysis, year, region, and activities involved in the case. As discussed below, each of the independent variables influence the elapsed time for a NEPA case, but there is still substantial variation in elapsed time that cannot be accounted for by the level of analysis, year, region, or activities involved.
3. Model Accuracy--Root Mean Square Error (RMSE) The root mean square error or RMSE for our model was 1.003. This can be interpreted as the "average" or "typical" miss in our prediction of elapsed time on a log scale. To ensure this value is unbiased, we performed crossvalidation analysis. 90% of the data was used to "train" or develop the model, and the remaining 10% was held back to "test" the model developed from only the 90%. For three different iterations where the training data set and testing data sets were randomly selected, the average RMSE was 1.001. This validates our RMSE, and indicates that if we use our model to predict the elapsed time for a future NEPA case, the typical error will be just over 1 on a log scale. Given that the overall average duration for elapsed time on a log scale is around 5, the relative error of prediction is approximately 20%.