U.S. Department of Transportation Federal Railroad Administration

FLORIDA'S TRAIN WHISTLE BAN

---

Office of Safety Final Edition September 1995

---

TABLE OF CONTENTS

Section I. INTRODUCTION II. STUDY METHODOLOGY Approach Study Hypothesis Null Hypothesis III. DATA COLLECTION Selection of Crossings Studied FEC Crossing Ordinances Accident Data Selected IV. DATA ANALYSIS Categorizing Data Daytime Control Group No-Ordinance FEC Control Group CSX Transportation, Inc.(CSX) Control Group V. CONCLUSION VI. ACTIONS FOLLOWING COMPLETION OF STUDY

Page 1 3 3 4 4 5 5 6 6 7 7 8 9 10 11 11

Appendices A Florida Statutes Applicable to Railroads and Other Utilities B Code of Metropolitan Dade County, (Sections 21-27.3 and 21-27.4) C Normalizing Data D Detailed Statistical Analysis E Form FRA F 6180-57, Rail-Highway Grade Crossing Accident/Incident Report Form F Chronological List of Whistle Bans by Jurisdiction G Letter to Florida Legislator John F. Cosgrove H Letter to Roman A. Yoder, Jr., Assistant City Manager, City of Palm Bay, Florida I Population of Florida (1980-90) J Fatal Highway Accidents in Florida & Selected Counties (1980-89) K FEC Crossing Accidents Through 1991 at Impacted Crossings (10:00 p.m. to 6:00 a.m.) L Letter from Public Utility Commission of Oregon re: Oregon Train Whistle Bans M Order No. 91-1164 Rescinding Whistle Prohibition Order for City of Eugene, Oregon N FRA's Emergency Order No. 15 O Conference Notice No. 3 P Letter to U.S. Representative, the Honorable Ileana Ros-Lehtinen Q Conference Notice No. 7 R Amended Emergency Order No. 15 S FRA Criteria for Allowing Whistle Bans

Page 16 18 20 22 28 29 33 46 55 56 57 58 60 71 79 109 116 123 139

Figures 1 Map of FEC 2 Special Advance Warning Sign 3 FEC Crossing Accidents: Impacted Crossings (10:00 p.m. to 6:00 a.m.) 4 FEC Ordinance-Impacted Crossings

Page 2 2 5 6

Tables 1 FEC Impacted Crossings, Nighttime Accident Experience 2 FEC Impacted Crossings, Daytime Accident Experience 3 No-Ordinance (FEC) Crossings 4 CSX Transportation (No-Ordinance) Crossings

Page 8 9 9 10

---

FLORIDA'S TRAIN WHISTLE BAN

In 1990, the Federal Railroad Administration (FRA) studied train whistle bans in Florida.1 The study shows a strong correlation between nighttime whistle bans and the number of accidents at highway-rail crossings. This report summarizes and updates the second edition of the study.2 The report also incorporates FRA administrative decisions issued subsequent to earlier editions of Florida's Train Whistle Ban.

I. INTRODUCTION
Railroads are powerless to restrain the growth of residential populations along their rights-of-way. Train whistle use is an important deterrent to highway-rail crossing accidents in densely populated areas. However, special interest groups formed in the late 1970s, sought ways to silence train whistles. Their attention concentrated on nighttime bans, which gained much support from nearby residents. One Florida-based group, Project Whistle Stop, Inc., approached Federal agencies and the State of Florida's Federal legislators to sponsor a national whistle ban.

When these efforts failed, the Florida State Legislature was persuaded to enact state whistle ban legislation. Effective July 1, 1984, local jurisdictions were allowed to establish nighttime (10:00 p.m. to 6:00 a.m.) train whistle bans. The bans only apply to certain crossings on the Florida East Coast Railway Company (FEC). Eligible crossings must be equipped with active warning devices.3 Figure 1 shows a map of the FEC system. Figure 2 shows the special advance warning sign attached below the traditional "advance warning" sign. Appendix A is an excerpt of the authorizing law, Florida Statute No. 351.03.

The Dade County Board of Commissioners passed the first widespread ordinance meeting State requirements. This ordinance affects 107 crossings. After numerous legal challenges, the Dade County FEC whistle ban began July 29, 1984. Appendix B is a draft of the Dade County Whistle Ban Ordinance.

Other jurisdictions followed. Seven counties and a dozen additional cities established bans. By December 31, 1984, 511 FEC public grade crossings were effected by the bans.4 Florida's east coast is the only locale in the nation where a ban on railroad whistles became extensive. Local bans have been established in other states. However, safety concerns generally prevail over noise concerns.

FEC's nighttime accident rate at affected crossings nearly tripled after whistle bans were imposed. The daytime accidents at affected crossings remained virtually unchanged. In contrast, nighttime accidents increased 23 percent at 89 FEC crossings where there were no bans.

II. STUDY METHODOLOGY

Approach

The Florida Whistle Ban analysis is straightforward. Use the "Scientific Approach" to examine accidents at whistle ban locations.5 Make a statistical comparison of the number of accidents reported before the ban to the number of accidents reported after the ban. Use control groups to identify or reduce the influence of outside factors.6 Finally, "normalize" data.7 Appendix C discusses Normalizing Historical Data. A chi-square test of homogeneity is used to compare the expected and observed accidents at impacted crossings. The detailed chi-square statistical analysis is found in Appendix D. To begin its study, FRA formed the study hypothesis.8 Hypothesis formation helps to direct the investigation.

Study Hypothesis

If nighttime whistle bans are imposed at highway-rail crossings, the number of accidents occurring at these crossings will significantly increase.

Null Hypothesis 9

The number of accidents at highway-rail crossings will not significantly increase if nighttime whistle bans are imposed.

III. DATA COLLECTION

Selection of Crossings Studied

The National Rail-Highway Crossing Inventory provides a list of FEC's at-grade, highway-rail crossings.10 All selected FEC crossings are public crossings. All selected FEC crossings have active warning devices. FRA identified 600 FEC crossings for its study. These crossings are found in 11 counties along Florida's east coast. Figure 3 shows FEC Crossing Accidents at Impacted Crossings. By 1990, 511 of FEC's crossings were affected by whistle bans. The remaining 89 highway-rail crossings were not affected by whistle bans. Figure 4 shows the growth in FEC ordinance impacted crossings subsequent to permissive state legislation in 1984.

FEC Crossing Ordinances

The effective dates for each crossing ordinance were obtained from FEC. A chronological list of whistle bans by jurisdiction and milepost boundaries is found in Appendix F.

Accident Data Selected

FEC's highway-rail crossing accident reports were segregated into two groups 11 . The "subject group" contained reports for accidents that occurred between 10:00 p.m. and 6:00 a.m. The "control group" contains reports for accidents occurring between 6:01 a.m. and 9:59 p.m. FEC reported a total of 785 crossing accidents between 1975 through 1989. Accident reports with missing or garbled data were eliminated. Also eliminated were reports for accidents at ineligible crossings.12 The study examined 681 accident reports. The subject group contains 253 accident reports. The control group contains 428 accident reports. Appendix E shows FRA's "Rail-Highway Grade Crossing Accident/Incident Report" (Form FRA F 6180-57).

For comparison, 224 CSX Transportation, Incorporated (CSX) highway-rail grade crossings were identified. The CSX crossings also have active warning devices. In addition, these crossings are located in the six Florida counties where both CSX and FEC operate.13

CSX's highway-rail crossing accident reports, totaling 324, also were segregated into two groups. The subject group contains 90 accident reports. The control group contains 234 accident reports.

IV. DATA ANALYSIS

Categorizing Data

The analysis focuses on changes in accident rates once bans are imposed. Florida whistle bans are not effective during the day. Therefore, nighttime accident rates should theoretically increase at crossings affected by bans. In addition, accident rates for the daytime control group were compared to accident rates for the nighttime control group. Since comparisons of daytime and nighttime rates were made for the same crossings, all conditions were identical except the whistle ban itself. Table 1 shows FEC's nighttime grade crossing accident experience. FEC reported 39 accidents at affected crossings during the pre ban period. During the post ban period, FEC reported 115 accidents at affected crossings.

FRA's statistical model predicts 49 post ban accidents if there are no whistle bans. Thus, 66 post ban accidents are unexplained. The 66 unexplained accidents resulted in 11 fatalities and 34 injuries.

The gap, depicted in Figure 3, widened while whistle bans remained in effect.

Daytime Control Group

Table 2 shows no significant change in accident rates during the daytime periods when whistle blowing is permissible.

No-Ordinance FEC Control Group

This control group establishes accident rates for crossings unaffected by whistle ban ordinances. Pre ban accidents were collected for the years, 1975-1984. Post ban accidents were collected for the years 1985-1989. Table 3 shows daytime accident rates were 30 percent lower in the 5 year post ban period. The nighttime rate was 23 percent higher. One explanation for the increase in nighttime rates could be changes in rail or highway operations. For example, annual locomotive miles reported by the FEC for the same periods increased by 11.5 percent.

CSX Transportation, Inc. (CSX) Control Group

Table 4 shows data for CSX. This interstate railroad is not affected by the Florida intrastate railroad whistle ban.

The trend of daytime CSX accident rates is comparable to FEC in the 1985-1989 period. In addition, each carrier showed a decline in accident rates (30 percent for FEC, 61 percent for CSX) between 1985-1989 when compared to the previous 1975-1984 period. Nighttime accident rates increased on both railroads-- 67 percent at CSX's crossings and 23 percent at FEC's non ordinance crossings. Overall, daytime and nighttime combined, CSX experienced a 36 percent decline in accidents--from .00911 accidents per crossing month in the 1975-1984 period to .00588 accidents per crossing-month in the 1985-1989 period. FEC's accident rates at no-ordinance crossings declined 17 percent-- from .00496 accidents per crossing month in the 1975-1984 period to .00412 accidents per crossing month in the 1985-1989 period.

Concurrently, the combined accident rate at FEC's impacted crossings increased 75 percent since the bans went into effect.

V. CONCLUSION

FRA conducted contingency table analyses using the chi-square statistic for subject and control groups. The purpose of a contingency table analysis is to determine whether a dependence exists between two qualitative variables. In this case, the objective of the study is to research whether the ordinance on whistle blowing increases freight crossing accidents.

The chi-square statistic for the study group of FEC nighttime accidents was compared to control groups. The differences between expected and observed accidents are great enough to reject the null hypothesis. The study confirms that nighttime whistle bans at impacted crossings cause highway-rail crossing accidents.

The value of the chi-square statistic determined that whistle ban ordinances were the only identifiable difference between crossings in the subject data set and those in the control groups. If nighttime whistle bans are imposed at highway-rail crossings, the number of accidents will significantly increase.

VI. ACTIONS FOLLOWING COMPLETION OF STUDY

In August 1990, FRA provided its study to officials of each Florida municipality with bans in effect. The study was also given to the Florida State Department of Transportation, and 15 State legislators. FRA requested that each recipient offer explanations, other than the whistle bans, which might account for the near tripling of the accident rates for subject groups. Several explanations were offered. Appendix G shows FRA's response to Florida State Representative Cosgrove. Appendix H shows FRA's response to Roman A. Yoder, Jr. The respondents' explanations include changes in population density, seasonal fluctuations in highway traffic, general increases in rail and highway traffic, and drug use (including alcohol). However, none of these variables explained the variance. If the above variables are responsible for the increase in crossing accidents, then the total number of fatal highway accidents should increase proportionately. Appendix I shows population increases in Florida. Appendix J shows that the number of fatal highway accidents did not increase significantly in Florida during the study.

Despite study results, no municipality rescinded its ban. Also, the Florida legislature offered no changes to the enabling legislation that authorizes the bans. Subsequently, 26 more crossings became subject to ordinances in 1990. The total number of crossings subject to bans increased to 537.

FRA does not object to whistle bans per se. However, state and local authorities need to compensate for the hazard created by whistle bans. Remedies acceptable to FRA include increasing law enforcement, installing immovable highway dividers, grade separating high-traffic crossings, or closing low-use crossings. Analysis of 1990 and preliminary 1991 data (first half of the year) shows a continuation of the post ban accidents. Appendix K illustrates FEC accident trends. Since July 1984, an upward increase in FEC crossing accidents has occurred.

A 1990 study of Oregon train whistle bans showed similar accident trends. The Public Utility Commission of Oregon (OPUC) discovered an increase in grade crossing accidents where whistle bans were in effect. Appendix L shows excerpts from the OPUC study. Based on the studies of OPUC and FRA, the whistle prohibition order issued by the OPUC was rescinded on September 13, 1991. Appendix M is a copy of the OPUC decision. Following its investigation of FEC whistle ban accidents, FRA issued Emergency Order No. 15 on July 26, 1991. This decision requires the FEC to sound train horns when approaching public highway-rail crossings. Specifically, FEC was ordered to follow the operating rules governing horn use that were in effect before the state-permissive train whistle ban. FRA recognizes that nighttime train whistles can be an inconvenience to residents near the railroad right-of-way. However, whistles can also save lives. Appendix N is FRA's Emergency Order No. 15.

FRA received 21 petitions requesting withdrawal or modification of Emergency Order No. 15. The Petitioners included 2 counties and 13 cities. About 31 percent of the impacted crossings are represented by the petitions. During the administrative review of this order, FRA addressed arguments concerning the accuracy of the FRA's whistle ban study, other potential causes for the accident increase, the FRA's justification for issuing the Emergency Order, and the FRA's willingness to consider alternative or mitigating remedies. These issues are discussed in Appendix O, FRA's Conference Notice No. 3, issued December 5, 1991. Appendix P is FRA's response to U.S. Representative Ileana Ros-Lehtinen. Her constituents questioned FRA's findings and conclusions.

On November 3, 1992, FRA issued Conference Notice No. 7. Conference Notice No. 7 provides performance specifications for full highway-rail crossing barriers, traffic divisional islands, and temporary crossing closures. Conference Notice No. 7 is found in Appendix Q.

In response to comments received from petitioners, FRA reviewed Emergency Order No. 15. Also, FRA studied alternate remedial measures. On August 31, 1993, FRA amended Emergency Order No. 15. The amended order is found in Appendix R. As one alternative to the Emergency Order, FRA proposed the temporary, night time closure of selected highways leading to crossings. Some communities have placed many crossings close together. The lower volume of highway traffic at night could be redirected to fewer crossings without a significant impact on traffic flow.

Nationwide, the FRA is actively working to close 25 percent of public highway-rail crossings. Many of these crossings are redundant. Redundant crossings can be eliminated with little impact on the traveling public. When crossings cannot be closed, FRA recommends a number of alternatives. Communities can grade separate crossings, i.e., build bridges, or install 4-quadrant gates. At crossings with standard gates, traffic divisional islands can be installed. These barriers prevent highway users from driving around crossing gates. Because barriers are a highway device, FRA is working with State and Federal highway authorities to define the requirements for installation. If measures are taken to assure highway-rail crossing safety, the FRA will modify the Emergency Order.

FRA has issued no other emergency orders regarding train whistles or horns. For safety reasons, the FRA will not approve train whistle ban ordinances without alternate safety criteria. These criteria are outlined in Appendix S. Locomotive horns and whistles are exempt from noise emission standards established by the Environmental Protection Agency (EPA). However, FRA has contracted the Volpe National Transportation Systems Center (VNTSC) to develop an optimal warning signal for locomotive whistles. If successful, whistle noise can be reduced for communities while not compromising safety. VNTSC also is investigating the installation of audible warning devices directly at crossings.

Body Footnotes

¹ U.S., Federal Railroad Administration, Florida's Train Whistle Ban, July 1990.
² U.S., Federal Railroad Administration, Florida's Train Whistle Ban, 2nd edition, September 1992.
³ All affected highway-rail grade crossings are required to be equipped with crossing gates, flashing lights, bells, and special highway advance warning signs.
⁴ During 1990, 26 additional crossings became subject to bans. The total FEC crossings subject to bans increased to 537.
⁵ The scientific approach is a systematic, controlled, empirical, and critical investigation of hypothetical propositions about the presumed relations among natural phenomena.
⁶ For example, changes in highway and rail traffic, could affect the number of accidents at particular crossings.
⁷ To compare before and after data whose histories are not equal (i.e., length of time), researchers "normalize" the data. This technique expresses data in a common unit of measure, e.g., accidents per month. ⁸ A hypothesis is a conjectural statement of the relation between two or more variables.
⁹ The null hypothesis is a statistical proposition which states that there is no relation between the variables (of the problem). The null hypothesis is a succinct way to express the testing of obtained data against chance expectation.
¹⁰ FRA maintains a computer-based file of all highway-rail crossings in the United States. States and railroads voluntarily provide changes in crossing information to FRA. In the year ended March 31, 1990, the FRA processed more than 90,000 inventory updates. Each crossing is assigned a unique number. This number allows precise crossing identification. The number is included in all crossing accident reports.
¹¹ The Federal Railroad Safety Act of 1970 (P.L. 91-458) and the Accident Reports Act (45 U.S.C. 38-34), require railroads to file accident reports with FRA. Any accident involving trains and an automobile, bus, truck, motorcycle, bicycle, farm vehicle, pedestrian, or other highway user at a highway-rail crossing must be reported to the FRA. Since 1975, data from these reports has been captured into FRA databases.
¹² Ineligible crossings include private, closed, and grade separated crossings, crossings located on abandoned track, and crossings without active warning devices.
¹³ The counties are Broward, Dade, Duval, Martin, Palm Beach, and Volusia.

Railroad-Highway Grade Crossing Warning Signs and Signals; Audible Warnings; Exercise of Reasonable Care; Blocking Highways, Roads, and Streets During Darkness.*

1. Every railroad company shall exercise reasonable care for the safety of motorists whenever its track crosses a highway and shall be responsible for erecting and maintaining crossbuck grade crossing warning signs in accordance with the uniform system of traffic control devices adopted pursuant to Florida Statute 316.0745. Such crossbuck signs shall be erected and maintained at all public or private railroad-highway grade crossings.

2. Advance railroad warning signs and pavement markings shall be installed and maintained at public railroad-highway grade crossings in accordance with the uniform system of traffic control devices by the governmental entity having jurisdiction over or maintenance responsibility for the highway or street. All persons approaching a railroad-highway grade crossings shall exercise reasonable care for their own safety and for the safety of railroad train crews as well as for the safety of train or vehicle passengers.

3. Except as provided in subsection (4), any railroad train approaching within 1,500 feet of a public railroad-highway grade crossing shall emit a signal audible for such distance.

4. 1. No railroad train of a railroad company operating wholly within this state may emit an audible warning signal between the hours of 10:00 p.m. and 6:00 a.m. in advance of any public railroad-highway grade crossing with train-activated automatic traffic control devices, which include flashing lights, bells, and crossing gates, where the municipality or county has in effect an ordinance that unconditionally prohibits the sounding of railroad train horns and whistles during such hours at all public railroad-highway grade crossings so signalized within that municipality or country and where the municipality, county, or state has erected signs at the crossings involved announcing that railroad train horns and whistles may not be sounded during such hours. Signs so erected shall be in conformance with the uniform system of traffic control devices as specified in Florida Statute 316.0745.

   2. Nothing in this subsection shall be construed to nullify the liability provisions of Florida Statute 768.28.

5. 1. Whenever a railroad train engages in a switching operation or stops so as to block a public highway, street, or road at any time from one-half hour after sunset to one-half hour before sunrise, the crew of the railroad train shall cause to be placed a lighted fusee or other visual warning device in both directions from the railroad train upon or at the edge of the pavement of the highway, street, or road to warn approaching motorists of the railroad train blocking the highway, street, or road. However, this subsection does not apply to railroad-highway grade crossings at which there are automatic warning devices properly functioning or at which there is adequate lighting.

   2. A person who violates any provision of paragraph (a) is guilty of a misdemeanor of the second degree, punishable as provided in Florida Statute 775.082, or 775.083.

History. s. 34, ch. 1987, 1874; RS 2264; GS 2841; ch. 7940, 1919; RGS 4529; CGL 6592; s. 1, ch. 73-336; s. 52, ch. 76-31; s. 5, ch.80-289; ss. 2, 3, ch. 81-318; ss. 1, 12, 14, ch. 82- 90; s.1, ch. 84-73; s. 39, ch. 86-243.

*Note. Expires October 1, 1992, pursuant to s. 14, ch. 82-90, and is scheduled for review pursuant to s. 11.61 in advance of that date.

Agenda Item No. 4 (a)
6-5-84

ORDINANCE NO.

ORDINANCE AMENDING SECTION 21-27.3 AND REPEALING SECTION 21-27.4 OF THE CODE OF METROPOLITAN DADE COUNTY, FLORIDA, RELATING TO PROHIBITION OF RAILROAD TRAIN WHISTLE AND HORN NOISE POLLUTION; PROVIDING FOR COUNTY-WIDE UNIFORM PROHIBITION FOR PUBLIC AT-GRADE CROSSINGS HAVING TRAIN-ACTIVATED TRAFFIC CONTROL DEVICES; PROVIDING SEVERABILITY, INCLUSION IN THE CODE, AND AN EFFECTIVE DATE

BE IT ORDAINED BY THE BOARD OF COUNTY COMMISSIONERS OF DADE COUNTY, FLORIDA:

1. Section 21-27.3 of the Code of Metropolitan Dade County, Florida, is hereby amended as follows:

   Section 21-27.3 Railroad train whistle and horn noise pollution prohibited--generally.

   1. Definitions.
      "Person" means any individual, corporation, partnership, other legal entity, or any agent or employee thereof.

   2. Applicability.
      The provisions of this section shall be applicable only to public railroad train crossings at grade within the incorporated or unincorporated areas of Dade County, which are equipped with train-activated, automatic traffic control devices, which shall include ringing bell, flashing light signals, and automatic crossing gates on both sides of the railroad train track.

   3. Prohibited acts.
      Notwithstanding anything in this Code to the contrary, it shall be unlawful and a public nuisance for any person operating a railroad train of a railroad company operating wholly within this state to blow or activate, or permit to be blown or activated, any horn or whistle from the railroad train between 10:00 p.m. and 6:00 a.m. at and in advance of any public at-grade crossing where the municipality, county, or state has erected a sign at the crossing involved announcing that railroad train horns and whistles will not be sounded during the aforesaid hours.

   4. Enforcement; costs and attorneys' fees; injunctions; criminal penalty.

2. Section 21-27.4 of the Code of Metropolitan Dade County, Florida, is hereby repealed in its entirety.

3. If any section, subsection, sentence, clause or provision of this ordinance is held invalid, the remainder of this ordinance shall not be affected thereby.

4. It is the intention of the Board of County Commissioners, and it is hereby ordained that the provisions of this ordinance shall become and be made a part of the Code of Metropolitan Dade County, Florida; and that the sections of this ordinance may be renumbered or relettered to accomplish such intention, and the word "ordinance" may be changed to "section," "article," or other appropriate word.

5. The provisions of this ordinance shall become effective on July 1, 1984.

PASSED AND ADOPTED:

Approved by County Attorney as to form and legal sufficiency. Initialed by NAG Prepared by: Initialed by PST

NORMALIZING DATA

"Normalizing" is a data averaging technique. This convention expresses data in a common unit of measure. For the purposes of this study, railroad-highway crossing accident data is expressed in "crossing accidents per month." This allows the comparison of data for the 10 years preceding the Florida whistle ban on July 1, 1984 to accidents for the 5 years following the ban. This study examined 68,024 crossing-months of pre-ordinance accident experience and 23,474 crossing-months of post-ordinance accident experience.

Normalizing Techniques

Two "normalizing" techniques are used in this study. The "Maximum History" technique uses all 15 years of available data (1975-1989). The "60 Month History" technique uses 60 months of pre- and post-whistle ban accident observations (1979-1984 and 1984-1989) Table C1 shows the nighttime crossing months, number of accidents and accident rates for each of these techniques.

Under either "normalizing" technique, the number of nighttime accidents in the post-ordinance period is three times larger than the accident rate before the ordinance was implemented.

The product of the number of crossings affected by each ordinance and the number of pre ordinance, or after ordinance months, is summed for all the impacted crossings.

Control Groups

Table C2 shows the daytime crossing months, number of accidents, and accident rates using normalizing techniques. Daytime accident rates increased 19 and 23 percent, respectively, under the "maximum history" and "60-month history" measures. The increase may be due to demographic changes and increases in traffic volumes. For example, FEC locomotive mileage increased 11.5 percent between the 1975-1984 10-year average and the 1985-1989 5-year average.

DETAILED STATISTICAL ANALYSIS

This appendix contains three contingency table analyses, which incorporate exposure data (a measure of a crossing's accident potential) for subject and control groups. Because definitive data was not available for highway traffic and train counts for each crossing, this study used crossing-months experience as a measure of exposure. In the first contingency table analysis, the number of crossing-months was equal for the subject group and control group for the before and after periods. However, the number of hours differed between the two groups. The subject group represented the nighttime period (8 hours) impacted by the ban. The control group, unaffected by the ordinance, represented the daytime period (16 hours). This difference in number of hours was considered in calculating the number of expected accidents. Since the time period for the control group was twice as long as that for the subject group, the opportunity for accident occurrence was two times greater for the control group.

The purpose of a contingency table analysis is to determine whether a dependence exists between two qualitative variables. In this case, the objective of the study is to research whether the ordinance on whistle blowing increased freight crossing accidents. The appropriate test statistic, in this analysis for a test of hypothesis is the chi-square (c) statistic. The chi-square test of homogeneity is computed for each analysis. This test computes a discrepancy measure based on observed and expected frequencies (if bans had not been imposed) for the individual cells. The value of the chi-square statistic determines if the differences between the observed and expected cell counts is large enough to reject the null hypothesis (H0).

A rejection of the null hypothesis asserts acceptance of the alternative hypothesis (H1). The hypotheses are:

H₀: Ordinance did not increase freight crossing accidents.
H₁: Ordinance did increase freight crossing accidents.

The test statistic (chi-square) to test the hypotheses is:

O_ij is the observed frequency for level i of the first classification method and level j of the second classification method.

r = # of rows.

c = # of columns.

The contingency table analyses for the three different control groups are presented below.

CONTROL GROUPS:

1. Daytime

Observed Accidents

	Before	After
Subject	39	115	154
Control	110	109	219
	149	224	373

Exposure (In Hours)

	Before	After
Subject	8	8	16
Control	16	16	32
	24	24	48

Proportions

	Before	After
Subject	8/48 = .167	8/48 = .167	.334
Control	16/48 = .333	16/48 = .333	.666
	.500	.500	1.000

Expected Accidents

	Before	After
Subject	(373)(.167) = 62	(373)(.167) = 63	125
Control	(373)(.333) = 124	(373)(.333) = 124	248
	186	187	373

2. No-Ordinance FEC Crossings

Observed Accidents

	Before	After
Subject	39	115	154
Control	13		21
	52	123	175

Exposure (Crossing-Months Experience)

	Before	After
Subject	23,474	23,474	46,948
Control	10,680	5,340	16,020
	34,154	28,814	62,968

Proportions

	Before	After
Subject	23,474/62,968 = .373	23,474/62,968 = .373	.746
Control	10,680/62,968 = .170	5,340/62,968 = .084	.254
	.543	.457	1.000

Expected Accidents

	Before	After
Subject	(175)(.373) = 65	(175)(.373) = 65	130
Control	(175)(.170) = 30	(175)(.084) = 15	45
	95	80	175

3. CSX Transportation (No-Ordinance) Crossings

Observed Accidents

	Before	After
Subject	39	115	154
Control	49	41	90
	88	156	244

Exposure (Crossing-Months Experience)

	Before	After
Subject	23,474	23,474	46,948
Control	26,880	13,440	40,320
	50,354	36,914	87,268

Proportions

	Before	After
Subject	23,474/87,268 = .269	23,474/87,268 = .269	.538
Control	26,880/87,268 = .308	13,440/87,268 = .154	.462
	.577	.423	1.000

Expected Accidents

Before	After
Subject	(244)(.269) = 66	(244)(.269) = 65	131
Control	(244)(.308) = 75	(244)(.154) = 38	113
	141	103	244

SUMMARY The absolute magnitude of the chi-square test results is very real. The data was analyzed with three well chosen control groups. Each analysis showed highly significant results at the .005 level of significance, which is more stringent than the normal (.05) level of significance. The ordinance on whistle blowing remains the only apparent explanation for the tripling in nighttime accidents at the impacted crossings.

Notes:
^**John E. Freund, Ronald E. Walpole, Mathematical Statistics, Fourth ed. (Englewood Cliffs: Prentice, 1987) pg. 438.

The Honorable John F. Cosgrove
Representative, 119th District
Florida House of Representatives
201 West Flagler Street
Miami, Florida 33130

Dear Mr. Cosgrove:

Thank you for your letter of September 12 regarding the Federal Railroad Administration's (FRA's) study, Florida's Train Whistle Ban. You enclosed a letter from Mr. John A. Cavalier in which several issues were raised regarding the study, its approach and findings. I have addressed each point in the enclosure to this letter.

The basic finding of the original study is that nighttime accidents are occurring at nearly three times the rate that they were before the whistle bans became effective. I have asked everyone who might be knowledgeable regarding this subject to account for this increase in accident rates. None have, but several have questioned the report, its procedures and even its facts. I trust that all the discussion about control groups, measures of rate, population densities, etc. will not serve to cloud the basic issue. So far, there is no other supportable explanation.

In answer to a question you forwarded, I have provided an account of FRA's other activities, which seek to improve highway-rail crossing safety. I am proud of FRA's record of accomplishment and current activities regarding crossing safety, but I am chagrinned by the continuing toll of crossing casualties. I would suggest it might be time for all of us concerned with public safety, and particularly with crossing safety, to ask, "what more can I do?"

I appreciate your interest in this matter. I hope we can work together to improve highway-rail crossing safety.

Sincerely,

J. W. Walsh
Associate Administrator for Safety

Enclosure w/attachments (6)

Enclosure

Quoting from Mr. John A. Cavalier's letter of September 7:

"1. I question the use of "control groups" based on multiplying the number of months times the number of crossings. A more accurate report would show the traffic count of automobiles at the crossing, the frequency of trains, and the length (sic) of trains for both CSX and FEC." (Florida East Coast Railway)

Four control groups were used in this study:

1. Pre-ordinance periods for the impacted crossings equal in duration to the post-ordinance periods through December 1989.

2. The daytime experience for the impacted crossings during pre-and post-ordinance periods of equal duration.

3. The nighttime experience for eligible FEC crossings for which no ordinance had been implemented as of December 1989.

4. CSX Transportation crossings equipped with devices similar to the impacted FEC crossings in those counties in which both FEC and CSX operate.

Multiplying the number of crossings times the number of months of experience to realize a figure for "crossing-months" of experience is a common statistical procedure used to measure and express collective experience. The technique does not even enter into those findings based on the first two control groups noted, and is used only to normalize accident rates in the latter two control groups.

Reliable current and historical highway traffic counts for individual crossings (State, county and municipal streets and highways) are not available to the Federal Railroad Administration (FRA). A coordinated and collective effort by concerned highway authorities may possibly produce such information from local records. Similarly, the historical frequency of trains and the current day-to-day schedules are not available in sufficient detail. Schedules are too variable. The length of trains is not available.

As a surrogate for traffic and train counts, I have compiled data regarding FEC's Locomotive Miles Operated, by year (See Attachment #1), and regarding Vehicle Miles Driven in Florida, also by year (See Attachment #2). The latter, by county, would be useful, but is not available to the FRA. Neither of these measures would account for the near tripling of the nighttime accident rate at FEC's impacted crossings, nor for the differential in the change in nighttime versus daytime accident rates.

"2. The study includes data from 1975 through 1989. Again, traffic counts for these periods plus population densities around crossings need to be considered. Growth along the east coast in condos, shopping malls, new highways, etc. are important factors, as well as the number of trains daily by both CSX and FEC."

In addition to the data for vehicle and train miles driven/operated noted above, I have compiled information on the number of registered motor vehicles and drivers in Florida for each year, 1975 through 1988. (1989 data is unavailable.) (See Attachments #3 and #4.) These are reasonable surrogate measures for population density and vehicle usage. We have asked the Florida Department of Transportation to provide this information by county, by year for the 11 counties in which the FEC operates. Once again, these figures do not explain the sudden increase in the FEC nighttime accident rate.

"3. Table 4 shows that CSX nighttime accidents up 67% versus FEC up 23% at no-ordinance crossings. Why? Again, each accident needs review as to cause. How many were a result of cars going around gates that are down? Should the fine for this be increased?"

The increase in nighttime accident rates at CSX Transportation and FEC no-ordinance crossings is probably attributable to the many factors Mr. Cavalier has already cited, e.g., traffic (both rail and highway) and population growth. The point in including these numbers in the original report was to provide control groups (bases for comparison) against which the subject group (FEC's impacted crossings) could be compared. In making such comparisons, one makes assumptions that the control groups are similar to the subject group in all or most essential characteristics. One includes multiple control groups (we used four) in order to minimize the impact of any discrepancies or biases which may be present in any one control group. "Why" these control groups showed increases could be the subject of another, probably more complex, study. For the purposes of this study, the fact that the increase in these control groups amounted to only a fraction of the increase in the subject group reinforces the point that something is unique about the subject group, and provides a base, along with other factors noted in the FRA report, on which to establish "expectations" for the subject group.

We have made a review of each FEC nighttime accident, both before and after ordinance implementation, 1975 through 1989, at both impacted and no-ordinance crossings. The FRA does not determine "cause" for highway-rail crossing accidents. Rather, we collect "circumstances" and "actions" of the parties and equipment involved. Reviewers are left to make their own judgments. The table, titled "FEC Nighttime Crossing Accidents, Tabulated Circumstances, 1975-1989, " (Attachment #5) provides a breakdown of our findings. The comparable columns are the middle two, which display accident experience for equal (EVEN) pre- and post-ordinance periods at the impacted crossings. Proportionally, the major changes are in the category where the motorist reportedly, "drove around or thru the gate." Pre-ordinance-- there were 17 such reports (43.6 percent of the 39 accidents)--versus 96 (83.5 percent) of the post-ordinance 115 accidents. The number of trucks involved went from 3 (7.7 percent) to 16 (13.9 percent) in the post-ordinance period. The number of occurrences where the "motorist passed standing highway vehicle," i.e., another vehicle already stopped at the crossing, jumped from two to nine, while the number of instances where a "motorist drove behind or in front of train and struck or was struck by second train" increased from zero to four. Also of interest, and tending to substantiate these numbers, is the reported number of instances where the motorist action is listed as "other." This is usually used to report instances where a vehicle is trapped on the crossing by standing highway traffic. Eighteen occurrences were reported in both the pre- and post-ordinance periods, a proportional decrease of 66.1 percent. This is intuitively acceptable in that the circumstances which would lead to a vehicle being trapped on a crossing would not be affected one way or another by whether or not the train used a whistle. The number of occurrences should remain the same.

Mr. Cavalier asked whether the fine should be increased for driving around or through a gate. This is certainly a local decision. But I would first ask, "How many time have motorists be cited for such an infraction, after an accident and when no accident occurred?" I would guess, especially in the latter instance, "None!" Increased fines are of little importance without effective enforcement.

"4. The chart on page 10 [FRA's Florida's Train Whistle Ban, July 1990] needs further analysis.

1. Is there a similar chart for CSX?

2. Why is the accident rate curve from July 1984 thru July 1986 relatively flat?

3. Why is the accident rate curve from July 1986 thru December 1989 at a higher rate?"

A similar chart for CSX is attached. (Attachment #6) Three factors mitigate against the immediate (July 1984) appearance of changes in accident rates under the conditions cited. The first ordinance was not effective until July 29, 1984. Additional ordinances went into effect in each month through March 1985 and sporadically in subsequent months, each accompanied by a flurry of controversy and publicity. Ordinances are still being implemented. Figure 3 [FRA's Florida's Train Whistle Ban, July 1990], graphically displays the phased implementation of the ordinances. Obviously, more effect will be apparent in the latter months and years. Also, there is a phenomenon that highway traffic engineers call the "novelty effect" which occurs anytime highway traffic control signal and sign standards are changed. This effect is most pronounced when publicity has preceded, appeared concurrent with, or followed implementation of a change. It causes a sometimes significant delay in the return to "normal," in this case, a wearing-off of the heightened awareness occasioned by the publicity which accompanied the passage and implementation of these ordinances. Lastly, a sharp decline or leveling in FEC's statistics in the fall of 1985 would temporarily have suppressed accident rates, countering forces (the whistle bans) which may have been tending to force rates higher. The delay Mr. Cavalier has noted is most likely attributable to a combination of these factors. The accident rate subsequent to July 1986 appears to be attributable, in large measure, to the bans on whistles.

"5. A study of bans in other states needs to be conducted."

I agree, and my staff has been seeking the cooperation of major railroads and the Association of American Railroads (AAR). Such a study will not be as straight forward as the Florida effort because, to our knowledge, no bans as widespread and as sweeping as the Florida provision have been implemented in other jurisdictions.

"6. What is the F.R.A. doing to improve safety at crossings other than relying on train horns that cannot be heard by automobiles with air conditioning or heaters operating and radio playing?"

Current activities include substantial financial and in-kind support of Operation Lifesaver, a nationwide program which promotes public education regarding hazards at crossings, engineering safety improvements and enforcement of crossing related traffic laws. The FRA is also sponsoring research: at the Volpe National Transportation Systems Center in Cambridge, Massachusetts regarding the efficacy of placing reflectors on the sides of rail cars in order to improve the conspicuity of the rail cars; and, at the Texas Transportation Institute in College Station, Texas regarding public perception and credibility of automatic warning devices and railroad response to, and findings regarding, notification by the public of warning device problems. The FRA is underwriting the cost of producing a pamphlet for national distribution promoting the evaluation of school bus driver training needs (re: crossing safety) based on a program developed in the Lamar Consolidated Independent School District, Texas. Demonstration projects, with FRA sponsorship, are being cooperatively conducted by Kansas State University, the University of Kansas, the Kansas Department of Transportation, the Atchison, Topeka and Santa Fe Railway, the Burlington Norther Railroad and the Union Pacific Railroad at rural crossings equipped with passive devices in southeastern Kansas. These demonstrations will report on motorists reaction to a variety of innovative uses of signs (some new) and reflectors. The FRA has recently negotiated and signed a financial assistance agreement, on behalf of three Department of Transportation (DOT) agencies, with the Law Enforcement Television Network (LETN). This grant will underwrite LETN's development costs for a series of police officer oriented program which will deal with law enforcement at highway-rail crossings, crossing accident investigations, the proper response to hazardous materials involvement in crossing accidents, trespasser prevention, etc. These programs will air on LETN and will reach more than 1,800 police stations nationwide, potentially more than 75,000 sworn officers. The FRA is also pursuing an open regulatory procedure, titled, "Grade Crossing Signal System Safety," FRA Docket No. RSCG-3. This procedure was prompted by concerns for crossing signal system maintenance, inspection and testing procedures currently in use within the industry. Further information is contained in the Federal Register of September 20, 1990, Volume 55, Number 183, pages 38707-38712. Lastly, the FRA collects and maintains a database regarding all highway-rail crossing accidents and is custodian of the U.S. DOT/AAR National Rail-Highway Crossing Inventory System, a computerized database, which contains records for all crossings in the United States. The FRA provides support from the databases, e.g., accident and inventory histories for crossings, accident predictions, etc., to public authorities and to railroads free of charge. FRA's crossing related concerns are broad based and comprehensive and certainly address more than railroad whistle bans along the east coast of Florida.

ATTACHMENT #6

Not Available

U.S. Department
of Transportation
Federal Railroad
Administration

January 8, 1991

Mr. Roman A. Yoder, Jr.
Assistant City Manager
120 Malabar Road, S.E.
Palm Bay, Florida 32907-3009

Dear Mr. Yoder:

Thank you for your letter of September 12 regarding the Federal Railroad Administration's (FRA's) study, Florida's Train Whistle Ban. You raised several questions regarding the study, its approach and findings. I have addressed each point in the enclosure to this letter.

The basic finding of the original study is that nighttime accidents are occurring at nearly three times the rate that they were before the whistle bans became effective. I have asked everyone who might be knowledgeable regarding this subject to account for this increase in accident rates. None have, but several have questioned the report, its procedures and even its facts. I trust that all the discussion about control groups, measures of rate, population densities, etc. will not serve to cloud the basic issue. So far, there is no other supportable explanation.

I appreciate your interest in this matter. If you have any further questions please write again.

Sincerely,

Philip Olekszyk
Acting Associate Administrator for Safety
Enclosure

Enclosure

Quoting from your letter of September 12:

"1. It is difficult to determine what trends may have been filtered out of the original data such as increases in accidents or volumes, particularly when the control groups (CSX and non-ban FEC crossings) utilize a ten-year pre-history versus a five-year post history."

We are aware of no trends which have been filtered out of the original data. As pointed out in the report, since no bans have been established for the CSX and the non-ban FEC crossings, there is no post-ban history for the crossings. In order to make a comparison, 10 versus 5 years was arbitrarily selected because it approximated the maximum post-ban history period (65 months). Data was normalized to a "per month" base in order to facilitate comparisons. If you would like to recommend a different split, and a rationale, we will recalculate the accident rates.

"2. What was average train traffic volume by year for post-and pre-ban crossings and for the control groups? This data would be useful to determine the number of accidents per rail mile traveled--a more useful measure than accidents per crossing or accidents per month."

The historical frequency of trains and the current day-to-day schedules are not available in sufficient detail to support such a comparison. Schedules are too variable. I have compiled data regarding FEC's locomotive Miles Operated, by year. (See Attachment #1.) This data offers no explanation for the near tripling of the nighttime accident rate at FEC's impacted crossings. FEC's non-ban crossings are exposed to the same rail traffic as the impacted crossings (See discussion on page 7 of the report). Similar data for CSX's operations within Florida is not available.

"3. No trend is shown for increases/decreases occurring over time, i.e., it can be assumed that in year one of the ten-year pre-ban history, that accidents were lower than in year ten of that ten-year history. Projecting whatever trends this data depicts would be useful as a measure of actual versus projected accidents."

Conditions change. "Old" data must be judiciously applied. The oldest data from the 10 year pre-ban period is now 15 years old. For example, your assumption that the number of accidents in year one of the 10 year pre-ban period was lower than in year ten of the 10 year period is wrong. The number of FEC crossings has decreased. The number of FEC crossings equipped with automatic warning devices has increased. Some sort of normalizing, e.g., accidents per crossing per month by type of warning device, is necessary in order to apply this data. However, in order to address your concern, I have prepared a graphic (Attachment #2), which tallies all FEC crossing accidents through the pre- and post-ban periods, from 1975 through 1989. A review of this data shows that the gains achieved by the FEC by eliminating some trackage and by equipping numerous crossings with gates, flashing lights and bells have been lost since the advent of the Florida whistle bans.

"4. Local population, average daily traffic or other measures of increased crossing use were not analyzed. It is apparent that the faster growing and more populated areas have the whistle ban while the less populated and slower areas do not. For example, the City of Palm Bay's population has increased by over 220 percent in the ten-year period 1980 to 1990, so even a 200 percent increase in crossing accident would still maintain an equivalent accidents per capita figure."

Local population and average daily traffic were not included for two reasons. First, such data was not readily available to the Federal Railroad Administration (FRA). Second, by judicious selection of the control groups such influences have been accommodated. However, on the first point, we have compiled some state-wide data regarding the number of registered motor vehicles and licensed drivers in Florida for each year, 1975 through 1988. (1989 data is unavailable.) (See Attachments #3 and #4) Also, on the second point (daily traffic), we have gathered data on Vehicle Miles Driven in Florida, Attachment #5. Data regarding the number of registered vehicles, drivers and vehicle miles driven, by county, would be useful, but is not available to the FRA. Florida's Department of Transportation has already provided some additional data and is considering further compilations. A coordinated and collective effort by concerned municipal or highway authorities could possibly produce such information (especially historical crossing specific traffic counts) from local records. The selection of the control groups, daytime at the same crossings, FEC crossings not impacted, and similarly equipped CSX crossings in the same counties, should have accounted for the factors you have cited. For example, if accidents per capita is a valid measure, and your population has increased so sharply, is this reflected in the daytime accident experience at those same crossings, or in the non-impacted crossings, both FEC and CSX? It is not. The data so far available to this office offer no explanation for the near tripling of the nighttime accident rate at FEC's impacted crossings, nor for the differential in the change in nighttime versus daytime accident rates.

"5. The study failed to utilize more advanced statistical modeling and measures of validity such as regression analysis and variance measures."

An analysis has been performed using a chi-square test of homogeneity for each control group versus the impacted (or treatment) group. The test computes a discrepancy measure based on observed and expected frequencies for individual cells. The value of the chi-square statistic determines if the differences between the observed and expected cell counts is large enough to reject the null hypothesis (H0 ). A rejection of the null hypothesis asserts acceptance of the alternative hypothesis (H1 ). The hypotheses are:

H0 : Ordinance did not increase freight crossing accidents.
H1 : Ordinance did increase freight crossing accidents.

In each analyses (daytime experience at impacted crossings, non-impacted FEC crossings and similarly equipped CSX crossings) the null hypothesis was rejected. [The full analysis is included as the Appendix D to this report.]

"6. The obvious bias shown in the language of the report, particularly in the Synopsis and Background sections and the lack of statistically valid procedures presented by the report lead to questionable findings."

This project, an assessment of the Florida whistle bans, was begun pursuant to a specific Congressional mandate. The request and our initial efforts were neutral. In fact, we were genuinely dubious about finding anything of substantive value. We were as surprised with the findings as I am sure you were. We have checked and rechecked the data and our procedures. We continue to seek explanations for the change, which might point to something other than the whistle bans. None have been offered, which have withstood scrutiny. If the language of the report is biased, we offer no apology. The findings are startling, even shocking, and if our bias in favor of safety is reflected in the report, we have done our job!

ATTACHMENT 1-5 and APPENDICES I-K

Not available.

APPENDIX L

PUBLIC UTILITY COMMISSION OF OREGON
LABOR & INDUSTRIES BUILDING, SALEM, OREGON 97310-0335 PHONE (503) 378-6660

December 19, 1990

Bruce George
Federal Railroad Administration
400 Seventh Street S.W.
Washington, D.C. 20590

RE: Oregon Train Whistle Bans

The State of Oregon is still very interested in the results of the FRA report entitled "Florida's Train Whistle Ban." Following the International Symposium On Railroad-Highway Grade Crossing Research and Safety in Knoxville, Tennesee, October 31-November 3, 1990, Oregon Public Utility Commission (OPUC) staff conducted a study of public crossings in Oregon where the OPUC had invoked a train whistle ban. We tried to model our study after the Florida study. We have 26 public crossings with OPUC invoked train whistle prohibitions. All of these prohibitions are 24 hour bans, not the nighttime bans in place in Florida. All the crossings with train whistle bans are equipped with flashing light and automatic gate signals and audible warning devices. We compared accidents for equal periods before and after the whistle ban took effect to evaluate the effectiveness of train whistling at crossings. We were able to evaluate 1,401 months of pre and post whistle ban at fully signalized crossings.

Our study revealed an increase in the number of crossing accidents from two to nine during the study period. Like the Florida study, we were unable to find any other factors that could have contributed to the increase in accidents.

I have attached a summary of our study. If you are interested in any additional information regarding the study or the crossings involved, please call me at (503) 378-6660. Have you received any feedback contrary to the findings of the Florida study? Specifically, have you heard anything from the cities, counties or Florida Secretary of Transportation regarding possible flaws or contributory factors to the accident increase cited in your study?

Craig J. Reiley
Manager, Crossing Safety Section

ORDER NO. 91-1164
ENTERED SEPT 13, 1991

BEFORE THE PUBLIC UTILITY COMMISSION OF OREGON

RX 355

In the Matter of the Petition of the City of EUGENE to Establish a Whistle-Free Zone on the Southern Pacific Transportation Company's Main Line, Between Hilyard and Van Buren Streets, in Eugene, Lane County. Oregon.

DISPOSITION: EUGENE WHISTLE PROHIBITION ORDER RESCINDED, EXCEPT AS TO SAFETY IMPROVEMENTS

Procedural History

On August 2, 1989, the Commission issued Order No. 89-1037 in this docket. The order granted, in part, the petition of the City of Eugene for a whistle prohibition order for ten public rail crossings on a section of Southern Pacific (SP) main line which runs through central Eugene. An appeal was filed by SP. Southern Pacific Transportation Company v. Oregon PUC, Marion County Circuit Court, Civil No. 90C10181.

On October 22, 1990, the Marion County Circuit Court, pursuant to a motion by SP, entered an order directing the Commission to reopen the record to consider a federal government report on Florida's nighttime train whistle ban.

On March 26, 1991, a hearing was held before Hearings Officer Simon J. Fitch to take additional evidence relating to the federal report. The following appearances were entered:

For Southern Pacific Transportation Company:

• Ian Whitlock
  Attorney at Law
  Portland, Oregon

For the City of Eugene:

• Milo Mecham
  Attorney at Law
  Eugene, Oregon

For the Commission staff:

• Machael Weirich
  Assistant Attorney General
  Salem, Oregon

Based upon the record herein, the Commission makes the following:

FINDINGS OF FACT

The "Florida Study"

In July 1990, the United States Department of Transportation, Federal Railroad Administration (FRA) issued a study entitled "Florida's Train Whistle Ban" (the "Florida Study"). The study reviewed the nighttime accident history of the Florida East Coast Railway Company (FEC) at highway-rail crossings where a nighttime whistle prohibition was in effect. The study covered a 65-month period beginning in 1984. The whistle bans, imposed by individual counties and cities, applied only to crossing equipped with gates, flashing lights, bells, and special advance warning signs. The whistle bans involved in the study were in effect only between the hours of 10:00 p.m. and 6:00 a.m. The advance warning signs read: NO TRAIN HORN, 10:00 P.M. TO 6:00 A.M.

The study reviewed accident statistics for 511 FEC crossings subject to whistle prohibitions, in each case comparing equal periods of time before and after the implementation of the ban. The number of nighttime accidents at these crossings increased 195 percent, from 39 accidents in the period before the bans to 115 after the bans.

The study compared these accident statistics with those for three control groups. The first control group consisted of the same 511 FEC crossings during daytime (6:00 a.m. to 10:00 p.m.) when whistles were still sounded at each crossing. For the same time periods, daytime accidents increased less than 1 percent, from 108 to 109.

The second control group consisted of FEC crossings with no whistle ban. In this category, accident rates between 1974 and 1984 were compared with rates between 1984 and 1989. Because the whistle bans began to be implemented in 1984, this provided similar "before and after" comparison periods. The daytime accident rate at these crossings was 30 percent lower in the five-year period than in the preceding ten years. The nighttime accident rate increased 23 percent in the five-year period, compared to the preceding ten years.

The third control group consisted of data from another railroad with Florida operations, CSX Transportation (CSX). Accident data for the same fifteen-year period (1975-89) were available for 224 similarly equipped CSX crossings in the six counties in which both railroads operate. CSX was not required to comply with the whistle bans because it operated interstate. CSX daytime accident rates decreased 61 percent between the five- and ten-year comparison periods. Nighttime rates increased 67 percent.

The combined daytime and nighttime accident rate at FEC's crossings with a whistle ban has increased 75 percent. The combined rate at crossings without a ban had decreased 17 percent.

The study concludes:

• The only identifiable difference between the crossings in the subject data set (the [whistle ban] ordinance impacted crossings) and the control groups remains the [whistle ban] ordinances. FEC compliance with the ordinances, the failure to use train whistles, remains the only explanation for the abrupt (200 percent) increases in the nighttime accident rate at the impacted crossings).

Florida Study, SP Ex. 1R, p.1.

Statistical Accuracy

Subsequent to the distribution of the Florida Study, the Federal Highway Administration conducted a statistical analysis¹ of the study data to determine the reliability of the conclusions. The analysis, made a part of the record in this docket, concluded that the initial study findings were correct and withstood critical statistical analysis.

Additional Information

FRA received questions from a Florida legislator and a Florida city official regarding the methodology of the study. FRA conceded that the study did not look at highway traffic counts at individual crossings, at frequency of trains or at train schedules. This data was not available to FRA. As a surrogate, FRA did look at FEC Locomotive Miles Operated, by year and month, and at Vehicle Miles Driven in Florida, by year.² It concluded that neither measure could account for the significant increase in the nighttime accident rate.

FRA was asked if it took population density and development such as condos, new shopping malls, and new highways into account. The study did not specifically do so. As a surrogate for population density, FRA looked at registered motor vehicles and drivers by county and year and concluded that the increased registration did not explain the sudden increase in crossing accidents. FRA did state that the increase in volume, since the increases at crossings without a whistle ban were probably attributable to such factors.

Accident Circumstances

In response to the foregoing inquiries, FRA also reviewed the individual circumstances of the nighttime accidents. Staff Ex. 2R, Attachment 5. The data reflected an increase after whistle prohibition in the proportion of accidents caused by motorists driving around or through the crossing gates from 43.6 percent to 83.5 percent of the total. The number of occurrences where a motorist passed another vehicle stopped at the crossing increased from two to nine. The number of times a motorist drove behind or in front of a train and struck or was struck by the train increased from zero to four. Staff Ex. 2R, p.3.

Oregon Study

Staff conducted a study of crossings in Oregon where whistle prohibitions have been ordered by the Commission. The data base, 18 crossings, is much smaller than that used in the Florida Study. In its study, staff identified those crossings where the Commission had ordered a train whistle ban and compiled the accident history for each crossing. The accident history was limited by the length of time either prior or subsequent to the effective date of a whistle ban. Staff found that accidents increased from two to six (200 percent) at the crossings in the Oregon study after the whistle ban had been in effect.

All six of the accidents at whistle-free crossings occurred at four crossings in Salem in an area where a 24-hour whistle prohibition was ordered by the Commission. RX 22, Order No. 84-158. Of the remaining 14 crossings in staff's study, the twelve crossings in Salem and two in Bend with whistle bans experienced no accidents after implementation of the ban. Staff Ex. 1R, Appendix "A."

The Oregon study does not state the time of day nor the circumstances of the Salem accidents. The number of accidents is too small to be statistically significant, although the increase is consistent with the trend identified with the Florida Study. The data used in staff's study was available at the time of the December 1988 evidentiary hearing in this docket but was not gathered or analyzed. In its prior order, the Commission concluded that the part which the train whistle, or the lack of a whistle, played in the Salem accidents is uncertain. Order No. 89-1037 at 15.

Since 1984 when the Commission received statutory authority to regulate the sounding of train whistles, Oregon experienced an overall decrease in the number of crossing accidents. Between 1984 and 1989, the last year that complete accident data is available, total crossing accidents in Oregon declined from 70 to 47, a reduction of 33 percent. The percentage of these accidents which occurred at gated crossings, however, varied dramatically from a low of 14 percent in 1988 to a high of 38 percent in 1989.

There have been no significant changes in conditions at the affected Eugene crossings since the first hearing.

"Railroad Accidents In Oregon" Annual Report

The Commission publishes annually a report entitled Railroad Accidents in Oregon: Statistics, Summary, and Analysis" (Railroad Accidents Report). The report is prepared by the Commission's Crossing Safety and Rail staff. The most recent edition was published in 1989. The Railroad Accidents Report reviews a wide variety of factors in relation to accident statistics including type of protection (warning device), accidents by county, government jurisdiction, railroad, month of occurrence, hour and day of occurrence, weather conditions, daylight and darkness, population density, size of train, number of tracks, type of railroad equipment, type of motor vehicle, train movement volume, speed of train, and highway volume.

The introductory section of the report states:

In the past two decades the Commission has responded to the legislative mandate to reduce crossing accidents at public grade crossings, wherever possible by closing some crossings and by improving the protection at crossings that remain open . . .

The most effective way to reduce accidents at crossings that remain open continues to be through the installation of automatic protection devices.

1989 Railroad Accident Report, p.1.

Effectiveness of crossing protection devices is compared by means of a formula which takes into account "exposure factor." Exposure factor is the daily average number of trains times the daily average number of vehicles using the crossing with an adjustment for train length and speed.³

In 1989, grade crossings protected by automatic gates represented 84 percent of the total exposure factor in Oregon and experienced 38 percent of the accidents. To date the Railroad Accident Report has not included statistics correlating accidents to crossings where whistle prohibitions are in effect.

The section of the report depicting train vehicle accidents by hour of occurrence and day of occurrence (day of the week) states:

"It is believed that the determining variables for the relationships shown are train and traffic volume, rather than any special characteristics of a particular hour of day."

Southern Pacific (SP)

In April 1990, SP performed a survey of its "closed out" cases involving Oregon crossing accidents for the previous five years. Thirty-two crossing accident files were reviewed. Of these, 13 occurred during the daylight, 18 during darkness, and one occurred at dusk. Of six pending crossing accident cases as of April 29, 1990, four occurred in the darkness.

Only four of the accidents occurred at crossings with gates. Of the 18 accidents occurring in darkness, ten occurred between 10:00 p.m. and 6:00 a.m., the hours of the Eugene Whistle ban. One of these occurred at a private crossing. The SP study did not contain information about the presence or absence of whistle prohibitions at crossings involved. The study included only three accidents from 1989, while staff reported 17. SP did not explain the discrepancy.

Parties' Positions

Southern Pacific asks the Commission to vacate its prior order and not impose a nighttime whistle prohibition at the affected Eugene crossings. The Commission staff has changed its position from that taken at the prior hearing and now agrees with SP that the order should be vacated. The City of Eugene asks the Commission to stand by its original decision, arguing that the Florida Study is not determinative of the Eugene crossing issues.

Previous Findings

The Commission incorporates by reference in this order the Findings of Fact in its prior order, Order No. 89-1037, pp. 3-15. With regard to visibility, the order states in part:

Visibility is limited at most crossings in the section. Buildings line the tracks on both sides, blocking the view of approaching traffic for locomotive crews. Similarly traffic on roadways and sidewalks has only limited views up and down the track before entering the crossing.

Id., p. 11. The previous order also includes the finding that "[e]ach of the ten public crossings on this section of line has at least one blind quadrant. Five of the crossings are blind in all four quadrants." Id., p. 12. "Many of the crossings with severely restricted visibility for motorists are also located on or near curves in the tracks. As a result, visibility is very poor for train engineers as well as motorists." Id., p. 13.

OPINION

Crossing safety is affected by a large number of factors. These include, among other factors, the type of warning devices in place, the volume and speed of vehicle and train traffic, visibility for train crews and motorists, track geometry, weather conditions, and time of day. These factors vary from crossing to crossing.

The City of Eugene correctly points out that the Florida Study does not examine the FEC accident data with the same level of detail as that contained in Oregon's Railroad Accidents Report. It is reasonable to conclude from the Railroad Accidents Report that factors such as train volume, traffic volume, and the type of crossing protection devices have a greater relationship to safety than does the sounding of the whistle. Nevertheless, the Florida Study does contain a very large amount of data and covers a significant period of time. The comparison to control groups and the statistical accuracy analysis performed lend credibility to the results. The Commission is persuaded that the Florida Study has some relevance to and should be given weight in the determination in this docket.

In its prior order in this docket, the Commission noted that this is a "difficult case." Order No. 89-1037 at 20. The Commission found that "poor track visibility, obstructed signal visibility, and high traffic problems create significant safety problems" and that "train whistles provide a necessary margin of safety in addition to that provided by the crossing signals." Id.

The Commission went on to find that safety improvements related to signal visibility would improve safety, but that high traffic volume in the daytime still tipped the balance in favor of a ban at night because of the increased annoyance to residents and the reduced danger because of reduced traffic volume. The Commission must now revisit this balancing test and take into account the new information from the Florida Study.

The study indicates that the absence of train whistles at night significantly increases the likelihood of accidents, even where crossings are protected with gates, flashing lights, bells, and warning signs. The Commission did not have statistical evidence of this nature in the record when it rendered its prior order. At that time, the evidence in the record did not show any definite relation between train whistles and accidents. Now there is such evidence in the record. The Commission concludes that, because of the specific characteristics of the Eugene section, in particular the limited visibility, train whistles also provide a necessary additional margin of safety at night. The risk to the public is likely to increase at the Eugene crossings at night if the whistle ban is implemented.

The Commission recognizes the annoyance and inconvenience suffered by some Eugene residents as a result of current whistle activity. Their interests are not insignificant and in an appropriate case would warrant protection where noise abatement could be made consistent with safety. On the other hand, Eugene residents, and crossing users in particular, have a strong interest in safe railroad crossings. Here, in light of the new evidence, the level of risk is sufficient, in the Commission's view, to tip the balance in favor of safety.

The Commission's decision in this order should not be interpreted as a determination that nighttime whistle prohibition is never appropriate. As the Commission said in the prior order in this docket:

Each petition for a whistle prohibition order presents a unique combination of safety and environmental factors. The Commission must balance these competing considerations and decide each petition on a case by case basis, based on the facts in the record.

Order No. 89-1037 at 20.

Safety Improvements Previously Ordered

Order No. 89-1037 at 21-22 and Appendix "E." The safety improvements were later modified pursuant to a staff motion. Order No. 90-1195. The Commission is aware of no reason why these improvements should not be made, notwithstanding the rescission of the whistle prohibition order.

CONCLUSIONS

1. Prohibition of routine train whistles at protected crossings in Eugene during nighttime hours will significantly increase the risk of accidents at those crossings.

2. The Commission's prior order in this docket should be rescinded, except as to the safety improvements.

ORDER

IT IS ORDERED that: