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

Previous | Index | Next

This section presents the major results of the statistical analysis of this research. The first set of findings relate to descriptive statistics on the length of the NEPA process, which is really the heart of this entire analysis. These descriptive statistics are presented, first, for the length of the NEPA process for the country as a whole during the past 30 years, and then for the individual former FHWA regions. Descriptive statistics for all other factors considered during the course of this analysis are then presented. Finally, the relationship between length of the NEPA process and other factors considered are presented, with its relationship to the project development process given special attention. The relationship between length of the NEPA process and all other factors are also presented according to the individual former FHWA regions.

It should be noted that this section generally focuses only on those findings that were found to be statistically significant (i. e. , within 0. 05 level of significance). If findings are not specifically reported in this section, it is usually because they were not found to be statistically significant. Even so, as reported below, some very interesting results were found.

4. 1 Descriptive Statistics on the Length of the NEPA Process

4. 1. 1 Overall Descriptive Statistics

As discussed in the previous section, the length of the NEPA process was denoted in this analysis as the variable NEPATIME, and was calculated on the basis of LASTEIS_YR minus BEG_NEPA. In a few cases, NEPATIME was found to exceed 15 years, generally because a Supplemental EIS process was involved in addition to the initial EIS process. Since the method for calculating NEPATIME didn't specifically exclude the time between the initial EIS process and the Supplemental EIS process, NEPATIME didn't give a true estimate of actual NEPA process time in those few cases. Therefore, those cases where NEPATIME exceeded 15 years were excluded from the descriptive analysis, in order to avoid incorrectly skewing the results.

A descriptive analysis of NEPATIME was run. The results, which are presented both tabularly and graphically in Appendix J, indicate that the mean value of NEPATIME is approximately 3. 6 years, with a standard deviation of 2. 4 years and values ranging from 1 to 12 years. There were three cases with NEPATIME values exceeding 15 years, leaving a valid sample size of 97 cases for analysis. Theoretically, this information indicates that the average period of time that has been required to comply with the NEPA process during the past 30 years has been about 3. 6 years, and the majority of the EISs prepared during that period have ranged between 1. 2 years and 6. 0 years (based on the standard deviation).

However, because the NEPATIME variable was found to be not normally distributed, the mean value may not be a true indicator of central tendency, so other indicators of central tendency were obtained. For instance, the median value, which is the midpoint among the full set of NEPATIME values that divides the observations into two groups of equal numbers, was determined to be about 3. 0 years. The reason that the median value is somewhat less than the mean value is because the higher values of NEPATIME that were included in the analysis (i. e. , those approaching the highest value of 12 years) tend to skew the mean on the high side. Therefore, in this case, the median value is likely to be a better indicator of the typical length required for complying with the NEPA process during the past 30 years.

A Kolmogorov-Smirnov test comparing the distribution of NEPATIME with a theoretical normal distribution was also run. As a result of this test, NEPATIME was found to be significantly different from the normal distribution. For subsequent tests, a new variable, which was not significantly different from the normal distribution, was created by taking the square root of NEPATIME (SQRTNTIM). For many of the subsequent statistical analyses, SQRTNTIM was utilized instead of NEPATIME due to the more normal distribution of that variable, although equivalent NEPATIME values are also reported when appropriate.

4. 1. 2 Descriptive Statistics by Region

The next step involved exploratory analysis of NEPATIME categorized according to the nine former FHWA regions. The results are summarized in Table 3, and are presented in greater detail in Appendix K. The interesting fact that emerged from this analysis is that there are apparent large differences between some FHWA regions with respect to the mean and median NEPATIMEs. The largest difference was found to be between Region 1 (in the Northeast) and Region 10 (in the Northwest). In the case of Region 1, which demonstrated the longest NEPATIME values among the regions, the mean NEPATIME value was approximately 4. 7 years, while the median NEPATIME value was approximately 4. 5 years. ^[6] At the other end of the spectrum, Region 10's mean NEPATIME value was approximately 2. 2 years, although the median value was only 1. 0 year. Interestingly, the minimum value for Region 10 was also 1. 0 year. The fact that the median and minimum NEPATIME values were both 1. 0 year is reflective of the fact that 6 out of the 11 cases in Region 10 had a NEPATIME value of 1. 0.

Regions 4 (in the Southeast) and 7 (in the Midwest), which had the next highest NEPATIME mean values, also appeared to be very different from Region 10. In the case of Region 4, the mean and median values were approximately 4. 4 and 3. 0 years, respectively. In the case of Region 7, the mean and median values were approximately 4. 3 and 4. 0 years, respectively.

After Region 10, Region 8 (the Rocky Mountain states) exhibited the next lowest mean NEPATIME value of approximately 2. 6 years. The median value was 2. 0 years.

Table 3: DESCRIPTIVE STATISTICS OF NEPATIME BY REGION

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

It should be noted that the regional comparisons provided above are based entirely on a review of the descriptive statistics provided in Appendix K. Actual statistical testing of the difference in the mean values of NEPATIME between regions is discussed in Section 4. 3. 3 below.

4. 2 Descriptive Statistics on Other Factors Considered

Similar to the NEPATIME variable, descriptive statistics were also analyzed for the other factors against which correlations would be made with NEPATIME or its more normally distributed square-root value, SQRTNTIM. Highlights of those descriptive statistics, which include number of valid cases in the sample, minimum value, maximum value, mean, and standard deviation, are presented in Table 4. Of particular interest is the total length of the project development process (OPMINSTR, or YR_OPEN minus PROSTART) during the 30-year NEPA period. Based on the sample, it can be concluded that the mean time for the entire project development process is approximately 13. 1 years, although the process has ranged from a little as 3. 0 years to as many as 36 years. The standard deviation of approximately 6. 7 years indicates that the vast majority of the cases ranged between 6. 4 years and 19. 8 years.

Frequencies for some of the key variables that could be considered to be either continuous or categorical in nature are presented in Appendix L. These frequencies indicate the number of responses for each categorical response that were received within the sample. For instance, in the case of the LAND_USE variable, 45 out of the 97 valid responses were identified as being rural in character, while 22 were identified as being suburban and 30 were identified as urban. Similar frequencies are presented in the Appendix for 26 additional variables as well.

These frequencies were then used to identify the appropriate types of further analyses to be performed (i. e. , whether to treat them as continuous or categorical variables in the subsequent testing).

4. 3 Statistical Relationship of the Length of the NEPA Process with Other Factors

Having developed the descriptive statistics for the length of the NEPA process and other factors, the next step was to refine the search for correlations and relationships between NEPATIME (or its surrogate) and other variables in the data set. The surrogate variable for NEPATIME that was found to be the most effective in making correlations was SQRTNTIM, which represents the square root of NEPATIME. The reason that SQRTNTIM was generally more effective in making correlations was because it more closely approximates a normal distribution than NEPATIME. Even so,NEPATIME equivalents are generally provided as well, since these values are more readily understood than the square root values.

Table 4: Descriptive Statistics for Key Variables

	N	Minimum	Maximum	Mean	Std. Deviation
INCEPTION	68	46. 00	88. 00	68. 04	9. 13
NEWLANES	95	0. 00	6. 00	2. 97	1. 63
LENGTH	91	0. 36	25. 90	6. 52	5. 54
LAND_AQ	61	0. 00	861. 00	153. 08	211. 77
FMISTOTL	97	$461,000. 00	$196,138,563. 00	$23,474,301. 51	$36,113,545. 82
FMIS_FED	97	$282,292. 00	$170,917,977. 00	$18,681,508. 42	$30,197,536. 57
NEPATIME	97	1. 00	12. 00	3. 61	2. 43
SQRTNTIM	97	1. 00	3. 46	1. 79	0. 63
FED_AGEN	97	1. 00	14. 00	6. 51	2. 97
ST_AGEN	97	0. 00	15. 00	4. 76	2. 84
EISPP	97	0. 00	1160. 00	265. 59	246. 74
PUBLICM	87	0. 00	11. 00	2. 51	2. 20
AGNCYMTG	34	0. 00	130. 00	8. 53	23. 38
PROJALT	97	0. 00	36. 00	3. 66	3. 97
RELHH	95	0. 00	981. 00	29. 60	106. 73
RELBUS	94	0. 00	106. 00	5. 78	12. 76
ACWETLI	97	0. 00	85. 00	2. 62	10. 36
OPMINSTR	97	3. 00	36. 00	13. 10	6. 66
PENG99	61	$2,056. 00	$31,811,256. 00	$2,648,349. 77	$4,813,666. 35
CENG99	62	$1,738. 00	$32,504,258. 00	$3,220,449. 29	$5,269,030. 55
ROW99	58	$8,637. 00	$115,309,521. 00	$7,753,319. 71	$16,687,161. 70
CON99	79	$180,139. 00	$188,493,520. 00	$26,966,478. 32	$38,332,523. 70
INCEP2OP	68	6. 00	50. 00	21. 94	9. 95
ALLAGENC	97	1. 00	25. 00	11. 27	4. 87
PENGTIME	61	1. 00	28. 00	8. 56	7. 35
CONTIME	78	1. 00	18. 00	5. 53	4. 11
CENGTIME	62	1. 00	16. 00	5. 85	4. 12
ROWTIME	57	1. 00	26. 00	6. 63	5. 36
PENG2CON	49	1. 00	36. 00	10. 35	7. 97
INCP2CON	55	5. 00	50. 00	21. 42	10. 15
Valid N (listwise)	6
Source: The Louis Berger Group, Inc. , 2000.

4. 3. 1 Relationship of the Length of the NEPA Process to the Entire Project Development Process

The first statistical relationship established with the length of the NEPA process (NEPATIME) was a comparison to total project length (YR_OPEN minus PROSTART). Based on mean and median values, it was found that the NEPA process comprises or, more likely, coincides with approximately 28% or 27%, respectively, of the total project development time. In other words, between the time that either preliminary engineering or the NEPA process begins, whichever is first, and the end of construction, the entire NEPA process generally comprises only about 27 - 28% of that total period.

A correlation analysis using a Pearson (parametric) test was then run in order to identify any significant relationship between the length of the NEPA process and the length of the project development process. The result of that test yielded a correlation value of 0. 495 and a significance level of 0. 000. Of all variables tested against NEPATIME, the correlation with YR_OPEN minus PROSTART is the greatest. This finding is not surprising since the length of the NEPA process accounts for 27 - 28% of the total project development process and would, therefore, likely have some influence on the total length of that process. Even so, the relatively weak correlation value indicates that NEPATIME only accounts for only a relatively small portion of the total variation in length of the project development process, further indicating that other factors would also likely influence the total process length.

4. 3. 2 Relationship of the Length of the NEPA Process to Other Factors

For continuous data variables, Pearson (parametric) and Spearman (nonparametric) correlation tests were run against NEPATIME and SQRTNTIM (i. e. , the square root of NEPATIME). Weak but significant correlations between SQRTNTIM / NEPATIME were found with the variables shown in Table 5. Of the variables included in the table, the strongest correlations were found with: the length of the project development process from project inception to project opening (INCP2CON, 0. 44); the length of the project development process from authorization of preliminary engineering to completion of construction (PENG2CON, 0. 43); and the number of total agencies involved (ALLAGENC, 0. 43). All of these correlations were weaker than that found between SQRTNTIM / NEPATIME and YR_OPEN - PROSTART.

It should be noted that the results of all correlation tests between SQRTNTIM / NEPATIME and other variables, as well as between the other variables themselves, are presented in Appendix M.

LEISDEC is a derived variable that is based on dividing the LASTEIS_YR into decades. Using the decades as a categorical variable, an ANOVA showed that there is a difference in the length of the NEPA process among the categories representing the 1970s, 1980s, and 1990s. Post hoc testing indicated that the significant difference occurred between the 1970s and the remaining two decades, whereas the 1980s and 1990s did not differ significantly from each other.

Table 5: CORRELATIONS BETWEEN LENGTH OF THE NEPA PROCESS AND OTHER VARIABLES

(¹) Significance level 0. 05 or better. (²) P= parametric; NP= nonparametric.

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

Appendix N for results of the post hoc testing by decade). It was found that EISs completed in the 1970s required significantly less time to prepare than those completed in either the 1980s or the 1990s. This is not an unexpected result since the requirements for preparing EISs evolved and became more complex over time.

Specifically, it was found that EISs completed in the 1970s took, on average, approximately 2. 2 years to prepare. By the 1980s, the average mean for EIS preparation increased to approximately 4. 4 years while the average mean increased to approximately 5. 0 years in the 1990s. Although an increase is shown to exist between the 1980s and the 1990s, the difference between those two decades was not found to be statistically significant (i.e. , the difference exceeds the 0. 05 level of significance).

Significant relationships were also found between SQRTNTIM and two variables reflecting other specific regulatory requirements. SECT404 is a nominal variable indicating whether a given EIS project also involved a Section 404 permit from the U. S. Army Corps of Engineers. As shown in Table 6, those cases that did involve Section 404 (denoted as "True" in the table) showed a significantly longer NEPA process time. The NEPATIME equivalent to SQRTNTM was found to be about 4. 3 years when a Section 404 permit was involved, but only about 2. 4 years when no Section 404 permit was involved. This finding is not a surprise given the increased compliance associated with the additional procedural and technical requirements of Section 404 that would likely have been initiated during the EIS process.

Table 6: CORRELATIONS BETWEEN LENGTH OF NEPA PROCESS AND OTHER REGULATORY REQUIREMENTS

SECT404	Mean SQRTNTIM	NEPATIME Equivalent	Significance
True	2. 073	4. 3	0. 000
False	1. 5439	2. 4
SECT4F
True	2. 1721	4. 7	0. 000
False	1. 6707	2. 8

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

SECT4F is a nominal variable indicating whether a given EIS project involved approval of a Section 4(f) Statement by FHWA. Similar to the Section 404 finding, those cases that did involve Section 4(f) approval (denoted as "True" in Table 6) showed a significantly longer NEPA process time. Specifically, the NEPATIME equivalent to SQRTNTIM was found to be about 4. 7 years when a Section 4(f) approval was involved, but only about 2. 8 years when no Section 4(f) approval was involved. This finding makes sense for the same reason that the Section 404 finding makes sense.

Other regulatory variables such as Coast Guard permits (PERMCG) and Section 106 approvals (SECT106) did not show significant differences in the length of the NEPA process between the with and without scenarios. Therefore, these variables are not shown in Table 6. However, the results of the tests performed for assessing correlation between the length of the NEPA process (SQRTNTIM) and all of the regulatory variables are provided in Appendix O.

AGNCYMTG is the number of meetings held with State and Federal agencies during the NEPA process. The huge range of values found for this variable (i. e. , 0 - 130 meetings) suggested that the variable should be disaggregated into two separate categories: (1) those below and including the median value of 2; and (2) those above the median. The resulting test showed a significant difference between the two groups with respect to the mean length of the NEPA process. As shown in Table 7, those projects where two or less agency meetings were held during the course of EIS preparation took approximately 2. 4 years to complete. When more than two agency meetings were held, it took approximately 4. 5 years to complete preparation of the EIS.

Table 7: CORRELATIONS BETWEEN LENGTH OF NEPA PROCESS AND NUMBER OF MEETINGS

AGNCMTG	Mean SQRTNTM	NEPATIME Equivalent	Significance
2 and below	1. 5506	2. 4	0. 008
more than 2	2. 1169	4. 5
PUBLICM
2 and below	1. 6313	2. 7	0. 002
more than 2	2. 0484	4. 2

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

A similar process was followed with the variable PUBLICM, which represents the number of public meetings held. The distribution of this variable also showed the median number of meetings to be 2. The 97 valid cases were then divided into those involving 2 or less meetings and those involving more than 2. Once again, there was a significant difference in the length of the NEPA process between these two groups, with an increased mean for NEPA process length associated with an increase in the number of public meetings. Specifically, as shown in Table 7, an average of 2. 7 years has been required to prepare an EIS when the two or less meetings have been involved while the average increases to 4. 2 years when more than two meetings have been involved.

Complete statistics for both variables, AGNCYMTG and PUBLICM, and their correlations with SQRTNTIM are presented in Appendix P.

It is important to note that neither the number of agency meetings nor the number of public meetings necessarily is a determinant of the length of time required to comply with NEPA. In fact, it is likely that the number of meetings of either type are actually determined by the complexity of the project and, therefore, are more of a byproduct of the length of the NEPA process rather than vice versa. The statistical analyses performed, however, do not specifically identify which variable is dependent on the other, so this theory is based more on speculation than fact.

A number of variables reflecting a variety of environmental issues were also considered for correlating against the length of the NEPA process. Most of these variables were found to be distributed so heavily in favor of the "False" response, that analysis would not have been appropriate or meaningful. Only one issue, ISSUNOIS (i. e. , which reflects noise as an issue), was somewhat more equitably distributed (30 True, 67 False responses). Therefore, this variable was tested against the length of the NEPA process. The result was that a significant difference in length of NEPA existed between those projects where noise was an issue and those projects where they were not. Projects involving noise issues had a longer mean NEPA process time than those without noise issues. Specifically, the mean NEPA process time was found to be only 3. 2 years when noise was not an issue while the mean increased to 4. 4 years when noise was an issue. Complete statistics for correlations between ISSUNOIS and length of the NEPA process (i. e. , SQRTNTIM / NEPATIME) are provided in Appendix Q.

Examination of other variables yielded no information of interest. No significant correlations or differences in NEPA process time were found with any other variable not specifically discussed above. It should be particularly noted in this regard that, other than the cost of right-of-way acquisition which was found to have a weak correlation with length of the NEPA process (see Table 5), no other cost variables were found to have any significant relationship with NEPA. It is very possible that this apparent lack of correlation may be attributed more to missing values and shortcomings in the FMIS database related to the cost variables (see Section 3. 3. 3 for further details on this data collection issue).

4. 3. 3 Relationship of the Length of the NEPA Process to Individual FHWA Regions

The next step involved an ANOVA, in which the transformed NEPATIME variable, SQRTNTIM, was tested against the entire group of former FHWA regions to see if any differences actually existed among the regions. As reported in Section 4. 1. 2, there were a number of apparent differences between some of the regions based on a review of the descriptive statistics. A determination as to whether or not these differences were statistically significant, however, was made following the use of ANOVA. This test works by examining the entire spectrum of values and categories. It looks at each category (i. e. , FHWA region) within the context of the full range of categories and tests if any category means are out of line with the others. If a significant difference is detected by the ANOVA, then Post Hoc testing, which analyzes where the differences exist between regions, would then also be conducted.

When applied to all regions, the results indicated that there was no significant difference among them (see Appendix R). Intuitively, however, the difference between Regions 1 and 10, which was the most extreme difference found in the descriptive statistics analysis, should have been significant. Following this reasoning, an Independent Samples t-test was run, using just the results from Region 1 and Region 10 (see Appendix S). In this scenario, a significant difference was revealed between the two regions in terms of SQRTNTIM. Similar results were obtained when the test was run using the raw variable (NEPATIME). These results confirm that, at a minimum, Regions 1 and 10 have historically been truly different from each other in terms of the length of the NEPA process. It is possible that some of the other individual regions may be significantly different from each other as well, but only Regions 1 and 10 were tested in this regard.

Since Regions 1 and 10 represented the greatest difference in mean values for SQRTNTIM, an attempt was also made to determine if any other variables exhibited differences between Regions 1 and 10. In this regard, Independent Samples t-tests were run in most cases while Crosstabulation analyses were run in two cases. Several variables were found to vary significantly between the two regions, with Region 1 exhibiting significantly greater values than Region 10 in all cases, which is consistent with the trend in NEPATIME. The following variables exhibited significant differences in this regard, as presented in Appendix T:

• NEWLANES (Number of new lanes added by construction of the project);
• ALLAGENC (Number of Federal and state agencies commenting on EIS);
• LNROWA99 (Transformed variable — ROW costs indexed to 1999 dollars);
• LNCON99 (Transformed variable — construction cost indexed to 1999 dollars);
• LNFMISFE (Transformed variable — FMIS Federal dollars indexed to 1999 dollars);
• LNCENG99 (Transformed variable — construction engineering costs indexed to 1999 dollars);
• SECT4F (EIS involving Section 4(f)); and
• PUBLICM (Number of public meetings).

4. 4 Use of Results as a Baseline for Evaluating Future Environmental Streamlining Initiatives

Although most of the relationships established through this analysis are between the length of the NEPA process and variables that would not be known until after the project has been largely completed, some useful relationships have been established which could be useful in evaluating the success of future environmental streamlining efforts. First, there appears to be a positive relationship between the length of the NEPA process and length of the total project development process. Second, the average length of time for completion of the NEPA process appears to comprise approximately 27 - 28% of the total project development time. This could be a useful benchmark for evaluating whether or not the NEPA process is efficiently carried forward in a given project. More importantly, shifts away from this average may indicate gradual loss or improvement in efficiency, or changes in the requirements that have changed the length of the process.

Other potentially useful relationships include the positive correlations between the number of state and local agencies that comment on an EIS and the length of time for the entire NEPA process. The same is true of agency and public meetings, the magnitude of right-of way acquisitions and the presence of special permitting or regulatory studies. This is not to suggest that any of these factors can or even should be limited. They do, however, signal the probability of more extensive time required to fully comply with the process. This should be of value to administrators.

It would be premature on the basis of the present data set and analysis to attempt to construct a model of the length of the NEPA process. At most, this analysis could serve as a baseline from which additional analysis can be started and to which additional data can be added. In other words, the data set lends itself well to be used as a baseline against which to evaluate the relative degree of success of future environmental streamlining initiatives, but its usefulness as a predictive model for determining either the length of the NEPA process based on certain project or environmental factors or for determining the length of the overall project development process based on the length of the NEPA process is limited at this point in time.