A 1976 lunchbox had a peanut butter sandwich in it, wrapped in wax paper, and an apple beside it. Nobody thought about that sandwich for a second. It was just lunch, and every other child in the room had something close to it.
Carry that lunchbox into most American schools today and it does not get through the door. Peanuts are banned outright in a great many of them, tree nuts with them, and in some the list runs on through dairy, eggs and wheat. Somewhere between those two lunchboxes, in a single lifetime, the food a child could safely bring to school became a matter for policy.
This page is everything I found while trying to work out what happened, and it starts with a problem I could not get around.
The video: 70s Americans Didn’t have allergies · the channel
The number that does not exist
Nobody measured food allergy in American children in the 1970s. The first serious American attempt came in 1987, when Bock followed 480 Colorado children from birth and recorded what actually happened when foods were reintroduced. (Pediatrics 1987;79:683.) The earliest usable American peanut figure comes from 1997. Everything anyone tells you about how rare this was when you were at school, including what I told you, is reconstructed from an absence of attention rather than from a count.
The National Academies put it carefully and the exact wording matters: “Although a general perception that food allergy is increasing exists, especially in westernized countries, very few studies support this likely change.” (National Academies, 2016, chapter 3.) They think the increase is likely. They also think almost nobody has demonstrated it.
That does not make the story wrong. It makes the defensible version narrower than the popular one, and the narrower version is still worth your time. At the end of this page there is a section on the papers that went against me, and that is the one to read if you only read one.
What can actually be measured
The figure you will have heard is 1 in 13 American children, which is 7.6%, and it comes from a survey of 51,819 households covering 38,408 children, where parents reported what their children had. Of those children, 61% had at least one allergy a physician had confirmed. (Gupta and colleagues, Pediatrics 2018.) The CDC’s own National Health Interview Survey, which asks specifically about diagnosed food allergy, gives a lower number: 5.8% in 2021, and 5.3% in 2024, which the CDC frames as about 1 in 20. Neither figure comes from feeding children the food and watching what happens, which is the only way to be certain.
If you see “1 in 13, according to the CDC”, the attribution is wrong. It is not a CDC number.
Peanut allergy in children did roughly triple. The same telephone survey run three times found self-reported childhood peanut allergy at 0.4% in 1997, 0.8% in 2002 and 1.4% in 2008. (Sicherer and colleagues, J Allergy Clin Immunol 2010.) Two things sit alongside that. Across all ages, peanut or tree nut allergy was 1.4% in 1997 and 1.4% in 2008, unchanged, so the rise is a children’s finding rather than a population one. And the response rate fell across the series from 67% to 52% to 42%, which matters more than it sounds, because the households that still answer a telephone survey about allergies are disproportionately the households that have one.
Emergency visits climbed steeply. In Illinois, paediatric emergency visits and hospital admissions for food-induced anaphylaxis went from 6.3 to 17.2 per 100,000 children between 2008 and 2012. (Dyer and colleagues, Ann Allergy Asthma Immunol 2015.) Every trend figure of this kind is drawn from billing codes, and the authors say plainly that they cannot separate a real rise in reactions from a change in how reactions get coded. Recognition, epinephrine availability and school policy all changed across the same window.
The hygiene hypothesis, and what became of it
David Strachan published two pages in the BMJ in 1989 showing that hay fever fell as the number of older siblings in a household rose, using the 1958 British birth cohort. (BMJ 1989;299:1259.) He was writing about hay fever and eczema rather than food allergy. He did not use the phrase hygiene hypothesis, which others attached later, but the word hygiene is in his title and the cleanliness idea was his: he pointed at declining family size, better household amenities and higher standards of personal cleanliness reducing cross infection between young children. He also found a weaker trend with younger siblings, which is part of why the older-sibling infection mechanism never held up.
Here is the part I would want if I were watching. The version I gave you, that we made our homes too clean, is the version the field itself has abandoned. The infection mechanism did not survive testing. Specific childhood infections, day care, pets and measles vaccination all gave inconsistent or null results. The researchers who built this area now argue that there is no confirmed evidence of a link between personal or household cleanliness and allergic disease, and that the popular framing does active harm by discouraging hand washing. (Bloomfield and colleagues, Perspectives in Public Health 2016.)
What replaced it is about losing contact with ancestral microbes through diet, antibiotics, birth mode and indoor living, and it goes by Old Friends or the biodiversity hypothesis rather than by hygiene. Writing with Perkin in 2022, Strachan said no mechanism had emerged as robust and described the mechanism behind his own family-size finding as a riddle wrapped in a mystery inside an enigma. The association is solid. What causes it is what nobody has pinned down. (Frontiers in Allergy 2022.)
The peanut advice that ran backwards
In 2000 the American Academy of Pediatrics advised that infants at high risk of allergy should avoid cow’s milk until one, eggs until two, and peanuts, tree nuts and fish until three. (Pediatrics 2000;106:346.) In 2008 it withdrew that advice, concluding there was insufficient evidence to justify delaying any of it beyond four to six months. (Pediatrics 2008;121:183.)
Then the trial. 640 infants with severe eczema, egg allergy, or both, randomised either to eat peanut or to avoid it. Among those who started with a negative skin test, 13.7% of the avoiders had peanut allergy at five years against 1.9% of the eaters. Among those who started with a positive test, 35.3% against 10.6%. (Du Toit and colleagues, NEJM 2015.)
The result is as clean as nutrition research gets. The story built on top of it needs care. The trial studied high-risk infants, the 2000 advice was aimed at high-risk infants, and the advice was withdrawn seven years before the trial reported. So the line running from a guideline to a national epidemic is a plausible narrative rather than a demonstrated one, and I should have said so more plainly than I did.
Birth
American caesarean rates went from about 5.5% of deliveries in 1970 to 16.5% by 1980. (Placek and Taffel, Public Health Reports 1980, for the 1970 figure; Placek and Taffel, AJPH 1987, for 1980.) In the birth certificate data they crossed 20% in the early 1980s, reached 22.8% in 1989, fell back to 20.7% by 1996, and then climbed again to 32.4% in 2024, where they have sat for years. (Births: Final Data for 2024.) Note the dip in the middle. That curve is not the smooth climb people draw alongside a rising allergy line.
Caesarean birth does change which bacteria colonise a newborn, and that is not seriously disputed. In 1,679 samples from 596 British infants, caesarean birth disrupted transmission of the mother’s Bacteroides strains and left hospital-associated organisms in their place. (Shao and colleagues, Nature 2019.) That study measured no allergy outcomes at all, which is worth holding on to, because it is the paper most often cited for a claim it never tested.
Antibiotics
In three large American health plans in 2009 and 2010, children aged between three and twenty-four months were the heaviest users of antibiotics of any age group, at between 0.91 and 1.70 dispensings per child per year depending on the plan. (Vaz and colleagues, Pediatrics 2014.) In the higher-prescribing plans that works out at roughly two to three courses before a second birthday, and rather fewer in the lowest. Those data are now sixteen years old and paediatric prescribing has come down since.
The farms
Two communities in the American midwest, both plain-living, both large families, both raised without television. The Amish farm one family to a farm, largely without machinery, and their children are around animals every day. The Hutterites farm communally and industrially, on a scale that keeps children away from the animals.
Asthma in Amish children runs at 5.2% against 21.3% in Hutterite children. Allergic sensitisation runs at 7.2% against 33.3%. House dust from Amish homes carried 4,399 endotoxin units per square metre against 648, nearly seven times as much. (Stein and colleagues, NEJM 2016.) When the Amish house dust was put into the noses of mice, it suppressed airway reactivity, and that protection vanished in mice lacking the core innate immune signalling machinery, which is the part that turns a correlation into something closer to a cause.
Two things to keep in proportion. Only 30 children from each community were immune-profiled, and the prevalence figures come from earlier community surveys rather than from those 60 children.
A second comparison sharpens it. Amish households showed allergic disease at 26.4% against 46.7% in Old Order Mennonite households. (Tantoco and colleagues, Ann Allergy Asthma Immunol 2018.) Both communities are plain-living and most families in both live on farms. What separates them is scale and machinery: the Amish farms are smaller, with fewer animals, less mechanised, and raw milk is drunk more often. The protection tracks how the farming is done rather than whether the family is plain.
What did not survive
Blood tests in American children did not show the increase. Comparing 7,896 children aged 6 to 19 across two national surveys, food sensitisation measured by IgE came out at 24.3% in 1988 to 1994 and 21.6% in 2005 to 2006, a difference whose confidence intervals overlap. The authors’ conclusion was a null: sensitisation did not increase, contrary to what they expected. The one change that was significant went the other way, with shrimp sensitisation falling from 11.2% to 6.1%. (McGowan and colleagues, J Allergy Clin Immunol Pract 2016.) Sensitisation is not the same as clinical allergy and the window starts in 1988 rather than 1975, so this does not settle the question. It is still the most inconvenient dataset in the field, and it is laboratory measurement rather than a survey.
The antibiotic link mostly disappears between siblings. Across a Swedish national cohort, foetal antibiotic exposure carried a hazard ratio of 1.28 for later asthma. Compare brother with brother inside the same family and it falls to 0.99. For antibiotics given for urinary or skin infections, which have nothing to do with the airway, the population figure of 1.54 falls to 0.85. (Örtqvist and colleagues, BMJ 2014.) The likeliest reading is that families prone to asthma are also families whose babies get prescribed more antibiotics, and that early wheeze is often treated as a chest infection before anyone calls it asthma.
The largest recent caesarean study found nothing. In 74,639 Japanese children, the adjusted odds ratios for wheeze, eczema, asthma and atopic dermatitis were 0.98, 0.99, 0.95 and 1.02. (Scientific Reports 2025.) Two limits: everything was assessed at one year of age, and food allergy was not among the outcomes. Where an effect is found elsewhere it is usually a 20% to 30% relative increase, and in the best-designed study it appeared only in children whose gut bacteria still looked caesarean at their first birthday. (Stokholm and colleagues, Science Translational Medicine 2020.)
The sibling effect that started all of this is heavily confounded. A Finnish total-population analysis of 324,306 children with siblings, comparing children within the same family, found that confounding accounted for as much as 75% of the apparent protection, most of all for asthma and eczema medication. A birth-order effect survived across every outcome they measured, so something real is there. It is a good deal smaller than the raw gradient suggests. (Luukkonen and colleagues, European Journal of Epidemiology 2024.)
And there is still no 1970s number. Nobody counted. That is where this page started and it is where it ends.
The video this comes from: 70s Americans Didn’t have allergies