News your connection to The Boston Globe
Deadly Delays: The Decline of Fire Response
To find your department  

> >


> >

Or browse complete list

About The Boston Globe investigation of fire department response times

This page gives details of The Boston Globe's investigation of fire department response times across the nation, as well as a description of the sources of the public records and the people involved in the project.

The home page for the investigation is
Report cards for individual Massachusetts fire departments can be found by using the drop-down menu at the top of this and every page, or by browsing the complete list.
An interactive map showing locations and details of more than 78,000 building fires in Massachusetts from 1990-2002 can be found here, as well as by clicking the red "See interactive version of this map" button on each fire department page.

Please send any questions to Bill Dedman, at

About the data

The Globe studied building fires reported from 1986 through 2002 across the US, and mapped those for 1990 through 2002 in Massachusetts. Fire departments report the fires to a database called the National Fire Incident Reporting System (NFIRS), which is maintained by the US Fire Administration.

To focus only on true emergencies, the Globe included only fires meeting certain criteria: only building fires that a fire department reported it extinguished, and only fires in the department's territory. That left 3.3 million building fires, reported by 20,000 fire departments.

Statistics Professor Elaine Allen of Babson College in Wellesley looked for the Globe at the question of whether death rates in fires was related to response time, and she found a statistically significant increase in the percentage of fires with a fatality, as response time increased. Her regression analysis controlled for the time of day, day of the week, property loss, whether the fire department was volunteer or full-time, and whether the fire department needed aid from another fire department. Fire deaths are many -- about 4,000 a year in the US -- but they are rare as a percentage of fires, making any statistical analysis of deaths somewhat tentative. Still, Prof. Allen found that as the response time increases by one minute, the probability of a death increases by half a percent. If fires are broken down into two groups -- response time below 8 minutes, or longer -- the increase of probability of a fatality is 4.1 percent in the slower-response fires. The biggest missing variable in such an analysis is the time that elapsed before anyone called the fire department.

Not every fire could be mapped, because of incomplete addresses reported by fire departments -- "behind Fred's" being the most inscrutable -- but the online maps show 89 percent of the building fires studied (78,502 out of 87,750 fires in Massachusetts).

Response times shown here are the elapsed time from the fire department's receipt of the alarm to the arrival at the fire. That response time therefore includes time for a dispatcher to handle the call and alert firefighters, turnout time for firefighters to get en route, and travel time. The National Fire Protection Association's national panel of fire chiefs, firefighters and technical specialists has adopted a standard urging communities with full-time fire departments to adopt a standard of 1 minute for dispatch time, 1 minute for response time, and 4 minutes for drive time. The federal database does not capture each of these times: just the receipt of the alarm and the arrival. The Globe judged every fire by this 6-minute standard. All times were rounded off to the nearest minute in the database.

The NFIRS database is voluntary in most states, so from year to year a different mix of fire departments report fires. This raises a question whether the decline in response times could be caused by a change in fire departments reporting fires -- for example, if Alaska did not report, but then started to report, the long distances in that state could drag down the national picture. To check this possibility, the Globe looked separately at volunteer and full-time departments. Using information from a US Fire Administration census of fire departments, the Globe put departments in five categories, based on the number of full-time and volunteer firefighters: entirely full-time, mostly full-time, mostly volunteer, entirely volunteer, or unknown, because they did not respond to the census. Both volunteer and career fire departments showed a sharp decline in the share of departments meeting an on-time standard of responding to 90 percent of fires within 6 minutes of receiving the alarm. In addition, the newspaper looked at response times only for fire departments that reported fires every year -- the equivalent of a "same-store sales" calculation in retail sales reports. Those departments also showed a downward trend in on-time performance.

Because this is apparently the first time that these public records have been used to study response times -- although state and federal officials have collected the times since the mid-1980s -- data quality issues have not received much scrutiny. Fire departments may vary in how scrupulously they record times. A fire crew could forget to tell a dispatcher that it arrived, so a later crew's arrival would be marked as the arrival time. And state and federal custodians of these records have apparently not performed the checks for data integrity that are common in business information management. For example, fire departments in Dallas, Houston and most other cities in Texas have reported perfection for more than a decade, arriving at fires at the exact moment they were dispatched. (State officials said they have no magic carpets for fire departments in Texas -- it must be a computer error in reporting times from computer-aided dispatch systems.) Other fire departments reported arriving before the alarm. To weed out such faulty times, the Globe excluded any fires with response times of 0 minutes, or 30 minutes or longer.

There is one more complexity in judging the response times. The federal incident reports ask when the fire department received the alarm. Because at many departments the first person to receive the alarm is a fire department dispatcher, the Globe used a 6-minute standard: one minute for dispatch time (from NFPA standard 1221), one minute for turnout time (from standard 1710), and four minutes for drive time (also from 1710). But some departments have told the Globe that they report as the alarm time a later time: when the fire department's dispatcher alerted firefighters. Those departments could be judged by a 5-minute standard, because they start the clock after the dispatching is done. But there is no way to know how many departments fill out the form in this way, so the Globe judged all departments by six minutes. This gives an advantage to departments such as the Boston Fire Department, which said it reports the later time; Boston would pass in either case: it gets to 90 percent of building fires within 5 minutes. To match up the incident reports perfectly with the NFPA standards, the NFIRS database would need to collect more detailed times on each step of the process from 911 call to arrival of each unit. Note that the public NFIRS database does not tell how many firefighters were in that first response; in a volunteer department it could be a single firefighter.

Fire station locations on the maps are the best available current information. The Globe started with an old list of station locations from the state fire marshal, and telephoned and faxed every fire department in the state for updates. Not every department provided information. If you know of a change in a station location, please let us know. Many fire stations on the map are unstaffed.

A file on the main page of has reports for 20,000 fire departments in the US. Not every state is included -- few California fire departments, for example, participate. For a more complete spreadsheet file, with information on fires in each time period, send an e-mail to Bill Dedman, at


Data analysis, mapping and reporting was done by Bill Dedman for the Boston Globe. The project was directed and edited by Mark Morrow, deputy managing editor/projects. The articles were copy edited by Lydia Rebac, Daniel Coleman and Christine Morris. David Butler and Reggie Myers created the graphics and maps. Pages were designed by Dan Zedek and David Schutz. John Tlumacki was the principal photographer. Photos were edited by Paula Nelson, Stephen Haines and Leanne Burden Seidel. Eric Bauer and Joel "Eddie" Medina of designed the online presentation and fire department report cards. Many members of the Globe's regional staff contributed, including David Beard, Doug Belkin, Thomasine Berg, Thomas Coakley, Maria Cramer, Kerry Drohan, Martin Finucane, Dean Inouye, Ric Kahn, Kay Lazar, Jennifer Peter, Mark Pothier, Emily Shartin and Kim Tan. The deputy managing editor/Sunday is Ellen Clegg.

The public records for this report were provided by the US Fire Administration from its National Fire Incident Reporting System. Gayle Kelch and Stanford Stewart of the USFA staff were helpful in understanding the data. Additional records came from the Massachusetts Department of Fire Services, where Fire Marshal Stephen D. Coan, Jennifer Mieth and Derryl Dion helped explain the records.

Richard Mullins helped solve several technical issues with the federal database.

Globe correspondents Todd Morrison and Martha Bartle made hundreds of telephone calls to fire departments to pin down station locations and details of fatal fires.

Vincent Dunn, a former deputy fire chief in New York City and author of several books on firefighter safety, was generous in advising the Globe and reviewing its findings.

Maps of each community in Massachusetts were created using ArcView, from ESRI. The fires were geocoded, or placed on the map, using MapShop for Media, a service of ESRI. Thanks to Kris Goodfellow of ESRI for her help. Additional data analysis, looking at road distances from fire stations, used ESRI's Network Analyst and street information from Tele Atlas. The street information shown on the online maps, however, came from the US Census Bureau.

Then the interactive maps at were created using BeyondGeo, a service of Blue Marble Geographics. Sam Knight of Blue Marble provided technical help.

Dr. Bill Huber of Quantitative Decisions helped with several steps of the mapping and spatial analysis.

Locations for schools, hospitals and geographic features were provided by MassGIS, the state Office of Geographic and Environmental Information, within the Massachusetts Executive Office of Environmental Affairs. Thanks to Michael Trust of that office for help with the data.

Other data came from the US Census Bureau's survey of local government finances, the Massachusetts Department of Revenue, the Federal Communications Commission, the National Fire Protection Association, the Commission on Fire Accreditation International, and the National Institute of Standards and Technology.

Globe Archives Today (free)
Yesterday (free)
Past 30 days
Last 12 months
 Advanced search