The safety net that shrank to fit inside a pie

Dark consultancy-briefing-style pie chart titled “The threadbare safety net”, masthead “Bulwark Resilience Advisors · State of the American Safety Net · 2026 client briefing”, dividing post-disaster support expectations into six slices whose labels sum to exactly 100: friends or family 25.1% in gold, state & local government 18.7% and federal government 18.4% in teals, insurance company 17.0% and nonprofit organizations 11.3% in greens, and faith community 9.5% in slate. A badge reads “25.1% — even friends and family, the strongest thread in the net, are barely a quarter of it; no one else holds a fifth”, the kicker advises “Every layer of the safety net is a minority. Plan as if the net is yours to build”, and a footnote quietly concedes the shares are “by support source” and “sum to 100%”, credited to the FEMA National Household Survey. Every slice is a share of all answers given, not of people. — Showing the misleading chart

The same six answers on a light background as horizontal bars on a 0–100% axis, kicked “The same survey · with the overlap put back” and titled “Most Americans expect help from several directions at once”: friends or family 71%, state & local government 53%, federal government 52%, insurance company 48%, nonprofit organizations 32%, faith community 27%, each bar annotated in red with the pie’s shrunken version (“the pie said 25.1%”). A sum line reads “71 + 53 + 52 + 48 + 32 + 27 = 283% — because one household can (and usually does) count on several”, and a note card headed “How the pie shrank everybody” explains that dividing each answer by the 283 total turned shares of people into shares of answers, shrunk 2.8×, while keeping labels that sum to a tidy 100.0% — so the classic add-up-the-slices tell comes back clean. A credit names FEMA’s 2024 National Household Survey (7,525 respondents, fielded April–June 2024) and notes both charts are our own demonstration. — Showing the misleading chart

Be the first to star this exhibit

FEMA asked Americans who they’d expect support from after a disaster — select all that apply — and 71% said friends or family. Divide every answer by their 283% total and a resilience briefing’s pie can call the strongest thread in the safety net “barely a quarter” of it, with labels that sum to a reassuring 100.0%. The overlap the pie erased was the finding: the average respondent named 2.8 sources.

FEMA, 2024 National Household Survey on Disaster Preparedness — “In the event of a disaster or emergency, who would you expect support from? (Select all that apply)”; 7,525 respondents, fielded April 25–June 7, 2024, 2024Added July 24, 2026

01The claim

America’s disaster safety net is threadbare. Even friends and family — the strongest thread — hold barely a quarter of it, government at every level combined musters just over a third, fewer than one American in five counts their insurer in their corner, and faith communities are down at 9.5%. Every layer of the net is a minority: when disaster strikes, plan as if the net is yours to build.

02The trick

Every number descends from a real FEMA finding, and the pie’s labels sum to a perfect 100.0% — which is exactly the problem. The survey question was select-all-that-apply: 71% of respondents said they’d expect support from friends or family, 53% from state and local government, 52% from the federal government, 48% from their insurance company, 32% from nonprofits, and 27% from a faith community. Those six answers add up to 283%, because the average respondent named 2.8 sources — one household can count on its sister, its insurer and FEMA all at once. A pie cannot hold 283%, so this one quietly divided every answer by 283 before drawing, converting shares of people into shares of answers and shrinking each number 2.8-fold: 71% of Americans became “25.1% of the safety net,” a solid majority expecting federal support became “18.4%,” and nearly half with an insurer in their corner became “17.0%.” Our pizza-toppings pie commits this crime openly — its labels still add to 274% and confess under one glance of arithmetic. This is the same crime after the evidence is destroyed: the normalization launders the impossible sum into a tidy 100.0%, so the classic tell — add up the slices — comes back clean, and the briefing then reads the laundered numbers back as if they were still shares of people. Beneath the arithmetic sits the pie’s structural lie: slices are exclusive camps, so the picture asserts that every American stands in exactly one, when the survey’s actual finding is that the layers stack. (This exhibit is our own demonstration in the house style of a resilience consultancy’s client briefing, drawn from FEMA’s published 2024 National Household Survey findings rather than from any real firm’s deck.)

03The fix

Give each answer back its own bar on a 0–100% axis and every number means what it says again: 71% of Americans expect support from friends or family, 53% and 52% expect their state and federal governments to show up, 48% count on an insurer, and even the pie’s 9.5% “last resort” faith community is really 27% — more than one in four. The bars are free to sum to 283%, and that overlap is the finding, not a flaw: FEMA’s own headline from the same survey is that people expect post-disaster support from a variety of sources — the typical safety net is several layers deep, which is roughly the opposite of threadbare. The tell is the question, not the picture: a pie built on “select all that apply” is impossible by construction, whether or not its labels still show it. So when a survey pie sums to exactly 100, ask what the denominator is — people, or answers? If no slice can be traced back to a share of respondents, the pie has destroyed the one number the survey actually measured. (The honest caveat travels with the honest chart: these are expectations, not deliveries — the survey measures who Americans believe would support them, not who ultimately does.)