Evaluating the Performance of Environmental Streamlining:
Development of a NEPA baseline for Measuring Continuous Performance

Previous | Index | Next

3. 1 Research Approach Overview

This research study has been designed in order to provide a more comprehensive, less subjective, and statistically-based approach to identifying NEPA process delays and evaluating their impact on time and cost of the overall project delivery process than any other research effort to date. It has been the intent of this study from the outset to analyze NEPA process delay implications from a very broad historical and geographical perspective, and to focus only on those projects that have actually been delivered in their entirety (i.e. , constructed and fully operating). It has also been the intent of this study to utilize several existing and extensive data sets in a manner that would allow at least some statistical analysis of the information in those data sets to be conducted on an essentially random sample of projects taken from the total known universe of relevant projects.

Given the complexity and untested feasibility of using these separate data sets, especially in conjunction with each other, a limited sampling of projects was first selected in an effort to test the data in a meaningful way. On the basis of the preliminary research, a Data Collection Strategy Memorandum (Louis Berger & Associates, Inc. , March 2000) was prepared to identify an approach that could provide all of the necessary data for undertaking the greater research effort. Three primary sources of data useful to the proposed research were identified and assessed:

• Northwestern University's Transportation Library EIS Database — This database consists of over 4,000 EIS documents prepared for FHWA projects around the country, and purportedly includes every FHWA EIS prepared since the inception of NEPA;
• FHWA's Fiscal Management Information System (FMIS) Database — This database, which contains more than 3 million records and includes a variety of financial appropriation information for all highway projects that have been financed using federal funds since the 1940s, comprises part of FHWA's overall accounting system; and
• Publicly-Available Internet and Commercially-Available Software Aerial Photo and Mapping Databases — These data sets include: (containing aerial photography since 1980 and USGS topoquads); (containing localized mapping based on U. S. Bureau of Census Tiger Files from 1990); and MapPoint 2000 CD ROM (containing more detailed, larger scale mapping based on U. S. Bureau of Census Tiger Files from 1990).

Of these, the EIS documents provided the means for collecting detailed information about each project included in the sample set such as proposed improvements, environmental conditions and impacts, duration of the NEPA process, etc. The FMIS database, in contrast, provided detailed information about dates, types and amounts of appropriations for specific project segments, and provided the ability to assess individual phases of the project development process. Finally, the aerial photo and mapping databases provided the means to ascertain whether or not a project has been fully constructed as proposed in the EIS.

From the preliminary research that was conducted, it was concluded that the three database sets together could, in fact, provide most of the data that was pertinent to the overall research study. However, it was also concluded that integrating them on the basis of individual EIS projects is confusing, complex and extraordinarily time consuming. Despite the difficulties, it was further concluded that these databases, used in combination, provide the best opportunity to develop a comprehensive, thorough and objective set of data that can be used in creating a statistical baseline for depicting the NEPA process and its relationship to the overall project delivery process. Details of the process required for integrating the three database sets according to individual EIS projects are provided in Section 3. 5 of this report.

It should be noted that it was the original intent of this research study to evaluate all three levels of NEPA documentation (i.e. , EIS, EA/FONSI, and CE) in terms of delay. However, it was determined that only EIS projects can be identified from a centralized database. Unlike in the case of EISs where the Northwestern University World Wide Web site provides a list of all EISs prepared by FHWA and the Northwestern University Transportation Library makes those EISs available for review, no such central source of information or readily available repository of EAs/FONSIs or CEs exists. In conducting the preliminary research, one FHWA division office was contacted in order to identify what happens to EAs/FONSIs and CEs upon completion of design of such projects. According to that particular office, the environmental documents are included in hard copy along with design documents and sent to a national archive in St. Louis, Missouri. Other archive locations for storing similar documents have also been noted to exist elsewhere in the country. It was implied that these documents are not generally maintained at the FHWA division offices after a project is completed and that they are put into permanent storage at one of these archive facilities. The ability to identify and retrieve such documents would be problematic, at best.

Given the apparent inability to access EA/FONSI and CE documents from any known sources, it was ultimately decided that this study should focus only on EIS projects. This decision also makes sense from a statistical perspective. A sample size of 100 projects had been predetermined to be used for this study since such a sample was considered to be sufficiently large to avoid sampling error. However, if that sample size was disaggregated to include all three levels of environmental documentation, then the sample for any one of the individual levels of NEPA documentation would not be large enough to protect against the potential for sampling error.

The decision to focus on EIS projects exclusively also makes sense from a practical standpoint. Even though the number of EISs that are prepared for Federal aid highway projects is dramatically fewer than the number of EAs/FONSIs and CEs prepared, the length of time required to comply with NEPA's EIS requirements is inherently greater. Therefore, the implication for the overall project development process in terms of delay time and cost is also greater when an EIS is involved.

Despite the fact that preliminary research was conducted on a select few projects in order to assess whether the basic methodology and the several sets of available data were usable and manageable, the full magnitude of the effort could not have been realized until the overall research study was underway. The most complex part of the study turned out to be the data collection element. Details of the problems and methods used in collecting the full range of data for the 100 sample projects are discussed throughout Section 3. 5 of this report.

The basic steps of the overall methodology used in this study, in the order that they were followed, are presented below:

Sample Selection

• From the list of EIS projects available from Northwestern University's World Wide Web Site, select a stratified random sample of 100 EIS projects to study, taking both geographic region and decade of each project into account;

Data Collection

• Obtain and review the EIS document(s) associated with each potential project to be included in the sample in order to identify locational and component parameters of the project;
• Obtain, parse and review FMIS Type 1 records, and match the appropriate records to each EIS project;
• Identify whether or not each EIS project has been fully constructed using information included in the appropriate FMIS Type 1 records and from the aerial photos and mapping available from the internet and CD ROM software;
• For those projects determined to be constructed, obtain, parse and review FMIS Type 2 records, and match the appropriate records to each EIS project;
• Record relevant data from all FMIS Type 2 records associated with each EIS project, and consolidate and record data on the basis of each EIS project as a whole;
• For those projects determined to be constructed, review the associated EIS document(s) to collect relevant project, environmental, and NEPA process data, and record the data on a project data form;
• Enter all relevant project data collected from the FMIS records, the EIS document(s) and aerial photography / mapping into a computerized database, and provide visual quality control checks for anomalies in the entered data;

Sample Selection Finalization

• Repeat the above steps until the total sample size of 100 constructed projects has been completed, taking from the next set of projects on the stratified random sampling list, as necessary;

Statistical Analysis

• Perform descriptive statistics analysis, including maximum and minimum values and, in some cases, frequencies, to identify additional anomalies in the data and to establish descriptive parameters of each data variable;
• Perform exploratory data techniques to look for differences in descriptive parameters among different FHWA regions;
• Create new data variables as necessary to complete statistical analysis by manipulating entered data variables;
• Examine data for normalcy of distribution and, where feasible, appropriate and necessary, transform data to obtain surrogates with a normal distribution;
• Test for differences among FHWA regions with regard to length of the NEPA process and other variables;
• Examine correlations between NEPA process time and other data variables; and

NEPA Process Baseline Development

• Utilizing the results of the statistical analysis, identify a NEPA process baseline or set of baselines under various conditions against which to evaluate future environmental streamlining efforts.

Tracing through the individual steps presented above provides a general indication of the overall methodology utilized in undertaking this research study. However, each of the individual components of the process involved its own set of substeps and analyses. A detailed discussion of each of these steps, including data limitations and problems encountered, is provided in subsequent subsections of this report.

3. 2 Description of Data Sources

As mentioned in Section 3. 1 above, there were three basic sources of data used for undertaking this research (i.e. , Northwestern University Transportation Library, FMIS Records, and on-line and commercially available aerial photo / mapping services). The composition, format and availability of these data sources are discussed in detail below.

3. 2. 1 Northwestern University Transportation Library

The Northwestern University Transportation Library and its World Wide Web site served as the starting point for undertaking this research. The library contains more than 4,000 EIS documents prepared for Federal-aid highway projects, dating back to 1970 when the requirements for NEPA documentation first went into effect. Since these EISs are purportedly available for all such projects undertaken in the last 30+ years, the projects covered by these EISs essentially form the total universe from which to draw the study sample. In this manner, the Northwestern University Transportation Library was really the backbone of the research.

The Library's internet on-line service provided a complete listing of FHWA's EIS projects that could be requested by state. The internet address at the time that most of the research was being conducted was It should be noted, however, that toward the end of the intense data collection process, the internet address changed to The ability to specifically request FHWA EIS data by state also became more difficult following the change in address and format of their web site. If the entire data collection effort for this project had occurred after the internet web site changed, the complexity of obtaining the EIS information in a usable way would have made it virtually impossible to continue.

The EIS database available at the time of conducting the research included the following information for every project: Title, Agency, State, EPA Number and Call Number, with an additional line for Notes which was usually blank. A sample printout of the EIS database requested according to state is provided in Appendix A.

As mentioned above, the complete on-line database was used to formulate the universe from which to select a stratified random sample for this research. Further discussion of the process used for selecting the research sample is provided in Section 3. 4 of this report.

In addition to serving as the sample universe, the on-line database provided the mechanism for requesting the actual EIS documents from the Library. After the potential sample EIS projects were identified, a system was established with the Library whereby e-mail requests were made for a set of documents. Initially, only five EISs could be requested and borrowed at a time, but eventually, the Library permitted EISs to be requested and borrowed for all projects within a given former FHWA region. The EIS documents available for loan from the Library generally included the Draft EIS, Final EIS, Supplemental Draft and/or Final Draft (as appropriate) and often, Technical Appendices. After receiving the requested documents through the mail, they were kept on loan for purposes of review for periods of up to a month before having to return them. This system continued until the entire research sample of 100 projects was selected.

Although the process of requesting and receiving EIS documents from the Library generally worked well, some delays in shipping, incomplete shipments and missing volumes did occasionally occur, which resulted in delays during this phase of the process. At one point during the data collection process, the Library physically moved into a new building, resulting in further delay while all of their EIS documents had been placed into boxes.

3. 2. 2 FMIS Records

The Fiscal Management Information System (FMIS) database was designed to keep careful track of funds approved and expended on all Congressional appropriations related to the Highway Trust Fund. The database is huge, involving more than 3 million individual records going back to the late 1940s. Each individual record can be as long as 350 ASCII characters.

The database is maintained by FHWA, whose staff uses the information for planning and executing program activities, evaluating program performance, and depicting financial trends and requirements related to current and future spending. The system provides instant access to current project data, as well as data on all closed Federally-funded highway projects. The data included in the system is entered in a variety of ways and from a variety of FHWA locations throughout the country.

FMIS data is organized into two basic record types, each involving numerous data fields. The first record type, denoted as Type 1, includes basic information about an appropriation, such as: appropriation number; project number; project location by state, county and MSA; a brief description of the project limits; a brief description of project activities; when the appropriation was approved; the date of last expenditure; etc. Additional FMIS data fields regarding environmental documentation were added to the Type 1 records around 1990, including NEPA Class of Action, EIS year, and EIS number. In the vast majority of the records, the environmental information was missing from the appropriate data fields, particularly for those records predating 1990, but also for many of the records entered after that date. There were also many records where one or more of the other data fields were blank as well. Missing data became particularly problematic when the descriptive fields regarding project limits and activities were left blank, since these data are critical for relating the FMIS records to each EIS project available from the Northwestern University Transportation Library (see Section 3. 5. 2 for details).

The second record type, denoted as Type 2, includes more detailed data about each appropriation, including but not limited to the following:

• nature of the project area (i.e. , urban, urbanized or rural);
• type of highway (i.e. , freeway, expressway, divided or undivided);
• functional classification (i.e. , principal arterial, minor arterial, major collector, minor collector or local);
• work class (i.e. , preliminary engineering, construction engineering, right-of-way or construction);
• work type (an extensive list of work types as presented in Appendix B);
• length of project;
• total funding;
• total federal funding;
• FHWA Region;
• a variety of appropriation milestone dates (i.e. , date of agreement, date reserved, date authorized, date awarded, date underway, date completed, date received, date of last action, etc. );
• number of lanes; and
• lane miles.

A further description of the specific FMIS data elements is provided in Appendix C. Of the Type 2 information available, the most critical data are those related to funding and start and completion dates for various activities. These data make it possible to obtain precise information for use in analyzing the time and cost required for project implementation. In this regard, there is no other comparably consistent source that exists for these data. It should be noted, however, that there is also an enormous amount of superfluous data in these records which are not relevant to the research at hand. It is the superfluous data that makes the overall database rather cumbersome and difficult to use.

Each record in the FMIS database, including Type 1 and Type 2 records, is based on an appropriation which funds a specific project activity or group of activities such as preliminary engineering, construction engineering, right-of-way acquisition, a particular aspect or aspects of construction, etc. for a given project or projects. These separate appropriations have been assigned their own 3-digit appropriation numbers within the FMIS. Each record also reflects a specific project or highway improvement. Separate 7-digit project numbers are assigned within the FMIS to each specific project or highway improvement.

Sample Type 1 records are presented in Appendix D while sample Type 2 records are presented in Appendix E. It should be noted, however, that the database information presented in both appendices has been reformatted to make the data more understandable and easy to use, since a single record taken directly from FHWA's database consists of a string of approximately 350 ASCII characters with no breaks between fields. The reformatted data presented in the appendices and which was actually used for this analysis was created by first using SPSS software to parse the data, and then Microsoft Excel software to present the data in an understandable format. The sample data presented in the appendices also excludes some fields of information that are extraneous to the research at hand.

The general methodology for using the FMIS database was to first request Type 1 records from FHWA's financial office on a statewide basis, and then organizing the data to focus on the particular county or counties where a given EIS project was found to be located. Once a project was determined to have been constructed, then the Type 2 records that are specific to that particular project were requested from FHWA. Details about how the Type 1 FMIS records were linked to each EIS project are presented below in Section 3. 5. 2, while details about how the Type 1 FMIS records were linked to the Type 2 records are presented in Section 3. 5. 4.

3. 2. 3 Aerial Photography and Mapping Sources

The various publicly-available internet and commercially-available software aerial photo and mapping databases were the primary data sources used to ascertain whether or not a given EIS project was fully constructed and is currently operating as intended. The three data sets used in this regard are as follows:

• This internet site provides aerial photography that generally dates from 1980 to 1998 and USGS topoquad mapping. The aerial photography consists of three basic types: Orbital Imaging, Spin-2 Satellite Imaging, and USGS aerials. All three aerial types are provided at a resolution of 16 meters per pixel length, although zooming to a resolution of 1 meter per pixel length is possible. This data source can be requested at the county and/or municipal level, depending on location. It should be noted that the actual availability of each of these aerial photo types and the USGS topoquad mapping varies by location, as does the date of each image.
• This internet site provides street mapping by county or municipality, and reflects localized mapping based on U. S. Bureau of Census Tiger Files from 1990. In many cases, the 1990 files have been updated to a more recent year. Some files were noted to have been updated as recently as 1999. Similar to the information, the mapping is provided at a resolution of 16 meters per pixel length, although zooming to a resolution of 1 meter per pixel length is possible.
• MapPoint 2000 — This commercially-available CD ROM software generally provides more detailed, and usually more current mapping based on U. S. Bureau of Census Tiger Files from 1990. Similar to the other data sources, the mapping is provided at a resolution of 16 meters per pixel length, although zooming to a resolution of 1 meter per pixel length is possible.

Details about how these data sources were used in conjunction with other information to identify whether or not an EIS project has been constructed and is currently operating are provided in Section 3. 5. 3 below.

3. 3 Limitations of Available Data

Although the three major data sources described above provide sufficient useable data to adequately conduct this research in the manner anticipated, several notable limitations of these sources, as well as other sources investigated, were found to exist which required that certain assumptions be made during the course of the research. These data limitations are discussed below.

3. 3. 1 Start and End Dates of the EIS Process

One type of data that was found to be unavailable from any central source was the official start and end points of the NEPA EIS process during the past 30 years. These data were necessary for identifying the total length of the NEPA EIS process for each project. Precise dates for Notices of Intent (NOIs), which are considered to be the official start dates of the NEPA process, and Records of Decision (RODs), which are considered to be the official end dates of the NEPA process, were obtainable from neither the FMIS database nor the EISs. During the preliminary research performed for this study, inquiries were made with FHWA headquarters, the U. S. Environmental Protection Agency (USEPA), and the Council on Environmental Quality (CEQ) to determine if there is a central source for these dates, and determined that such a source does not exist.

A physical check of the Federal Register was also conducted to determine the potential for using that source. The Federal Register was found to be available on the World Wide Web from 1995 to the present at the following address: http://www.gpoaccess.gov/fr/index.html. This is a searchable form of the Federal Register with a browse feature by term and by year. For entries prior to 1995, however, a hard-copy search must be performed in a Federal repository.

Further complicating a physical search is the change in indexing that occurred in 1984. Environmental Impact Statements are listed as a heading in the Congressional Information Services, Inc. (CIS) Federal Register Index which began its publication in 1984. Prior to that year, there is no such listing in the annual or quarterly Federal Register indexes. In the CIS Federal Register Index, EISs are listed under the Federal Highway Administration as "Hwy construction project plan, County State. EIS required." However, this information is only applicable to the Notices of Intent, as no such notation was available either prior to or after 1984 in either the standard Federal Register Index or the CIS Federal Register Index for Records of Decision.

For the sample projects tested during the preliminary research performed for this study, the ROD dates could not be found in the Federal Register for any of them. Even using the date of the signed FEIS to target a date for the ROD, the actual ROD date could not be verified using the Federal Register. It was, therefore, concluded that searching hard copies of the Federal Register is not an efficient use of resources, especially given the likelihood that the ROD date will not be listed for a given project.

In the case of NOI dates, they can generally be found on the standard Federal Register Index, although they are very difficult to find in many cases. Due to the fact that the headings vary over time, their locations within the index are inconsistent, often requiring hours to find a single specific entry. Total reliance on the Federal Register Index for identifying NOI dates, therefore, was considered to be a rather inefficient exercise, especially given that a reasonable surrogate date could be developed.

The possibility of uniformly contacting FHWA division offices to obtain dates for NOIs and RODs was considered during the preliminary research effort for this study. Based on previous experience with these types of inquiries, however, it was concluded that this would not have been a cost-effective option because of the time required to find the person who has the information, institutional memory problems, and the possibility that uneven data, and consequently statistical bias, would result. Given these limitations, the effort involved in uniformly attempting to solicit this information was concluded to be not worthwhile, even if the information was able to be obtained part of the time.

It was also concluded that the EIS data presented in the FMIS database was sketchy, at best. As discussed previously, the EIS data field was not included in FMIS until 1990. Therefore, for all projects that preceded 1990, which would constitute the majority of the projects considered as part of this research, the FMIS database would be of no value in identifying start and end points for the EIS process. For those projects that were entered onto the FMIS database after 1990, it was found that data were only sporadically included in the EIS data fields. It was also unclear exactly what data was actually reflected in the "EIS year" field, although it is likely that the field reflects year of EIS completion rather than year of EIS initiation. For these reasons, the FMIS database was concluded to be of little value in identifying the total length of the NEPA EIS process for each project.

Therefore, it was decided that the start and end points of each EIS project should be based directly on information available in the EIS documents themselves. Of these, the end points were easy, given that every FEIS document has an FHWA signature date identifying when the FEIS was completed. The year of the signature was selected as the end point, rather than arbitrarily attempting to add some additional time period (probably a few months) after the signature date to reflect the ROD date. Although it is likely that most RODs were approved within a few months of the signature date, there is no basis to verify that assumption, without conducting a statistical analysis to further test the assumption. Therefore, the EIS end date used for each of the projects included in the research sample was the date of signature. Only the year of the signature was recorded for use in the analysis.

Identification of the EIS start dates, however, required greater judgement on the part of the EIS reviewers. Since most of the EIS documents included some sort of project history or milestones discussion, a real or proxy start date was generally able to be identified. In a few cases, the actual date of the NOI was reported in the EIS, but in most cases, the start date had to be estimated based on other information presented in the project history. For instance, if reference was made to the date of a scoping meeting or an initial project meeting, the date of project initiation was assumed to be similar. Given that only the year of project initiation was recorded for use in the analysis, this seems to have been a reasonable assumption. In every case, there was some indication provided in the EIS document that allowed a reasonable estimate to be made about the EIS start date.

Although it is recognized that more judgement is involved in identifying the start of the EIS process in this manner than would exist if a detailed review of Federal Register documents were conducted for every project, the level of effort required to do the more detailed analysis was cost and time prohibitive. It is unlikely that the real start date for the EIS process would vary by more than one year in either direction from that shown for each project, and it is probable that even if there was a few month difference between the real date and the date inferred from the EIS document, that both dates would fall within the same recorded year. Therefore, the length of the EIS process as estimated in this study is considered to be reasonable for analysis, although it does not specifically reflect the time between date of signature and receipt of ROD.

3. 3. 2 Cost of the NEPA Process

Detailed project cost data is generally available in the FMIS database, beginning with preliminary engineering and proceeding through project completion. However, there is little information concerning the cost of the NEPA process itself, and any planning that occurred before the NEPA process. Although the FMIS database includes some categories for project planning prior to engineering, none of the FMIS project data reviewed during the preliminary research contained any such costs. As a general rule, NEPA documentation costs are included in preliminary engineering budgets, and cannot easily be separated out.

It is theoretically possible to provide a very rough estimate of the cost of the NEPA process using a series of wide cost ranges, based upon the number of documents and the number of agencies involved. However, the development of such a cost estimate would be extremely subjective and the utility of such a crude method for estimating cost would be questionable. Because time appears to be the principal concern of the NEPA process, the costs of the process itself have been given a lower priority. Therefore, no effort was made to develop a cost estimate of the NEPA documentation process for a particular project, unless the cost was specifically included in the FMIS database. Even then, costs would only have been included for those projects studied since 1990. In reality, NEPA cost information was unavailable for any of the sample projects included in this research.

3. 3. 3 Cost of Other Phases of EIS Projects

As mentioned in Section 3. 2. 2 above, the FMIS records contain information about milestone dates and costs associated with various phases of the project development process (i.e. , preliminary engineering, construction engineering, right-of-way acquisition, and construction), as well as other information. Although the milestone dates and cost information are usually provided for each individual record (i.e. , combined appropriation number and project number), an underestimation of total cost by EIS project is anticipated to have occurred due to the unavailability of certain descriptive information in the FMIS records required to link those records by EIS project. Details about the process required to identify and link the FMIS records on the basis of a particular EIS project are provided in Section 3. 5. 2.

The descriptive data field of each FMIS record usually includes some identifying reference to the particular segment of highway represented by a particular appropriation number / project number combination. However, if the descriptive data field has been left blank, there is no way to relate that particular FMIS record to the overall EIS project. Therefore, any FMIS records in which the descriptive data field has been left blank are essentially ignored as part of this research assignment since they are of little value without being able to relate them to a specific EIS project. It is estimated that such records represent less than 10% of the total records, although a thorough assessment in this regard was not performed.

The problem with ignoring these records is that they do contain cost information by appropriate project development phase(s) which would not be counted in the research. For instance, if a particular record with missing descriptive information includes a cost value for preliminary engineering, that record may, in fact, be associated with part of an overall project addressed in an EIS document, but the cost for that record would not be added to the other records comprising that EIS project. In this manner, the total cost of preliminary engineering would be undercounted for the overall EIS project. Obviously, as the number of FMIS records with missing descriptive information that should have been included as part of the overall EIS project database increases, the more undercounting that would occur. Although there is no way of estimating how many applicable FMIS records may have been ignored or how many EIS projects may have been undercounted, there is a likelihood that at least some of the project costs have been understated to some extent. Therefore, some caution is urged when evaluating the results of the impact of the NEPA process on project costs by phase.

Concerns regarding milestone dates related to unassigned FMIS records are less of an issue than costs. Because many of the milestone dates for individual records are included within the range of dates for all records encompassing an EIS project, it is unlikely that there are any important dates that have not already been incorporated. Therefore, the analysis of the impact of the NEPA process on project schedules by phase should not be compromised in any manner due to missing data and unassigned FMIS records.

3. 3. 4 Environmental Permitting Information

During the preliminary research performed as part of this study, it was concluded that, unlike Northwestern University's Transportation Library in the case of EISs, no central source for obtaining copies of environmental permits for specific projects exists. Although the focus of this research effort has always been to provide benchmarks related to the NEPA process rather than to emphasize delay and cost contributions associated with environmental permitting, any information regarding such permitting is considered to be useful in understanding interrelationships of NEPA with other factors and processes. An indication of the likelihood that environmental permits were actually acquired for a given project was possible on the basis of reviewing the EIS documents associated with that project. Verification of the receipt of such permits and details of the permit requirements, however, can only be provided through a review of the actual permit materials. Due to the unavailability of the permit documents from a centralized location, a simple identification of whether or not a particular permit (e. g. , Section 404 Wetlands Permit, U. S. Coast Guard Bridge Permit, etc. ) was required for a project was noted during the EIS reviews. It should be noted that this limitation in data availability is not considered to be a critical shortcoming to the intent or implementation of this research.

3. 4 Selection of the Research Sample

As discussed in Section 3. 2. 1 above, the Transportation Library at Northwestern University, which has an on-line internet service purportedly identifying all of the FHWA EIS projects prepared since inception of NEPA, served as the basis for selecting a representative sample of FHWA EIS projects for analysis as part of this study. In effect, the Library's total database of approximately 4,000 EIS projects served as the universe from which the research sample was drawn.

It was determined from the outset that a sample size of 100 projects was desirable from both statistical and cost perspectives. In fact, such a sample size was the smallest considered to be acceptable in terms of minimizing sampling error. Referring to the widely-used Central Limit Theorem,^[5] estimates of average project delays due to NEPA should be unbiased, although some degree of sampling error is expected. This sampling error decreases as sample size increases. It was estimated that the sampling error associated with a sample size of 100 projects is approximately 9. 1%, indicating that there is up to a 9. 1% possibility that the particular sample chosen for this research could be expected to differ from the total universe of projects. Although larger sample sizes would result in smaller sampling error, the potential reductions in error begin to be overshadowed by the extensive additional cost and effort required to collect data and analyze for a larger sample size. Trading off the cost of sampling and analysis versus the degree of estimated statistical error, it is considered that an error of less than 10% is acceptable for this research study. Therefore, a sample size of 100 projects is believed to provide a reasonable level of statistical validity at a reasonable cost.

The data from the Library's World Wide Web site, which was available at a statewide level, was downloaded and categorized according to each of the nine regions formerly used by FHWA. The projects were then sorted by year within each region, and then numbered to allow for random number selection within each region. As part of the stratified random sampling process, the 100-project sample was disaggregated into the nine separate regions, with 11 projects to be included in each in order to ensure geographic diversity. Since this process only resulted in 99 projects, the 100^th project was randomly selected from the entire database and added to the appropriate region.

The original intent of the stratified random sampling process was to also stratify the projects within each region by each of the three decades since inception of NEPA. A further constraint was originally intended to be imposed which would limit the number of projects to be selected from the 1970s to three within each region, or roughly 27% of the sample. The remaining eight projects in each region would then be split between projects from the 1980s and the 1990s. This limitation was primarily proposed to be employed given the over-representation of NEPA EISs in the 1970s and the fact that the EISs prepared during that decade were dramatically different in terms of scope and requirements from those prepared in later decades.

The first cut at developing a random stratified sample did in fact stratify on the basis of decade as well as former FHWA region. Each of the nine regions was initially assigned 15 randomly selected EISs using the above-stated decade stratification, with the intention of selecting 11 projects that were found to have been actually constructed within each. However, in most cases, additional projects had to be added to each regional sample pool when it was found that 11 completed projects could not be selected from the initial 15. In some cases, more than 20 documents were analyzed for a given region before the sample of 11 could be completed. As new projects were added to each initial regional sample pool, stratification by decade was eventually eliminated although stratification by region continued.

The stratification by decade was primarily eliminated because, in some cases, several hundred randomly numbered projects within each region failed to produce the necessary number of projects by decade. Since the vast majority of projects in each region dated back to the 1970s, none of those projects could be considered further after the initial allocation of 1970s projects was made. Many other projects were eliminated from the random number pool when it was determined that FEIS documents were not available from the Northwestern University Transportation Library. Availability of an FEIS was essential since the preferred alternative would have been described in that document.

Another reason for eliminating stratification by decade was the difficulty in finding EIS projects prepared in the 1990s that were fully constructed. Roughly at the midpoint of the data collection process, it was concluded that any EISs completed between 1995 and the present were unlikely to have been constructed, based on other projects of that vintage that had already been researched. Thus, all projects from 1995 to the present were removed from the project pool and replacement documents were randomly selected from the Northwestern University Transportation Library database. All remaining projects with EISs having signature dates in the early 1990s were then prescreened in terms of their probability for having been fully constructed before requesting any documents from the Library, in an effort to reduce the number of EISs requiring review.

For these reasons, stratification by region became very problematic and an inordinate amount of additional time was determined to be required in order to continue on that basis. Therefore, it was ultimately decided to eliminate that constraint by allowing additional 1970s projects to be randomly selected within each region. The final sample consisted of 38 EIS projects completed in the 1970s, 52 EIS projects completed in the 1980s, and 10 EIS projects completed in the 1990s. It should be noted that some of the projects that have been identified as 1980s projects, particularly those completed in the early part of the decade, were largely prepared during the 1970s and may be more indicative of that decade. For purposes of this analysis, however, they are appropriately included as 1980s projects.

By the time 100 constructed and completed projects was identified for inclusion in the research sample and placed into the final study database, a total of 162 projects had been investigated in terms of initial EIS review and making a construction completion determination. Most of the additional 62 projects had been eliminated because they were determined to have not yet been fully constructed, or construction completion could not be confirmed either way. In a few cases, projects had to be eliminated because certain critical information (e.g., milestone dates from the FMIS records) were found to be missing in their entirety.

The complete list of 162 projects considered and the 100 actually selected for the research sample is presented in Appendix F. This list presents the projects according to former FHWA region, and also identifies those projects added to each region after the initial 15 projects were considered.

3. 5 Data Collection Methodology

The data collection methodology consists of several basic steps, as presented in summary in Section 3. 1. The details of each basic step are provided in the following subsections.

3. 5. 1 Initial Review of EIS Project Details

Upon receipt of each set of EIS documents from the Northwestern University Transportation Library, they were reviewed in a cursory fashion to determine the specific parameters of each project (e. g. , type of project, limits of project, location of project, etc. ). If the project consisted of multiple components (e. g. , a roadway widening and a new interchange), they were noted. Graphics included in the EIS documents were particularly useful. It should be noted that this review focused on the design of the preferred alternative as stated in the FEIS, as it was assumed that the other alternatives identified were not ultimately designed and constructed.

3. 5. 2 Linking FMIS Type 1 Records to EIS Projects

One of the key challenges of dealing with the FMIS database was determining how to aggregrate all relevant records on the basis of each project as identified in an EIS. It was determined during the preliminary research conducted for this study that the number of individual appropriation number / project number combinations included in the FMIS records that are related to a given EIS project is often voluminous. This is due to the fact that each individual FMIS record may only represent one or several small aspects of work associated with the entire project (e. g. , an appropriation for some specific types of construction along some smaller segment of highway included as part of the overall project identified in the EIS).

After sorting the FMIS Type 1 records by county, they were linked to a specific EIS project (i.e. , the preferred alternative shown in the FEIS) via the Type 1 descriptive data field information provided and the dates related to those records. When the descriptive data field in a given FMIS record appeared to be somehow related to the overall project shown in the FEIS, then the type of work and milestone date fields were also checked for correlation to the FEIS project. If the location, description of work and milestone dates for a given record all appear to coincide with the specifics of the project shown in the FEIS, then it would be linked to that project. Similarly, all other records having locations, descriptions of work and milestone dates that appear to coincide with the FEIS project would also be linked to that project. Once all Type 1 records that appear to be linked to an EIS project have been identified, then Type 2 records, which have more detailed information, would be requested from FHWA's database.

For purposes of illustrating this process, the 20-mile circumferential Route I-287 project from Montville, New Jersey to the New York State Line is used. This project, which was the subject of an FEIS prepared in the early 1980s, was evaluated as part of the preliminary research conducted for this study. Even though it is not one of the projects that was randomly selected as part of the sample pool, the process involved in linking the FMIS Type 1 records to the overall EIS project is similar to that used for projects included in the pool.

In order to identify all of the FMIS Type 1 records related to the Route I-287 project from Montville, New Jersey to the New York State Line, the entire file of Type 1 records in New Jersey first had to be opened using Excel software. There were a total of 10,113 records in the New Jersey file. In order to make the number more manageable, they were then sorted in the following manner:

• Records were sorted in ascending alphabetical order by the County field, and records not pertaining to the relevant counties (i.e. , Morris, Passaic and Bergen) were removed. This selection left 1,653 records.
• Most New Jersey projects pertaining to Interstate routes have the route number as the first set of digits in the Project Number field (non-Interstate routes and Interstate routes in some other states, however, do not necessarily have obvious identifying Project Number structures). The Project Number field was sorted in ascending numeric order and records not containing "287" as the first three digits of the Project Number field were removed. This selection left 134 records. For cases where route designation numbers are not part of the Project Number field, it is necessary to rely on the descriptive information provided in the record in order to identify the specific route or highway. This field could be sorted or searched for numeric or text strings to produce similar results.
• The Project Description and Work Description fields for all records pertaining to I-287 were then reviewed and those records not pertaining to the subject project were deleted. Judgements about which records to retain were made based on descriptions of the geographic limits of the work in the Project Description field and the type of work in the Work Description field. Records relating to work on I-287 from Montville to the New York State Line were retained (examples include descriptive entries for "Franklin Avenue to S. of Ramapo River" or "Passaic-Bergen Cty. Line to Camp Gaw"). All records relating to the correct geographic area of the subject project but with work items that did not appear to be related to the project were excluded (e. g. , roadway resurfacing). Examination of the milestone date fields was also important to confirming the exclusion of these types of records. Of the 134 total records related to Route I-287 as a whole, 28 were determined to be directly related to the subject project from Montville, New Jersey to the New York State Line.

Needless to say, as the above illustrative example indicates, the process of linking the FMIS Type 1 records to EIS projects is complex, judgmental and time consuming. However, the process was proven to be workable and to provide reasonable information for use in the overall research analysis.

3. 5. 3 Verification of Project Construction Completion Status

An important requirement of this research is that each project included in the research sample be constructed to the point where it resembles the project as proposed in the FEIS and when it is fully operational to traffic as intended. The process of determining project completion routinely included several steps, the first of which involved review of any available aerial photography and mapping. The use of publicly-available internet and commercially-available CD ROM aerial photo and mapping services was discussed as a data source in Section 3. 2. 3.

When recent aerial photography was available, analysis of the photos was useful in determining project completion where a new facility such as an interchange, bridge, or roadway segment was a component of the preferred alternative identified in the FEIS document. In some cases, depending on the date of the aerial photo and the status of completion shown to exist at that time, certain assumptions could be made about the project's current completion status, if necessary. For instance, if a 1998 aerial photograph showed that three out of four ramps of an interchange project were already completed and the fourth was partially completed, it could probably be assumed that the project has since been completed in its entirety and is currently operating.

When aerial photography for a given project area was not available to show whether or not the construction of the preferred alternative had been completed, or was in fact available but did not reflect any construction completion, aerial mapping was used if available. The problem with mapping, however, is that the actual use of an improved facility as shown on the mapping can not necessarily be verified. For instance, in the case of a roadway widening from four lanes to six, mapping available through the internet and CD ROM software would not necessarily identify that six lanes exist and are functioning. Instead, the roadway would show up on the mapping as a major facility, but without any indication of the number of lanes or status of widening completion. In other cases, however, aerial mapping was sufficient for making a conclusion about project completion, particularly when entirely new facilities were proposed.

Appendix G provides examples of aerial photography and mapping available from the internet and CD ROM software that were used to determine whether or not projects were actually constructed. The first example included in the appendix presents a project that was determined not to have been constructed and operating, while the second example presents a project that was determined to have been constructed and is currently operating.

Another indicator of a project's completion status is the FMIS database. Since one of the data fields in the FMIS reflects the year of construction completion, it can be assumed that the project has been completed if all of the records associated with a given EIS project indicate that construction has been completed. However, if some records that have been determined to be part of the overall EIS project show blanks in that column, then verification of construction completion becomes problematic. When possible, this indicator was used in conjunction with whatever aerial photography and mapping information was available.

In cases where project completion could not be verified via analysis of either aerial photography, available mapping or FMIS records data, state transportation agencies were contacted for project completion information. Verification in this manner, although likely to be fairly accurate, often required random ad hoc contacts of several individuals before finally reaching one who was actually able to provide the information requested. The level of institutional memory available within a state DOT was a key factor in the ability to collect valuable input from their staffs. Where local knowledge of a project was not readily available, or in cases where a project was constructed in an area not familiar to local DOT staff, verification of construction completion could not be verified in a timely manner. Record keeping issues and the lack of availability of EIS documents at the state DOT level were found to be particularly problematic for older projects. This procedure of contacting state DOTs was only used as a last resort due to the time involved in locating a knowledgeable individual, and so as not to burden the state DOT staff with specific questions about a project that may be a distant memory.

Other difficulties in establishing construction completion were also encountered during the course of the data collection. For instance, some of the older EISs had very little graphic information with which to establish the construction details needed to determine if the project was indeed built. In cases such as the Hammond Railroad Relocation in Indiana, the FEIS document had been reproduced at a quality too low to capture the rail lines. Project completion was impossible to determine even with the existence of some FMIS records related to the project.

3. 5. 4 Connecting the FMIS Type 1 Records to Type 2 Records

As described in Section 3. 5. 2 above, FMIS Type 1 records have unique Appropriation and Project Numbers for each project documented by FHWA. Once a Type 1 record has been linked to the project detailed in an EIS, Type 2 records are obtained from FHWA's Finance Division using the pertinent Appropriation and Project Numbers. This process requires FHWA staff to program the database to extract the pertinent records and provide them in ASCII format. They are then parsed for evaluation so that the necessary data can be gathered.

As discussed in Section 3. 2. 2, the Type 2 records provide more detailed information about each expenditure that is broken into categories such as preliminary engineering, landscaping, right-of-way acquisition, paving, safety, etc. For every Type 1 record, there may be several or even many Type 2 records having the same combination of Appropriation and Project Number. Matching the Type 1 and Type 2 records proved to be a tedious and often convoluted process.

An example of a typical set of Type 1 and Type 2 records is presented in Figure 1. As seen in the figure, the Type 1 records acquired from FHWA's database are comprised of 16 data fields. Two separate Type 1 records are shown, each having the same Appropriation Number but different Project Numbers. Upon requesting Type 2 records on the basis of those two Appropriation Number and Project Number combinations, a total of six separate Type 2 records appear, four exhibiting the first number combination and two exhibiting the second combination.

FIGURE 1: FMIS Type 1 and FMIS Type 2 Records

FMIS Type 1

REC-TYPE	STATE	APPROP	PROJECT	OK	LOCATION	WORK	NEPA	EISYR	EISNO	CON-TROL	V151	MSA	COUNTY	APPYR	LAST EXP
1	24.0	075	9057003	X	MD 16.W.OF PARSONS CRK. BRDG.TO SLAUGHTER CREE	GRADE, DRAIN & PAVE	#NULL!	.00	.0	1	.00		DORCHESTER	72.0	.0
1	24.0	075	9057004	X	MD 16.PARSONS CREEK TO SLAUGHTER CREEK BR.	RIGHT OF WAY ACQUISITION	#NULL!	.00	.0	1	.00		DORCHESTER	71.0	.0

FMIS Type 2

TYPE	STATE2	APP2	PROJECT2	LINE_ NO	CTY_ CODE	URB_ RURL	FAC_ TYPE	FUNC_ CLA	WORK_ CLA	WORK- TYPE	EXPEND1	STEP	MILES	TOT_COST	FED_FUND
2	24.0	075	9057003	30.0	19.00	R			3	I000	#NULL!	9	197.00	6574050.00	3298860.00
2	24.0	075	9057003	31.0	19.00	R			3	Y060	#NULL!	9	.00	152390.00	57260.00
2	24.0	075	9057003	32.0	19.00	R			3	Y080	#NULL!	9	.00	5010.00	2510.00
2	24.0	075	9057003	.0	#NULL!						3358630.00		.00	.00	.00
2	24.0	075	9057004	20.0	19.00	R			2	ROWA	#NULL!	9	.00	339570.00	88690.00
2	24.0	075	9057004	.0	#NULL!						88690.00		.00	.00	.00
														7071020	3447320

3. 5. 5 Final Review of EIS Project Details and Completion of Project Data Forms

Upon making a determination that a particular project has been constructed and is currently operating, critical information about the project and its environmental conditions was collected from the associated set of EIS documents and recorded on a project data form. Specific types of information that were collected and recorded on the basis of the EIS review generally included the following:

• Project name and location;
• Project setting type;
• Project milestone dates;
• Type of improvement;
• Project size and length;
• Estimated total project cost;
• Details of NEPA process;
• Major objectives of project;
• Types of permits required;
• Types of environmental issues or controversies;
• EIS document details;
• Number of project alternatives; and
• Details of any environmental mitigation measures proposed.

In addition to the information collected from the EIS documents, pertinent information from the FMIS records, when available, were also recorded on the project data form for each project. The FMIS information recorded in the data form was based on all compiled FMIS data records related to the specific EIS project. Types of information recorded from the FMIS data records generally included the following:

• Earliest year of agreement for a variety of project activities;
• Latest year completed for a variety of project activities;
• Total cost for a variety of project activities; and
• Federal cost for a variety of project activities.

The format and contents of the project data form template used to record the project-specific information, which actually consists of eight separate sheets, are shown in Appendix H. This template was developed at the outset of the research. It should be noted, however, that the actual information collected and recorded on the data form varied by project, especially as the research proceeded and the potential for obtaining certain information from each EIS or set of FMIS records became clearer. Some of the information originally proposed for collection in the project data form was determined to be too difficult to obtain or too subjective for undertaking statistical analysis and was therefore not collected beyond the initial set of documents reviewed during the preliminary research. It was also found that many of the data categories included from the FMIS records were rarely completed and were therefore eliminated from future data collection efforts as well.

3. 5. 6 Development and Verification of Database

The information from the project data forms was entered into a Microsoft Access database format that was suitable for statistical handling and analysis. A total of 197 separate data fields of information, reflecting the types of information included on the project data forms, were initially created. However, as discussed in Section 3. 5. 5 above, many of the data fields were dropped when it was determined that they added little value to the analysis. An abbreviated database consisting of 78 data fields was then developed, from which a final database of 74 variables was created. A list of the variable names assigned to each of the abbreviated 78 data fields used and their respective definitions is provided in Table 1. The actual final database that was created for each of the 100 projects included in the research sample grouped by the final 74 variables is presented in Appendix I.

Whenever a project was missing data for a given variable, the entry was generally coded with a "999" so that it would be excluded from any statistical analysis. For many of the original variables included in the database, the vast majority of the entries were coded with "999" and therefore, these variables were generally eliminated from further analysis. The variables that made it into the final database usually had substantially fewer "999" entries.

It should be noted that all of the cost variables included in the final database were standardized to 1999 dollars using the Consumer Price Index (CPI). The basic formula used in this regard is shown below:

(Cost x CPI 1999)= 1999 Cost
(CPI from Cost Year)

It was important to standardize these values since the costs as included in the FMIS records reflect 30 separate years of data. By standardizing the data, inflationary increases that have occurred during those years can be taken into account. The costs included in the database as shown in Appendix I reflect the standardized values.

Once all of the data were entered into the final database, a visual quality control check was performed on the data to identify any anomalies such as milestone dates that didn't make sense in relation to other milestone dates. When necessary, the original data sources were consulted to verify that the data entered were correct. If a transcription error was identified, the entered data was corrected. If the data anomalies could not be explained or corrected, the data field entry was changed to "999" in order to delete the data from further analysis. Upon completion of the visual quality control check, the database was ready for undertaking statistical analysis.

Table 1: Final Database Variables and Their Definitions

3. 6 Statistical Analysis Methodology

The statistical analysis methodology consists of several basic steps, as presented in summary in Section 3. 1. The details of each basic step are provided in the following subsections.

3. 6. 1 Identification of Descriptive Statistics

Descriptive parameters of the database variables were developed, first as a means to identify the general characteristics of each variable and then as a method of identifying any further anomalies in the data that were not caught during the visual quality control check as discussed in Section 3. 5. 6. above. In this regard, the final database was entered into the SPSS Base 10 statistical software package, which allows a variety of descriptive statistical analyses to be performed.

The database was initially examined to determine if any values fell outside of a range of values that was considered to be appropriate for a given variable. Any values that suggested a potential error in the data or a suspicious outlier were identified and flagged for further scrutiny. These values were then examined in relation to the mean, standard deviation, skewness, range, and maximum and minimum values of continuous data variables. Histograms, stem-and-leaf, and box-and-whisker plots were created to provide a visual view of the data (examples of these types of graphic representations are provided in Appendix J). Probability plots, used to evaluate the degree to which continuous variables vary from the normal distribution, were created for most continuous variables. In the case of nominal variables, a frequency analysis usually sufficed to indicate any anomalies.

Next, a number of computed variables were created from the variables included in the final database. One of these variables, denoted as PROSTART, was created to reflect the earlier date of the two variables, BEG_NEPA and AUTHPENG. This variable was necessary since some projects indicated that the NEPA process began even before Preliminary Engineering officially began.

Most of the computed variables were created by subtracting one variable from another and were established to create timelines. The specific variables that were created in this manner, using some of the variables shown in Table 1, are as follows:

In the above cases where one variable was subtracted from another, the results were checked to ascertain if an inappropriate value resulted. For example, if the year in which a process began was subtracted from the year in which it ended and a negative result occurred, it indicates that the end date was earlier than the begin date. An effort was then made to determine if there was a data entry error or if the source data was incorrect. If the issue could not be resolved, the case was eliminated for analyses involving either the computed variable, the begin date or the end date.

In addition to the new variables that were created to reflect timelines, one computed variable was created by adding one variable to another. This new variable, which was also created by using some of the variables shown in Table 1, is as follows:

Another variable, denoted LEISDEC, was created to reflect the decade in which the EIS was completed (i.e., LASTEIS_YR). All cases in which the EIS was completed in the 1970s were coded as "1970s" under the LEISDEC variable. Similar codings were performed for those EISs completed in the 1980s and 1990s.

Finally, a number of variables were created by transforming other existing variables into different numeric equivalents such as square roots or logarithms. The following new variables were created in this manner:

It should be noted that the addition of the created variables to the database increased the total number of variables to 114, although not all of those variables were necessarily used in the statistics analyses.

The computed variables and the original variables were then run through the program's Exploratory Data module, which repeated some of the descriptives, added others (e. g. , medians, m-estimator measures, etc.) and consolidated much of the information into a single set of tables. An additional advantage of the Exploratory Data module is that it permits the compilation of data by categories, so that exploratory data could be generated by former FHWA Region and allow the results to be compared.

In developing the descriptive statistics, a number of assumptions were made. These include the following:

• All data were drawn from a normal population;
• All cases are independent of each other;
• Significance was set at 0. 05 and, in cases where the statistical package calculated the precise significance, it is provided; and
• Where the length of time for a process (e. g. , Preliminary Engineering or the NEPA process) was accomplished within a single year, the process was shown as taking one year.

3. 6. 2 Selection of Data Analysis Methods

Descriptive and analytical methods were limited to procedures available in the SPSS Base 10 package. The specific methods used from that package are described below.

Identification of Data Distribution

Many statistical tests require that the data being analyzed be normally distributed. Although the Central Limits Theorem permits the normalcy requirement to be relaxed for any sample where n$100, an effort was made to either transform non-normally distributed data to a form approximating a normal distribution, or to test the data using procedures that do not require a normal distribution.

Continuous data were examined for normalcy using both graphic and non-graphic methods. First, the variable was subjected to a Kolmogorov-Smirnov test against a theoretical normal distribution. If the significance level was greater than 0. 05, the variable was assumed to not differ significantly from the normal distribution. If the significance level was #0. 05, the variable was treated as being not normally distributed.

Data Transformations

Data identified as being non-normally distributed were then visually examined using one or more histograms and probability plots to identify the most likely transformation method. If a given variable could be appropriately transformed using the techniques available in the statistical package, this was done. In some cases, the square root of the original variable resulted in a normal distribution. In other cases, the log_n, log_10, reciprocal or exponential transformations were tested. Transformation resulted in a new variable, but the original variable was retained. The key variables created through such transformation are identified above in Section 3. 6. 1.

Parametric Versus Non-Parametric Analyses

Some variables in the final database could not be easily transformed. Others were clearly not continuous data and, therefore, were not normally distributed. For these variables, non-parametric test procedures were utilized. Unlike parametric procedures, non-parametric procedures do not require a normal distribution. Non-parametric tests are generally considered to have less power than parametric tests (i.e. , there is a somewhat higher risk of a Type II error occurring, which involves the failure to reject a hypothesis when it is actually false). Since the norm in significance testing is to test a null hypothesis (e. g. , there is no difference in the length of the NEPA process between Regions 1 and 5), the lower power of the non-parametric tests could result in failing to reject that hypothesis. Therefore, wherever possible, parametric tests were used. In some cases, the transformation was tested parametrically and the original variable was tested non-parametrically to compare the results.

3. 6. 3 Statistical Relationship Tests Between Variables

A number of statistical relationship tests were designed and conducted in order to answer a number of questions, all of which relate to the length of the NEPA process. Specifically, these tests were designed to identify which variables, if any, are correlated with the length of the NEPA process, and which factors, if any, appear to differ with respect to the length of the NEPA process. A list of the correlations that were tested in this regard are shown in Table 2.

The general approach to identifying any statistical relationships began by first looking at the descriptive statistics as a whole, and then by looking at the descriptive statistics broken down by categories suggested by the data variables themselves. For example, a total of 97 out of the 100 cases of NEPATIME information were considered to be useable for analysis, after those cases that were considered to be extreme outliers were eliminated. These 97 cases could then be grouped into categories (e. g. , by FHWA region) and the descriptive statistics could be obtained by category (FHWA region) as well. The same process could then be followed for dividing the 97 cases into, for instance, those associated with a Section 4(f) study and those not associated with a Section 4(f) study. Similar analyses could be performed for a variety of other types of categories as well.

The next step was to attempt to find significant differences in the length of the NEPA process according to the categorical variables. Finally, any correlations between the length of the NEPA process and other more or less continuous variables were sought. A preliminary correlation matrix was run, relating the length of the NEPA process to all other continuous variables. Those variables that appeared to have a linear relationship or close to it, were selected for further analysis. Further analysis involved either transforming those variables which could be made to approximate the normal distribution and then correlating them to the length of the NEPA process, or ascertaining the correlation through use of nonparametric procedures.

Where the variables against which the length of the NEPA process were to be tested were categorical, relationships were examined using ANOVA (Analysis of Variance) or an Independent Sample t-test. The ANOVA was used where there were multiple categories to be tested. The t-test, on the other hand, was used where the categories were dichotomous. In one case (i.e. , the variable NEWLANES), the variable was considered to be an ordinal variable, requiring that a correlation with NEPATIME be obtained by using Spearman's Rho, a non-parametric correlation procedure.

Table 2: Statistical Correlation Tests Conducted for the Length of the NEPA Process and a Variety of Other Variables

• (LASTEIS_YR — BEG_NEPA) and FHWAREG (1, 3 and 4 combined)
• (LASTEIS_YR — BEG_NEPA) and FHWAREG (1, 3 and 4 separately)
• (LASTEIS_YR — BEG_NEPA) and LAND_USE
• (LASTEIS_YR — BEG_NEPA) and MSA
• (LASTEIS_YR — BEG_NEPA) and GROWTH
• (LASTEIS_YR — BEG_NEPA) and PROJTYP1
• (LASTEIS_YR — BEG_NEPA) and NEWLANES
• (LASTEIS_YR — BEG_NEPA) and LENGTH
• (LASTEIS_YR — BEG_NEPA) and LASTEIS_YR ('70s, '80s and '90s separately)
• (LASTEIS_YR — BEG_NEPA) and PH_NO
• (LASTEIS_YR — BEG_NEPA) and (FED_AGEN + ST_AGEN)
• (LASTEIS_YR — BEG_NEPA) and EPA_RATE
• (LASTEIS_YR — BEG_NEPA) and SECT404
• (LASTEIS_YR — BEG_NEPA) and PERMCG
• (LASTEIS_YR — BEG_NEPA) and SECT4F
• (LASTEIS_YR — BEG_NEPA) and SECT106
• (LASTEIS_YR — BEG_NEPA) and (COMPPENG — AUTHPENG)
• (LASTEIS_YR — BEG_NEPA) and (COMPCENG — AUTHCENG)
• (LASTEIS_YR — BEG_NEPA) and (COMPROWA — AUTHROWA)
• (LASTEIS_YR — BEG_NEPA) and (COMPCONS — AUTHCONS)
• (LASTEIS_YR — BEG_NEPA) and (COMPCONS — AUTHPENG)
• (LASTEIS_YR — BEG_NEPA) and (COMPCONS — INCEPTION)
• (LASTEIS_YR — BEG_NEPA) and (YROPEN — AUTHPENG)
• (LASTEIS_YR — BEG_NEPA) and (YROPEN — INCEPTION)
• (LASTEIS_YR — BEG_NEPA) and PUBLICM
• (LASTEIS_YR — BEG_NEPA) and AGNCYMTG
• (LASTEIS_YR — BEG_NEPA) and ISSULAND
• (LASTEIS_YR — BEG_NEPA) and ISSUFARM
• (LASTEIS_YR — BEG_NEPA) and ISSUECON
• (LASTEIS_YR — BEG_NEPA) and ISSUNOIS
• (LASTEIS_YR — BEG_NEPA) and ISSUAQ
• (LASTEIS_YR — BEG_NEPA) and ISSUVIS
• (LASTEIS_YR — BEG_NEPA) and ISSUWETL
• (LASTEIS_YR — BEG_NEPA) and ISSUT&E
• (LASTEIS_YR — BEG_NEPA) and ISSUCULT
• (LASTEIS_YR — BEG_NEPA) and ISSUWQ
• (LASTEIS_YR — BEG_NEPA) and ISSUINDR
• (LASTEIS_YR — BEG_NEPA) and ISSUEJ
• (LASTEIS_YR — BEG_NEPA) and ISSUCOM
• (LASTEIS_YR — BEG_NEPA) and ISSUOTHR
• (LASTEIS_YR — BEG_NEPA) and PROJALT
• (LASTEIS_YR — BEG_NEPA) and CHGFEIS
• (LASTEIS_YR — BEG_NEPA) and RELHH
• (LASTEIS_YR — BEG_NEPA) and RELBUS
• (LASTEIS_YR — BEG_NEPA) and ACWETLI
• (LASTEIS_YR — BEG_NEPA) and EIS_COST (standardized to an index year from LASTEIS_YR)
• (LASTEIS_YR — BEG_NEPA) and FMISTOTL (standardized to an index year from Median year between AUTHPENG and COMPCONS)
• (LASTEIS_YR — BEG_NEPA) and TOT$PENG (standardized to an index year from Median year between AUTHPENG and COMPPENG)
• (LASTEIS_YR — BEG_NEPA) and TOT$CENG (standardized to an index year from Median year between AUTHCENG and COMPCENG)
• (LASTEIS_YR — BEG_NEPA) and TOT$ROWA (standardized to an index year from Median year between AUTHROWA and COMPROWA)
• (LASTEIS_YR — BEG_NEPA) and TOT$CONS (standardized to an index year from Median year between AUTHCONS and COMPCONS)

Notes: In some of the above cases, there was either insufficient information available for a given variable or insufficient variety of data within the variable to actually allow correlation tests to be performed. Therefore, the above list presents the full listing of correlation tests that were considered, rather than those that were actually performed.

The variable LASTEIS_YR — BEG_NEPA is also denoted as NEPATIME.

Source: The Louis Berger Group, Inc. , 2000.