The Ten-Year Gap

1The Document With a Date Problem

Somewhere in your station there is a binder. It has a seal on the cover, a resolution number from the board or the council, and a date. The date is the problem. If your community risk assessment and standard of cover were adopted in 2016, they describe a town that no longer exists: its population, its housing stock, its road network, its call load, its hazards, and the roster you assumed would answer the tones.

Nobody decided to let the plan go stale. It happened the way most drift happens, a few percent at a time. Calls went up 4 percent one year. A subdivision opened past the edge of the four-minute map. Three members aged out and two joined. A distribution center went in along the highway. No single change was large enough to trigger a rewrite, and the document had already done the job it was politically required to do: it existed.

A standard of cover is a photograph. A community is a motion picture. The longer the photograph sits on the shelf, the more decisions get made from a picture of the past.

This report does two things. First, it quantifies what ten years actually does to an American community and to the demand curve of the department protecting it, using national data from the Census Bureau, NFPA, the U.S. Fire Administration, and the Energy Information Administration. Second, it gives you a concrete, repeatable method for measuring the drift in your own CAD and RMS data: seven measurements, with thresholds, that tell you whether your deployment plan still describes your community or merely commemorates it.

The pattern that emerges from the national data is consistent: over a decade, the underlying variables of fire service planning move 20 to 50 percent, while the documents governing deployment are commonly written once, adopted once, and reviewed only when an accreditation cycle, a grant application, or a bad night forces the question. We call the distance between the two the ten-year gap.

2What Ten Years Does to a Community

Start with the national ledger. Between mid-2015 and mid-2025 the United States added roughly 20 million people, growing from about 321 million to 341.8 million.^[1] Six percent growth sounds manageable, almost flat. It is not evenly distributed, and fire protection is not delivered to averages. It is delivered to specific road segments, specific structures, and specific people, and all three moved.

2.1Growth concentrates exactly where deployment is thinnest

The fastest-growing places in America are not core cities with established multi-station departments. They are outer-ring suburbs and exurbs, many of them protected by combination or volunteer departments built for a rural past. All five of the fastest-growing U.S. cities over 20,000 population in the latest Census estimates are outer-ring Texas suburbs: Celina (up 24.6 percent in a single year), Fulshear (21.0 percent), Princeton (18.1 percent), Melissa (14.5 percent), and Anna (10.2 percent).^[2] The county pattern is the same: among large metros, the fastest-growing counties sit on the outer edges.^[3]

Stretch the window to a decade and the numbers stop sounding like growth and start sounding like replacement. Princeton, Texas counted 6,807 residents in the 2010 census, 17,027 in 2020, and an estimated 43,524 by mid-2025.^[2] Celina nearly quadrupled in five years, from 16,739 to 64,427.^[2] A risk assessment written for either town at the start of the window is not outdated. It is describing a different municipality that happens to share the name.

2.2The population is aging faster than it is growing

In 2015, 47.8 million Americans were 65 or older, 14.9 percent of the population.^[5] By 2024 that figure was 61.2 million, 18.0 percent.^[4] The 65-plus population grew roughly 28 percent in nine years while the total population grew about 6 percent. Older adults now outnumber children in 11 states and in roughly 45 percent of U.S. counties, up from 3 states and 983 counties in 2020.^[4] By 2030, every baby boomer will be past 65.^[6]

For a fire department this is not a demographic curiosity. Age is the single strongest driver of EMS demand, and EMS is already close to two-thirds of the run load nationally. A community whose 65-plus share moves from 15 to 20 percent has changed its demand profile even if not one new house was built.

2.3The building stock changed shape

The country added roughly 14 million housing units over the decade.^[33] More telling than the count is the composition: 608,000 multifamily units were completed in 2024 alone, the most since 1986, and 54 percent of them were in buildings of 50 or more units.^[7] Many suburbs that were exclusively single-family in 2015 now have their first large residential occupancies: different preplans, different water demands, different aerial needs, different evacuation math. If your SOC's critical-task analysis was built entirely around the two-story single-family dwelling, the housing market has been quietly editing it for you.

2.4The roads got slower and the day got flatter

Americans drove a record 3.28 trillion vehicle-miles in 2024, finally exceeding the pre-pandemic peak and up from a then-record 3.15 trillion in 2015.^[8][9] The average commuter lost 63 hours to congestion in 2024, the highest ever recorded, and congestion is no longer confined to rush hour: it has spread into midday and weekends, pushed by hybrid work schedules and delivery traffic.^[10] Meanwhile the share of people working primarily from home went from 5.7 percent in 2019 to 13.3 percent in 2024.^[11] Your daytime population, the one that was supposedly at the office when your 2016 plan assumed light residential daytime demand, moved back home and stayed there.

Travel-time assumptions are the load-bearing wall of every standard of cover. If the 2016 drive-time model was built on 2016 traffic, every contour on that map is now optimistic, and the optimism is worst at the exact hours your volunteer daytime response is weakest.

2.5Drift runs in both directions

Growth is only half the story. More than half of U.S. counties were smaller in 2020 than in 2010,^[12] and in 2025, 65 percent of counties recorded more deaths than births.^[3] A shrinking community drifts too: a smaller tax base carrying the same station footprint, an older population generating more EMS per capita, a hiring pool that thinned out a decade ago. The ten-year gap is not a growth problem. It is a change problem. The plan assumed a community that held still, and no American community held still.

3What Ten Years Does to Demand

The community is one side of the gap. The demand curve is the other, and it moved faster.

3.1More calls, same fires

In 2015 U.S. fire departments answered about 33.6 million calls, 64 percent of them medical.^[13] The Congressional Fire Services Institute now puts the annual figure at more than 36 million, up from roughly 20 million in 2000.^[14] In the 2024 national run profile, EMS and rescue account for 65.1 percent of runs. Fires account for 3.7 percent.^[15] The national volunteer community puts it plainly: call volume has more than tripled in 40 years.^[18]

Structure fires, the incident type most deployment standards were originally engineered around, have been roughly flat for a decade, averaging around 500,000 per year against a 1980s average near 860,000.^[17] The work that grew is the work most 2016-era standards of cover treat as a footnote: medical calls, service calls, good-intent calls, and false alarms, which together now make up more than 90 percent of the run load.^[15]

Over ten years the workload did not just grow. It changed species, and it changed species faster than the documents describing it.

3.2The people answering the calls

While demand climbed, the response side fell. NFPA counted 814,850 volunteer firefighters in 2015. By 2023 the count was 635,100, the lowest ever recorded.^[18] The steepest single drop came right at the front of the window: 132,250 volunteers lost between 2015 and 2017, with 83,550 of that loss concentrated in communities under 2,500 population.^[19] The remaining force is older: in small communities, the share of firefighters over 50 rose from 15.9 percent in 1987 to 34 percent in 2020.^[34]

Normalize it and the trend is unambiguous: firefighters per 1,000 residents fell from 3.57 in 2010 to 3.14 in 2020.^[20] A standard of cover that assumed a 2015 roster, 2015 daytime availability, and a 2015 average age is carrying an effective response force assumption that no longer clears the audit.

4The Hazards Your 2016 Risk Assessment Has No Category For

Drift in degree is one problem: more calls, slower roads, fewer hands. Drift in kind is worse. Several of today's most consequential hazard classes were not rare in 2016. They were effectively absent, which means an honest 2016 CRA has no row for them at all. A risk you never scored is a risk you never planned for.

4.1Batteries, at three scales

Americans bought about 114,000 plug-in electric vehicles in 2015. In 2024 they bought over 1.5 million, and more than 5.7 million EVs are now on the road.^[21] Every one of them is a different extinguishment problem, a different overhaul problem, and a different tow-and-storage problem than the vehicle fire your tactics assumed.

At grid scale, utility battery storage went from 869 megawatts nationally in 2018 to more than 26 gigawatts by the end of 2024, with another 19.6 gigawatts planned for 2025 alone.^[22] These installations are landing in farm-field townships and small suburbs on interconnection-queue logic, not on fire protection logic. When the Moss Landing facility in California burned in January 2025, the fire consumed a majority of roughly 100,000 lithium-ion batteries, forced more than a thousand residents to evacuate, and reignited a month later.^[23] The departments that will handle the next one are, statistically, departments whose current CRA does not contain the phrase "energy storage."

At the household scale, FDNY documented 104 lithium-ion battery fires in 2021 and 268 in 2023, with 18 deaths that year.^[24] The hazard climbed from novelty to leading fire cause in one budget cycle. It is not a New York problem; New York is simply where the count is kept.

4.2The big-box logistics wave

Since 2020, developers have delivered 1.8 billion square feet of industrial space, more than the entire previous decade, peaking at a record 610 million square feet in 2023 against a prior-decade average of 169 million per year.^[25] The buildings are also bigger: only 6 percent of U.S. warehouse buildings were built in the last ten years, yet they account for 18.6 percent of total inventory by floor area.^[26] If one of these landed on your highway corridor after your CRA was adopted, your maximum-risk occupancy changed, your water supply math changed, and your mutual aid assumptions changed with it.

4.3Occupancies that changed without a building permit

Some of the sharpest risk changes never pulled a permit. The 65-plus population grew 13 percent just since 2020,^[4] and most of that population is aging in place: the single-family house in your first-due is now an unlicensed, unsprinklered assisted-living occupancy of one, with mobility and oxygen concentrator to match. Short-term rentals turned known single-family dwellings into high-turnover transient lodging, with supply growing over 22 percent in the peak year of the boom.^[32] None of these show up on a zoning map. All of them show up at 0300.

5Why Plans Freeze

None of this is a secret. The standards community has been saying "keep it current" for years. CPSE is explicit that credible organizations continuously monitor and manage their evaluation of and response to risk,^[30] and the CFAI accreditation model operationalizes it: performance monitored at least quarterly, performance gaps for the total response area determined at least annually, the governing body formally notified of any gap between capability and the adopted level of service at least annually, and community expectations revisited at least every three years.^[29] Accreditation itself runs on a five-year cycle with an annual compliance report.^[28] NFPA's new consolidated deployment standard, NFPA 1750, published in December 2025, folds NFPA 1201, 1710, 1720, and 1730 into a single organization-and-deployment framework and points to NFPA 1300 for the risk assessment itself.^[27]

So the cadence exists on paper. The problem is who it binds. CPSE crossed 300 accredited agencies for the first time in 2022,^[35] against roughly 29,500 U.S. fire departments.^[18] For everyone else, the CRA and SOC are usually an event, not a process: commissioned for a station bond, a consolidation study, a grant, or a new chief's first year, adopted with a resolution, and then filed. The document was scoped as a deliverable. Nobody was assigned to keep it true.

There is also an incentive problem, and it deserves to be named. A deployment plan reviewed annually generates asks: staffing, apparatus, a station study, uncomfortable conversations with the board about a number that used to be green and is now yellow. A plan on a shelf generates silence. For an elected official, silence is cheaper. For a chief, silence feels like stability right up until the audit, the lawsuit, or the fire that lands past the edge of the old map. Attendance is not accountability, and possession of a document is not planning.

The gap does not form because anyone decides to ignore the data. It forms because no year's drift is big enough to force the question, and the decade's drift is too big to admit.

6Anatomy of the Drift: A Worked Example

To see how the national numbers compound in one place, take a composite township department. Call it Ridgeview. Ridgeview is not a real department; it is built from the national figures in Sections 2 through 4, scaled to a single first-due. The math is real even though the department is not.

In 2016, Ridgeview protected 28,400 people from one station with a combination roster. It ran 3,240 calls, 62 percent EMS. Its adopted standard of cover called for a 9:00 total response time at the 90th percentile, and it was hitting 9:10. The CRA scored the usual suspects: single-family dwellings, a strip mall, a school, a lumber yard, two motels.

Now apply ten years of the drift documented above, at rates the national data says are ordinary for an edge-of-metro township: population growth of 2.2 percent a year, call growth of 4 percent a year, the aging curve, the housing mix shift, and the volunteer decline.

Read the table's last column top to bottom, then read the last row. That is the ten-year gap in one screen: every operational variable moved 20 to 50 percent or spawned an entirely new category, and the adopted performance standard, the number the board thinks it is buying, moved zero.

Now notice the trap in the annual view. Call growth of 4 percent a year is about 130 extra calls: one added run every three days. No trustee notices one run every three days. The 90th percentile slipped roughly nine seconds a year. Nobody rewrites a deployment plan over nine seconds. Each year's drift sits comfortably inside noise, and ten years of noise compounds into a plan that misstates demand by half.

No single year's change was big enough to justify a review. That is not an excuse for the gap. That is the mechanism of the gap.

7Measure the Drift in Your Own Data

The national numbers make the case that drift is real. They say nothing about your drift. That answer is sitting in your CAD, your RMS, and free public data, and extracting it is a few days of work, not a consulting engagement. Below are seven measurements. Each one compares what your adopted CRA and SOC assume against what your data now shows, and each comes with a materiality threshold: the line past which the honest conclusion is that the plan no longer describes the community.

The setup is the same for all seven. Pull ten years of incident data if you have it, five at minimum, with timestamps, incident type, and location. Pull your adopted CRA and SOC and write down, on one page, the specific numbers they assume: baseline call volume, incident mix, performance objective and measured percentile, drive-time coverage, effective response force, and the hazard inventory. That one page is the photograph. The measurements below are the motion picture.

7.1Demand growth

Total calls per year, plotted for every year since the plan was adopted, plus the compound annual growth rate. Compare the latest year against the baseline year written in the plan. Material drift: cumulative change beyond ±15 percent. At the national average of roughly 3 to 4 percent annual growth, an ordinary department crosses this line in four to five years, roughly one accreditation cycle.

7.2Incident mix

Share of runs by major category (EMS, fire, service, good intent, false alarm) now versus the mix table in the plan. Material drift: any major category moving more than 5 points. Mix shift is the quiet one: total volume can look stable while EMS growth and fire decline restructure the workload underneath,^[16] changing unit types, training load, and staffing logic without moving the headline number.

7.3Spatial drift

Map incident locations by year. Compare the demand-weighted center and the top ten demand cells against station locations and the drive-time contours in the plan. New subdivisions, annexations, and corridor development show up here first. Material drift: a sustained demand cluster forming outside the adopted travel-time contour. A cluster your stations cannot reach inside the objective is a deployment decision being made by a developer instead of by the department.

7.4Performance drift

90th percentile total response time, by year, by planning zone. Not the average; averages bury exactly the calls the standard exists to protect. Material drift: the 90th percentile moving more than 60 seconds against baseline, or objective compliance falling more than 5 points. Plot it as a trend line, not a scorecard. A number that misses by a little, every year, in the same direction, is not noise. It is a trajectory.

7.5Concurrency and reliability

Percentage of incidents that begin while another incident is still active, and first-due unit reliability: how often the assigned first-due actually made the run. Material drift: overlap rate up more than half again from baseline, or first-due reliability under 90 percent. Concurrency grows faster than call volume, as the Ridgeview math shows, and it is the variable that turns "we made the number on average" into "the second call waited."

7.6Hazard inventory diff

Take the hazard and occupancy list in the CRA and diff it against the community as it exists today: licensed care facilities, buildings over 50 units, warehouses over 100,000 square feet, EV registrations, battery storage projects in the utility interconnection queue, short-term rental permits, solar farms, new schools. Material drift: any present-day hazard class with a nonzero count and no row in the adopted CRA. This is the only measurement on the list that requires no math at all, and it is routinely the most alarming one.

7.7Response capacity

Responders per incident by hour of day, trended by year, against the effective response force the SOC assumes. Volunteer and combination departments should split weekday daytime from nights and weekends. Material drift: daytime turnout below the SOC's assumed force on more than 10 percent of working fires, or a turnout trend down more than 25 percent from baseline. A standard of cover that assumes fifteen on the fireground is fiction if the data shows seven, and the fiction is worst at 1400 on a Tuesday.

7.8Where the data comes from, and how often to run this

Everything above runs on data you already own or can get free: your CAD and RMS for demand, performance, concurrency, and turnout; Census ACS five-year estimates for population, age, housing, and commuting at the tract level; county GIS and permit records for the hazard diff; state registration data for EVs; the EIA and your utility's interconnection queue for battery storage. As of this year there is also a national yardstick: NERIS replaced NFIRS in January, and its public data feed opened in August 2026, which for the first time lets a department compare its demand profile against live national data.^[31]

Cadence: run all seven measurements once a year, on a calendar trigger, not a crisis trigger. That happens to be the discipline CFAI already demands of accredited agencies, determining performance gaps at least annually and formally reporting them to the governing body.^[29] You do not need the accreditation to steal the habit. One page a year, seven numbers against seven thresholds, presented to the board whether the news is good or bad. Cross a threshold, and the plan gets reopened. That single rule is the difference between a living standard of cover and a commemorative one.

8The Bottom Line

Over roughly a decade, the United States added 20 million people, aged into a country where seniors outnumber children in nearly half its counties, moved its growth to the exurban edge, rebuilt its housing stock around large multifamily, set driving records, and deployed three entirely new classes of battery hazard. Fire service demand rose past 36 million calls while the volunteer force fell to its lowest count on record. Those are not projections. They are the last ten years, measured.

Against that, the typical community risk assessment and standard of cover were written once, adopted once, and aged in a drawer. The gap between the two is not a paperwork deficiency. It is the distance between the service your governing body believes it funds and the service your data says you deliver, and it is the first exhibit in every station bond debate, every staffing request, and every courtroom where a response time becomes evidence.

The gap is also, unlike most problems in this business, cheap to fix. The data exists. The thresholds fit on one page. The cadence is one afternoon a year. Measure the change. Match the service. Defend the number.

§Data Sources and Method

National figures in this report come from U.S. Census Bureau population, housing, and commuting estimates (Vintage 2015 through Vintage 2025; the decade population comparison spans two estimate vintages), NFPA survey data as published in the cited secondary sources, U.S. Fire Administration NFIRS run profiles, NVFC and IAFF workforce compilations, FHWA Traffic Volume Trends, the Texas A&M Transportation Institute Urban Mobility Report, the U.S. Energy Information Administration, Argonne National Laboratory, and industry construction research from Cushman & Wakefield, CBRE, NAHB, and AirDNA. Where an exact national figure is not publishable (NFPA's most recent total-call estimate), the report uses the conservative citable bound and says so. Ridgeview is a composite: its parameters are drawn from the national data and its concurrency figures are modeled as described in the note to Table 4. No proprietary IMT customer data was used.