How NFPA estimates fires and fire losses

More information

The National Estimates Approach to U.S. Fire Statistics (PDF) by John R. Hall Jr. and Beatrice Harwood, Fire Technology, May 1989

How are estimates of specific fire causes calculated?
NFPA analyzes each data element in NFIRS separately. Different data elements capture different aspects of the fire scenario.

Equipment involved in ignition refers to the equipment that provided the heat that started the fire. In many cases, the equipment was operating as designed but human error, such as unattended cooking or leaving something that could catch fire close to equipment, was a factor. Heat source describes the heat that started the fire.  By themselves, some heat sources, such as operating equipment, match, or lighter, do not provide enough information for prevention efforts. The data element of factor contributing to ignition describes how the heat source and item first ignited came together to start a fire. This field includes playing with heat source, equipment unattended, exposure to other fire, and combustible too close to heat source. NFIRS also has a field called “cause of ignition” that groups fires into five categories, intentional, unintentional, failure of equipment or heat source, act of nature, and other or unclassified (intended for exposures).

For structure and building fires only, USFA assigns 1 of 16 cause categories to each fire incident using a hierarchy of definitions based on data found in these data elements. Each cause category is mutually exclusive meaning that each fire incident is assigned one and only one cause code.

How are missing and unknown data handled?
NFPA handles unknown or missing data separately for each data element and allocates them proportionally. This means the percentages of each factor or code are calculated on fires with known information for that data element. The totals (sum of fires with known and unknown data) are multiplied by the percentages of known data associated with each factor or code to create an estimate for each factor. If 8% of fires with known causes were intentionally set, then we assume that 8% of total fires were intentionally set.

The large number of fires with factor contributing to ignition coded as “none” suggested that some fire officers were choosing “none” rather than choosing a specific factor.  Consequently, “none” in that field is treated as unknown. For equipment involved in ignition, a code of “none” in which the heat source does not indicate some type of equipment is also treated as unknown. For example, this would be done when “operating equipment” was entered as the heat source, but the equipment involved was coded as “none.” Incidents are sometimes coded this way when the equipment was operating properly. Also, some people think of equipment as something plugged in, so wiring might not be called equipment.

Heat source has a code choice of “heat from open flame or smoking materials, other.” NFPA treats this as a partial unknown in the open flame or smoking material category, and distributes these fires among the remaining open flame or smoking materials category. This category includes cigarettes, pipes or cigars, heat from undetermined smoking material, matches, lighters, candles warning or road flares, backfires from internal combustion engines. And flames or torches used for lighting. USFA does not treat this heat source code as a partial unknown. Instead, the code is treated simply as its own category as defined by the NFIRS 5.0 Complete Reference Guide: “other open flame or smoking materials.”

Very little information is required for six categories of confined structure fires in NFIRS. How are these handled?
Allocating unknown data is appropriate if we can safely assume that if the unreported (unknown or missing data) were actually known, it would closely resemble the information we already know. Confined fires include: cooking fires confined to the vessel of origin; confined chimney or flue fires; confined fuel burner or boiler fires; confined incinerator fires; confined commercial compactor fires; and trash, rubbish or waste fires in or on a structure that did not extend beyond the structure of origin. NFPA analyzes confined fires separately from non-confined fires. 

Later in the process, results from the analysis of confined and non-confined fires are summed and percentages of total fires are calculated. For most data elements, 10-20% of confined fires have usable data. For home fires, this is enough to draw some general conclusions. Because of the large amount of unreported data, these estimates will be less stable than for non-confined fires. Because of the minor nature of these fires, they are not a major factor in fire losses. Because 80 to 90 percent of the confined fires have unreported (missing) data on the fire module as the completion of the fields are not required for these types of fires, USFA analysts do not use this methodology. USFA considers 80 to 90 percent to be too large to infer that the fires with unreported data should be distributed in the same proportion as those with known information. With confined fires accounting for almost half of all structure fires, NFPA feels that these fires cannot be ignored. 

For equipment involved in ignition, all confined cooking fires are assumed to have cooking equipment involved, while all confined chimney or flue fires and fuel burner or boiler fires were assumed to have involved heating equipment. Incinerators and compactors are self-explanatory. The confined trash fires are not analyzed further for equipment involved in ignition, but they are for other factors. 

USFA’s cause hierarchy makes similar assignments for confined fires involving equipment-specific incident types, but analyzes confined trash fires together with non-confined fires.

What is the difference between home structure fires and residential building fires?
Homes include one- or two family homes (Including manufactured homes), and apartments or other multi-family homes. Residential properties include all homes, as well as hotels and motels, dormitories, residential board and care, rooming houses, and unclassified residential properties. Non-home residential tends to be more heavily regulated than homes. 

Structures include buildings, but also include: open structures, such as bridges, open stairs, and walkways; open platforms, like loading docks without a roof; tents; underground structure work area like tunnels and mines; connective structures like fences, utility poles, and pipelines, and unclassified structure types. Unclassified structure types are the most common. The areas of origin for non-building home fires indicate a substantial portion of these fires occurred in areas we associate with homes. 

USFA defines residential structures and buildings as follows: A structure is a constructed item of which a building is one type. The term residential structure commonly refers to buildings where people live. To coincide with this concept, the definition of a residential structure fire includes only those fires confined to an enclosed building or fixed portable or mobile structure with a residential property use. Such fires are referred to as residential buildings to distinguish these buildings from other structures on residential properties that may include fences, sheds, and other uninhabitable structures. Residential buildings include, but are not limited to one- or two-family dwellings, multifamily dwellings, manufactured housing, boarding houses or residential hotels, commercial hotels, college dormitories, and sorority/fraternity houses.

NFPA’s approach will include some structures that are not buildings and exclude non-home residential properties such has hotels and dormitories. USFA’s approach will exclude some fires that on open stairs or decks and will include non-home residential properties such as hotels and dormitories. 

USFA and NFPA have very different estimates for some fire causes. Why is that?
The main reason has to do with where the causes are pulled from and how unknown data is handled. USFA uses a hierarchy that assigns each fire to just one cause. NFPA analyzes each data element separately, and within each element, confined and non-confined fires are also analyzed separately. Fires that have unknown or missing data are distributed within each data element. Data elements are not mutually exclusive. For example, a child who starts a fire while playing with a cigarette would be captured as both smoking and playing by NFPA. Also, USFA’s hierarchy only treats fires in which the factor contributing to ignition was coded as “none” as unknown if the incident has no other causal information provided.  The same is true for equipment involved in ignition.