Daniel sent us a question that seems simple on its face. How many people around the world lack access to clean, safe drinking water? And is that number going down? But the most commonly cited figure, two point two billion people without safe water, comes from a twenty nineteen WHO UNICEF report. That number is already stale, and the methodology behind it is surprisingly fragile. In March of this year, the UN released its mid-decade SDG six progress report, and the world is not on track. So we've got fresh data and some new controversies to dig into.
The first thing to understand, before we can even answer whether the number is decreasing, is what "access to clean water" actually means in official statistics. Because that's where everything gets slippery.
Of course it does. Define your terms and the crisis either shrinks by a factor of three or balloons into something unmeasurable.
The gold standard definition comes from the Joint Monitoring Programme, the JMP, which is run by WHO and UNICEF. They use something called indicator six point one point one under the Sustainable Development Goals. The metric is "safely managed drinking water." That means three things. The water source has to be improved, so a protected well, a piped connection, a borehole. It has to be located on the premises. And it has to be available when needed and free from fecal and chemical contamination.
And the third one, the contamination piece, is where the whole thing starts to wobble.
Because the JMP doesn't just have a binary "has water or doesn't." They use what's called the ladder framework. The bottom is surface water, rivers, lakes, irrigation ditches. Next is unimproved sources, unprotected wells, tanker trucks from sketchy vendors. Then limited service, which means an improved source but it's more than a thirty minute round trip to fetch it. Then basic service, an improved source within thirty minutes. And at the top, safely managed. The headline number depends entirely on which rung of this ladder you call "access.
When someone says two point two billion people lack access to clean water, they're using the top rung. That's the safely managed number. If you drop down one rung and ask who lacks even basic access, meaning improved water within thirty minutes, the number collapses to about seven hundred million.
And that's the core tension. The two numbers describe fundamentally different problems. The seven hundred and three million without basic access, those are people who need wells drilled, pipes laid, infrastructure built. The additional one point five billion who have basic access but not safely managed, they have a tap or a protected well, but the water might not be reliably safe, or it might be intermittent, or it's shared between multiple households. That's a contamination and reliability problem, not an infrastructure desert.
The crisis is simultaneously enormous and also much narrower than the headline suggests, depending on what you're trying to solve for. It's like going to a doctor and saying "I'm sick" versus specifying whether you have a broken leg or a bacterial infection. Same word, completely different treatment plan.
That's a good way to frame it. And that brings us to the measurement problem. How does the JMP actually count these people? Because you might imagine there's some global sensor network, real-time water quality monitoring. There isn't. They use household surveys. The Demographic and Health Surveys, DHS, the Multiple Indicator Cluster Surveys, MICS, and national census data. Surveys ask a question like, "What is your main source of drinking water?" And people self-report. They say "piped water" or "a borehole" or "the river." That's the proxy.
So it's the honor system for water quality. And a borehole that tested clean three years ago but has been slowly accumulating nitrates from nearby fertilizer runoff, that's still an improved source on paper.
That's exactly the blind spot. And it gets worse. The safely managed tier requires water to be free from fecal and chemical contamination. But most countries don't test water quality in their surveys. The WHO published a report in twenty twenty-one on water quality monitoring, and it found that only forty-five percent of countries have national water quality testing programs. Forty-five percent. The rest use predictive models. They look at sanitation coverage, population density, and they estimate contamination risk statistically.
For more than half the world's countries, the "safely managed" designation is a modeled guess, not a measurement. That's like estimating how many people have malaria by looking at mosquito net distribution data and rainfall patterns, rather than actually testing blood samples. You might get close, but you're going to miss a lot.
Even where testing exists, it's usually at the source, not at the point of use. You test the borehole, it's clean. But the water gets carried home in a jerrycan that hasn't been washed in six months, stored in an open container, dipped with a cup that also touched raw food. The contamination happens between the source and the mouth.
Which is why point-of-use testing often tells a much grimmer story. You mentioned Ethiopia.
Ethiopia's twenty twenty-four DHS survey showed sixty-eight percent basic access. But a Stanford study in twenty twenty-five did E. coli testing at the point of use, in people's homes, and found only forty-one percent of sources were actually free of contamination at the point where someone drinks it. That's a twenty-seven point gap between what the survey says and what the water actually contains.
Twenty-seven points. That's not a margin of error, that's an entirely different reality.
This isn't unique to Ethiopia. It's the norm wherever water quality testing is sparse. The JMP itself acknowledges this. Their reports include a note that safely managed estimates for many countries are based on limited data and modeling assumptions. But that disclaimer doesn't make it into the headlines.
Of course it doesn't. "Two point two billion people lack safely managed water, but that number is partly modeled and we're not totally sure" doesn't fit on a chyron. So let's talk about where the problem is actually concentrated. The urban versus rural split.
Eight out of ten people without even basic water access live in rural areas. But here's what's interesting. The gap is narrowing. Rural access grew from seventy-one percent to eighty-one percent between twenty fifteen and twenty twenty-four. That's real progress. Urban access, meanwhile, stagnated at ninety-six percent. The remaining challenge is increasingly in peri-urban areas and informal settlements.
That urban ninety-six percent number masks something.
Jakarta is the perfect case study. Sixty percent of the city relies on unregulated groundwater. The municipal piped system is classified as an improved source, so those households show up in the statistics as having access. But in many neighborhoods, that piped water is available for maybe four hours a day. People fill buckets when it flows and ration it. The groundwater they're pumping to fill the gap is causing the city to sink, literally, subsidence of up to twenty-five centimeters per year in some areas. And that groundwater is increasingly contaminated with saltwater intrusion and sewage.
The survey says improved access, the reality is four hours of unreliable flow, sinking land, and contaminated backup sources. This is the difference between a ticked box and actual water security. And it makes me wonder, how does the JMP framework even handle intermittency? If the tap exists but it's dry most of the day, does that trigger any kind of downgrade?
It's supposed to, in theory. The "available when needed" criterion in the safely managed definition is meant to catch intermittency. But in practice, survey questions about reliability are notoriously vague. They ask things like "In the last two weeks, was there any time when you did not have sufficient water?" People adapt their expectations downward. If you're used to water coming for two hours a day, and you've structured your whole life around filling containers during that window, you might answer "no, I had sufficient water" because you got what you expected. The survey captures perceived adequacy, not objective reliability.
The metric is measuring satisfaction with an unacceptable baseline. It's the water equivalent of asking someone in a food desert if they had enough to eat, and they say yes because they're comparing themselves to a famine, not to a balanced diet.
And Jakarta isn't an outlier. It's a preview of what happens when cities grow faster than their water infrastructure. The JMP framework wasn't designed to capture intermittency well. If you have a piped connection, you're counted as improved, even if that pipe is dry for twenty hours a day.
Where's the worst of it?
Sub-Saharan Africa accounts for sixty percent of the global population without basic water access. That's the epicenter. But the country-level numbers are surprising. India alone has a hundred and sixty-three million people without basic access. That's more than any country except Nigeria and Ethiopia combined. And India's numbers are improving rapidly, but the absolute scale is still staggering.
China has effectively solved basic access. Ninety-nine percent plus. The challenge there has shifted to groundwater quality. Arsenic, industrial contamination, nitrate pollution from agriculture. They've solved the "is there a tap" problem and are now dealing with the "is what comes out of the tap safe" problem. That's actually the direction the whole world is moving, just at different speeds.
The headline numbers are squishier than they look. But the more important question is the trend. Is the number actually going down?
The answer is yes and also not really, depending on which number you track. Let's start with the good news. The number of people without basic access has fallen from one point one billion in the year two thousand to seven hundred and three million in twenty twenty-four. That's a thirty-six percent reduction in absolute terms. And because the global population grew over that same period, the percentage of the world without basic access dropped from eighteen percent to eight point six percent. That's a genuine, substantial improvement.
On basic access, the world is winning. Slowly, but winning.
Now the safely managed story. The number of people with safely managed access has increased from four point two billion to five point eight billion since twenty fifteen. That's one point six billion people gaining top-tier access in a decade. But the number of people lacking safely managed water has barely budged. Two point two billion in twenty fifteen, two point one billion in twenty twenty-four. A hundred million person improvement over nine years. The percentage dropped from thirty percent to twenty-six percent.
Why does the absolute number barely move when one point six billion people gained access?
Population growth in the countries with the lowest access levels ate the gains. Sub-Saharan Africa added roughly three hundred million people during that period. Most of them were born into communities without safely managed water. So you're running just to stay in place. The tap is being installed, but the population is growing faster than the plumbing.
Running up the down escalator. And the SDG target?
The twenty twenty-six UN progress report on SDG six is sobering. At current rates, universal basic access will be achieved by twenty forty-three, which is thirteen years late. For safely managed water, the target year is twenty thirty, but the projected achievement date is twenty fifty-five at best. And that's if nothing goes wrong, no climate shocks, no funding collapses. The gap is widening for the poorest quintile. The people who need the most progress are seeing the least.
Twenty fifty-five. A full quarter century after the target. So the SDG six point one goal is effectively dead. Not delayed, dead. Nobody's going to say that out loud at the UN, but the math doesn't lie.
The funding picture makes it worse. The World Bank's twenty twenty-five GLAAS report, that's the Global Analysis and Assessment of Sanitation and Drinking-Water, found that countries with the lowest access levels spend the least per capita on water. Often under five dollars per person per year. The estimated cost to meet SDG six point one is about a hundred and fourteen dollars per person per year in rural areas. The funding gap is a hundred and fourteen billion dollars annually, and only twenty-six billion is currently flowing.
The places that need the money most are spending five dollars per person, and the target is a hundred and fourteen. That's not a gap, that's a chasm. And I'm guessing the twenty-six billion that's flowing isn't all going to the places with the lowest access.
It's not. A lot of water funding is politically allocated, not need-based. Middle-income countries with functioning bureaucracies absorb a disproportionate share because they can write grant proposals and navigate World Bank procurement processes. The Central African Republic doesn't have a water ministry with a hundred staffers who know how to apply for Green Climate Fund financing.
The administrative capacity to absorb aid is inversely correlated with the need for aid. That's the tragedy of a lot of development funding, not just water. It's like a food bank that only gives groceries to people who can fill out a twelve-page application in triplicate. The people who are hungriest are the least equipped to navigate the bureaucracy. But water has an extra complication.
Groundwater depletion is the sleeper crisis. The twenty twenty-five NASA GRACE-FO satellite data showed the Indus Basin aquifer, which underlies India and Pakistan, losing eight cubic kilometers of water per year. Eight cubic kilometers. That's a volume of water roughly the size of a small Great Lake, vanishing annually. As shallow wells dry up, rural communities fall back to surface water, rivers, ponds. That's a reversal. People who had improved access lose it because the aquifer beneath them is being mined faster than it recharges.
Surface water means waterborne disease, means women and children spending hours fetching water again. Climate doesn't just threaten future progress, it's actively reversing past gains.
The twenty twenty-four drought in the Horn of Africa pushed twelve million people back to unimproved sources. Twelve million people in a single year. That's equivalent to wiping out a decade of access gains in that region. And these droughts are becoming more frequent. The IPCC's sixth assessment projected that what used to be a once-in-a-generation drought in East Africa will be a once-per-decade event by twenty thirty.
The trend line isn't a smooth upward slope. It's a sawtooth. Gains during normal years, sharp reversals during drought years. And the droughts are accelerating. It's like trying to fill a bathtub while someone periodically pulls the drain plug.
Which brings me to the technology angle, because there's a seductive narrative that point-of-use treatment, chlorine tablets, ceramic filters, UV pens, can bridge the remaining gap without needing to build massive infrastructure. And there's some truth to it. A twenty twenty-three Cochrane review, the gold standard for evidence synthesis in public health, found that household water treatment reduces diarrhea by fifty-two percent when used consistently.
When used consistently. That's doing a lot of work.
The adherence problem is the whole ballgame. The same Cochrane review found that consistent use drops to about thirty percent after six months. People start with good intentions, they use the chlorine tablets, they boil the water. Six months in, the tablets run out and the next batch isn't free, or the boiling fuel is needed for cooking instead, or the treated water tastes slightly different and family members complain. The last mile problem in water isn't technological, it's behavioral.
Like adopting a feral cat. The technology exists. The cat is right there. But consistent long-term engagement is the actual challenge.
That's a weirdly apt metaphor, and I want to push on it. With a feral cat, you're fighting against the animal's instincts and past experience. With water treatment, you're fighting against habit, convenience, and sometimes taste preferences. What's the behavioral economics equivalent of a can of tuna?
I think it's making the default option the safe option. If the water coming out of the tap is already treated, you don't have to remember to add chlorine tablets. The behavioral intervention isn't convincing people to do something, it's making the safe choice the path of least resistance. But that requires infrastructure, which brings us back to the funding problem.
Bangladesh's arsenic mitigation program is the canonical example of this gap between intervention and adoption. In the nineteen nineties, Bangladesh discovered that millions of its tube wells, which had been drilled to solve the surface water disease problem, were contaminated with naturally occurring arsenic. A massive testing program ensued. Twenty million wells tested. One point five million painted red as contaminated. The solution was to switch to green wells or use deeper aquifers. But a follow-up study found that ten years later, forty percent of red-tagged wells were still in use because the nearest green well was too far away, or the deeper well water tasted different, or people simply didn't believe the test results.
Twenty million wells tested, an enormous logistical achievement, and a decade later nearly half the contaminated wells are still being drunk from because the alternative was inconvenient. That's not a failure of technology or even of information. That's a failure of the behavioral and economic assumptions baked into the intervention. They assumed that if you tell people the water is poisoned, they'll walk an extra half mile to the safe well. But walking an extra half mile every day, carrying water, is a real cost. And if the poisoned water doesn't make you visibly sick today, the immediate cost of the walk outweighs the long-term risk of the arsenic.
It's a temporal discounting problem. The arsenic might give you cancer in twenty years. The walk to the green well costs you thirty minutes right now. Humans are terrible at trading off present costs against future risks, especially when the present cost is concrete and the future risk is probabilistic. And it's the same dynamic that makes maintenance such a critical leverage point. A twenty twenty-five World Bank study found that thirty percent of rural water systems in sub-Saharan Africa are non-functional within five years of installation. Not thirty years, five years. A handpump breaks, the spare part isn't available locally, the village doesn't have the funds to send someone to the district capital to buy it, and the pump sits broken. Everyone goes back to the river.
The highest return on investment isn't drilling new boreholes. It's keeping the existing ones working.
The World Bank analysis explicitly says that spending on maintenance and local supply chains has a higher ROI than new construction. But maintenance is boring. Donors want to cut ribbons on new wells. There's no photo op for "we fixed the gasket on a pump installed in twenty eighteen." The incentive structure of development funding is biased toward construction over maintenance.
That's the governance problem. Nobody wants to fund the unglamorous work of keeping things running. It's the infrastructure equivalent of how news media covers plane crashes but not the millions of flights that land safely. The drama is in the new thing, not in the thing that quietly works. Meanwhile, Cape Town almost ran out of water entirely.
Cape Town's Day Zero crisis in twenty eighteen is a fascinating counterpoint to the infrastructure-first narrative. The city was weeks away from turning off the taps, the first major city in the world to face that prospect. And they averted it not by building a desalination plant or drilling new boreholes, but through demand-side management. Water restrictions, public pressure campaigns, pressure reduction in the pipe network. Water demand dropped fifty percent in three years.
Behavioral change and pressure management. No new supply, just using less of what they had. That's the demand-side story that gets overshadowed by the shiny infrastructure projects. And what's striking about Cape Town is that they didn't just nudge people with polite suggestions. They published a map of household water consumption by neighborhood. Public shaming as a water conservation tool. It was extraordinarily effective and also deeply uncomfortable as a governance model.
The map was controversial, but it worked. People who were using more water than their neighbors got phone calls from the city. There was a hotline where you could report your neighbor for watering their lawn. It was a wartime mobilization mentality applied to municipal water. And contrast that with Singapore's NEWater system. Singapore took the opposite approach, massive investment in membrane technology, reverse osmosis, UV disinfection, turning wastewater into drinking-quality water. It works brilliantly for a wealthy city-state with a technocratic government and the ability to invest billions. But that model doesn't transfer to rural Ethiopia or the peri-urban slums of Dhaka. There's no one right answer. The solution has to match the governance capacity and the economic reality of the specific place.
We've got a measurement problem, a funding misalignment, a maintenance crisis, climate reversals, and a behavioral adoption gap. The actual number of people without clean water is somewhere between seven hundred million and two point two billion, depending on definitions. The trend is positive for basic access, barely moving for safely managed, and fragile everywhere. Where does that leave someone who encounters a water access statistic in their work or reading?
I'd say there are three questions to ask whenever you see a water access number. First, what tier of the JMP ladder does this refer to? Is it safely managed, basic, or something else? The difference between two point two billion and seven hundred million is entirely about which rung you're standing on. Second, was water quality tested or modeled? If a country doesn't have a national testing program, the number is an estimate built on assumptions. Treat it as directional, not precise. Third, is this point-source or point-of-use? The borehole might be clean, but what's in the cup? Those three questions will tell you how much to trust the statistic.
That's a useful mental toolkit. And the corollary is that any number that doesn't specify the tier is probably useless. A headline that says "X million people lack water" without defining what "lack water" means is noise, not signal.
The leverage point that I think is most underappreciated is maintenance. The thirty percent non-functionality rate in sub-Saharan Africa is a scandal hiding in plain sight. We've already paid for those systems. The concrete is poured, the pump is installed. And a third of them are broken within five years. Fixing what we already built is the highest-impact, lowest-cost intervention available, and it's systematically underfunded because it's not newsworthy.
It's the infrastructure equivalent of preventive medicine. Everyone agrees it's more cost-effective than emergency surgery, but the funding goes to the emergency room because that's where the drama is.
The climate piece makes this urgent in a way that didn't exist when the SDGs were drafted in twenty fifteen. The Indus Basin losing eight cubic kilometers a year, the Horn of Africa drought pushing twelve million people back to surface water, these aren't projections. They're measurements from the last two years. The next decade is going to determine whether SDG six point one is merely delayed or completely abandoned. If we get a series of severe drought years in sub-Saharan Africa and South Asia, the trend lines could flatline or reverse.
Which brings us to the open question that I think defines the next era of water policy. The JMP is currently redesigning its monitoring framework for the post-twenty-thirty period. There's a debate about whether the safely managed metric should move from a binary classification, safe or unsafe, to something risk-based. A probability of contamination over time, accounting for seasonal variation, intermittency, and climate vulnerability. That would change the headline number dramatically, and probably make it larger.
Because a probabilistic metric captures what the binary metric hides. A water source that's clean in the dry season but contaminated during floods would show up as risky year-round, not as safe with occasional problems. It's a more honest representation of the lived experience of water insecurity. But it also makes the problem look worse, which is politically uncomfortable. Governments that are currently reporting steady progress would suddenly see their numbers drop, not because conditions changed, but because the ruler changed.
That political discomfort is precisely why the debate is happening now, years before the post-twenty-thirty framework takes effect. Nobody wants to be the minister who announces that water access in their country just fell by fifteen percentage points because of a methodological change. The politics of measurement are as fraught as the measurement itself.
There's a technological dimension to this. Satellite-based water quality monitoring is becoming viable. NASA's ECOSTRESS, the ESA's Sentinel-2, they can detect algal blooms and sediment plumes from orbit. By twenty thirty, we may have near-real-time global water quality data that makes the current survey-based approach look as archaic as measuring poverty by asking people if they feel poor.
Which would be a revolution in accountability. Right now, a government can report that eighty percent of its population has safely managed water, and there's no independent way to verify that. Satellite monitoring changes the game. But it also raises privacy and sovereignty questions. Who gets to monitor whose water? And what happens when the satellite data contradicts the government's official statistics?
We saw a preview of this with air quality monitoring. When satellite data on PM two point five pollution started contradicting official ground-station numbers in various countries, it created a diplomatic friction that's still unresolved. Water is arguably more sensitive because it's not just about public health, it's about agricultural capacity and national security. A satellite that can detect groundwater depletion is also a satellite that can identify which countries are mining their aquifers unsustainably.
Which means the data exists, or will exist soon, but whether it gets incorporated into official statistics is a political question, not a technical one. The number of people without clean water is decreasing, but the rate of decrease is slowing. The easy progress, drilling wells in accessible areas, connecting urban households to existing piped networks, has largely been made. What remains is harder. It's about maintenance, governance, behavioral change, and climate resilience. The infrastructure that's already in the ground is decaying faster than it's being replaced. The aquifers that supply a third of the world's agriculture are depleting. And the funding system is structured to reward construction over everything else.
The two point two billion number is a useful rallying cry. It gets attention, it mobilizes resources, it makes the problem feel urgent. But as a policy tool, it's nearly useless. The seven hundred and three million without basic access, that's the number that tells you where to put pipes and pumps. The rest of the gap is about water safety, reliability, and behavior. Different problems, different solutions.
The next time someone drops a water access statistic on you, ask which rung of the ladder, whether the water was tested or modeled, and whether the measurement is at the source or at the lip of the cup. The answers will tell you more than the number ever could.
Now: Hilbert's daily fun fact.
Hilbert: In the eighteen sixties, the term "flenser" entered Arctic whaling parlance to describe the men who stripped blubber from whale carcasses on the shores of Nunavut. The word derives from a Danish verb meaning "to slice into strips," and the job was considered so gruesome that flensers were paid in whisky as well as wages.
Paid in whisky to strip blubber in the Arctic. That's a compensation package.
I have so many follow-up questions that I'm going to suppress for the sake of the closing.
This has been My Weird Prompts, with thanks to our producer Hilbert Flumingtop. You can find every episode at myweirdprompts dot com, and if you enjoyed this one, leave us a review wherever you listen. We'll be back with a new prompt next time.
