The Cheapest Thing That Works

On the first of January this year the Irish state signed up more than seven hundred and sixty thousand people to a pension without asking them, a number that has since passed eight hundred and thirty-five thousand, or would have if every employer had registered: six thousand six hundred still had not by March. Everyone aged between twenty-three and sixty, earning over twenty thousand euros, without a workplace pension already, was simply enrolled. Money began leaving their pay in the same month.

They were given a way out. From the first of July a window opened in which anyone could leave and get their own contributions back, and it closes at the end of this month. By the eighth of July, a week in, the agency running it had recorded 4,997 opt-outs. That is about six in every thousand enrolled. Seven weeks of the window were still to run at that point, and it closes at the end of this month, so the final figure will be higher. What it will not be is anywhere near the number who would have joined had they been asked.

Now look at Poland, which did the same thing seven years earlier. The PPK scheme auto-enrols employees, the employer pays in, the state adds a top-up, and the money is the worker's. Seven years on, participation stands at 61.43 per cent on the scheme operator's own monthly numbers, on the scheme's own flattering count of who is in. Roughly two Polish workers in five are not in it, though not all of them refused: the scheme auto-enrols only up to age fifty-five and older workers have to ask, so part of that two in five was never handed anything to turn down.

Inside Poland the pattern is sharper than the national figure. In firms with more than a thousand employees, participation runs at 90.53 per cent. In firms of ten to forty-nine, 33.49. Same law, same offer, same year, fifty-seven points apart. The Polish public sector, which nobody would call feckless, sits at 35.25 per cent against the private sector's 70.61, so this is not a simple story about poor people either.

I can find nothing about Irish character that explains the first number and nothing about Polish character that explains the second. Part of what differs is what was set up around the person: how the enrolment was run, who did the paperwork, whether the human resources department in a building of two thousand people had a process, whether a workshop of twenty had anybody whose job that was. Only part, and my own figures say so. Polish firms under ten read 74.65 per cent, which breaks the size story, and the public sector, which has more process than almost anyone, sits near the bottom of it. What a Polish worker expects the state to do with a pension fund over twenty years is doing work here that no human resources department can undo.

I think the same gap explains a great deal else. Everyone knows how they are supposed to live and almost nobody lives that way, and the first explanation people reach for, that they do not know, collapses on contact with the data, and the second, that they do not care, is doing far less work than it looks like. What is actually unequal is more interesting than either, and I did not expect it to be spread across Europe as unevenly as it turns out to be.

Nobody in this story is ignorant

If information were the missing ingredient, the people who got it from a cardiologist standing over their own hospital bed would be the ones who changed.

The EUROASPIRE surveys interview people across Europe who have already had a heart attack or a coronary procedure. Among 8,261 of them across twenty-seven countries, nineteen per cent were still smoking, and among those who had smoked before the event, fifty-five per cent were smoking after it. Two thirds were not managing thirty minutes of activity five times a week. Thirty-eight per cent were obese. I cannot think of a group less in need of a leaflet. Their cardiologist has already told them, in a hospital bed, with a scar to point at.

The general picture matches. Around twenty-four per cent of EU adults smoke and another twenty per cent used to and stopped, so quitting is manifestly possible and enormously common. Just over half of Europe's current smokers have already tried and failed at least once. Asked what they used in the past year to help, sixty-four per cent named nothing at all from a list of every available aid, and three per cent had used medical support or a stop-smoking service.

Exercise is worse and simpler. In the last European health survey to ask, in 2019, nearly half of EU adults, 48.4 per cent, did no leisure-time aerobic activity in a typical week, and only 13.6 per cent met both of the WHO's recommendations. Asked plainly in a separate survey, forty-five per cent of Europeans say they never exercise or play sport, up from thirty-nine per cent in 2009 though a point down since 2017. On food, twelve per cent of Europeans manage five portions of fruit and vegetables a day.

Even where the stakes are immediate and the instruction is medical, the gap holds: pooled across forty-four studies and nearly two million people worldwide, sixty per cent took at least eighty per cent of their prescribed cardiovascular medication.

Doctors are the control group for the knowledge theory, and the comparison that counts is not doctors against everybody, since doctors are graduates on graduate incomes. A meta-analysis of eighty-nine studies covering twenty thousand European physicians found smoking prevalence of twenty-five per cent, the joint highest of any world region, although that pools decades and the rate falls to sixteen per cent in studies after 2015. Knowing exactly what a cigarette does to a lung moves the number. It does not close the gap.

What the cheap thing is for

The mistake I keep finding in advice, including advice I have given, is that it treats the bad option as an absence of the good one. It is a product, it is competitively priced, and it is doing a job that nothing on the recommended list even bids for.

At nine in the evening the job is to end the day. The cigarette, the drink, the delivery order and the twenty minutes that turn into ninety all perform it reliably, immediately, at a price payable tonight. The recommended alternative does not solve tonight. It solves a statistical problem belonging to a person you have not become yet and may not become.

What the cheap thing is cheap in is not money. It costs no planning, no equipment, no spare hour, no rearranged evening and no second attempt, and it works the first time. That is the currency the evening is short of, and Europe holds it very unequally.

Start with time, since everyone assumes they know this one. In the eleven countries covered by Eurostat's last time-use round, people in full-time work record between 2 hours 58 minutes of leisure a day in Bulgaria and 5 hours 2 minutes in Germany. Couples with a child at home give up between twenty-two minutes and an hour and forty-seven of that, and inside those couples the women give up more than the men. Shift work, which wrecks sleep and mealtimes together, covers 17.4 per cent of EU employees, from 34.9 per cent in Romania to about eleven in Denmark.

Then money, where the spread is worse than most people think. The cheapest basket meeting national dietary guidelines costs between €2.33 and €3.23 per person per day across the euro area, which is seventy to ninety-eight euros a month for one person before anything else in a life. In 2025, 8.5 per cent of the EU population could not afford a meal with meat, fish or a vegetarian equivalent every second day, rising to 20.3 per cent among people at risk of poverty, and running from 1.2 per cent in Cyprus to 18.1 in Romania. 8.8 per cent could not keep their home adequately warm.

And credit, which is where the trap closes. In June the euro-area average rate on a new consumer loan was 7.50 per cent against 3.51 on a mortgage. The small unsecured loan that a household in trouble actually needs costs more than double the large secured one that a comfortable household uses, and where you live decides how much more: from 4.05 per cent in Malta to 12.89 in Estonia, and that is bank lending, before the instalment shops and small-loan firms where the real cost of being short lives.

Behind all of it sits the number I keep coming back to. Across the EU, 29.2 per cent of people live in a household that could not meet an unexpected required expense out of its own resources: 61.2 per cent of those below the poverty line, and 23 per cent of everybody else. It runs from 15.3 per cent in the Netherlands to 50.5 in Greece, and it is not a poor-country phenomenon, since Germany reads 31.9. Ask what "an unexpected expense" means and the survey turns out to ask each country about a different sum, calibrated to its own poverty line, so on Eurostat's own rule the question clears something over three hundred euros in Bulgaria and something near two thousand four hundred in Luxembourg.

I wrote the supply side of this argument earlier this month: lower the cost of an action and the action becomes more frequent, and that law does not care whether the action is betting or saving. This is the demand side. I think the cost of an action is not a property of the action. It is paid out of a reserve, and I did not understand mine until I watched a bad month empty it. Across Europe that reserve is distributed worse than income is.

The money nobody collects

Non-take-up looks at first like the best case against everything I am about to argue. Here is money with a person's name on it, sitting on a table, and it goes uncollected. Then look at what collecting it takes.

Non-take-up is, to my mind, the largest unclaimed sum in European social policy, and almost nobody writes about it. The OECD, citing Belgian research, puts non-take-up of working-age social assistance at between 37 and 51 per cent. About a third of French households entitled to the RSA minimum income do not receive it in a given quarter, at an average loss of three hundred and thirty euros a month. Among single French people over sixty-five entitled to the minimum old-age pension, only about half claim it, and the gap is wider for women than men.

Eurofound gathered national estimates for sixteen member states and found every one of them had at least one benefit whose non-take-up reached forty per cent or more. The estimates are old and use incompatible methods, so they cannot be ranked against each other, but the direction is unanimous.

Hold that next to Ireland's five thousand opt-outs, because the pairing is what convinced me. France ran both halves of this on one benefit. The chèque énergie reached five and a half million households automatically; when a tax reform left roughly a million more, on a consumer institute's estimate, that the state could no longer identify, it opened an online claim for them, and about a hundred thousand claimed. The government's own verdict was that the procedure did not answer the problem. Ireland is the same shape from the other side: when refusing takes a form, one person in a hundred and sixty fills it in. The two populations are not each other's twins, since Ireland's twenty-thousand-euro floor excludes the very workers the take-up studies count. What repeats across both is the form. The behaviour is identical in both cases: people do not do the paperwork. Only the default changed sides.

What refuses to fit

Several of the numbers refuse to line up by income, and the ones that refuse hardest are the ones I would most have liked to have.

Binge drinking is the clearest refusal. Among Europeans who drink, monthly heavy drinking is lowest among the least educated: 12.7 per cent for those with lower-secondary schooling or less, against 22.3 for the upper-secondary group and 20.2 for graduates. In Poland the graduate is twice as likely to binge monthly as the early school-leaver, 17.6 against 8.5, and the same holds in Hungary, Czechia and Croatia. By income the line runs the wrong way too, from 16.4 per cent in the poorest fifth to 21.3 in the richest; in Ireland the richest fifth binges twice as often as the poorest.

Obesity gives me less than I would like as well. The EU gradient is real but modest, twenty per cent of low-educated adults against eleven per cent of graduates, and for men it thins to four and a half points and inverts outright in Estonia, Latvia, Romania, Lithuania and Bulgaria.

Then the finding that troubled me most, and that should trouble anyone who thinks money is the mechanism. Sweden hands out large lottery prizes, and because the winners are effectively random among players you can read the result as an experiment. Following the whole Swedish population through administrative registers, researchers found no evidence that wealth affects adult mortality or use of health care, with estimates precise enough to rule out effects a sixth the size of the observed wealth-mortality gradient. Their own conclusion is that in affluent countries with extensive safety nets, the causal effect of wealth is not a major source of the health-wealth gradient. Handing people money does not make them live longer.

Education reform does not obviously do it either. Nineteen compulsory schooling reforms across twelve European jurisdictions produced a pooled 2.8 per cent reduction in men's twenty-year mortality and nothing coherent at all for women.

Nor does redistribution, at least not simply. The most redistributive states in Europe, the Nordic ones, do not have the smallest health inequalities: across twenty countries, relative mortality inequality among men ran from 1.50 in Spain to 2.89 in Lithuania, with the Nordics sitting above the western European norm and Spain and Italy unexpectedly low. England ran the first systematic national strategy to close its own gap and, after thirteen years, missed its own targets. A larger transfer budget, aimed directly at this, did not dissolve it. The Nordic welfare states, the most redistributive in Europe, do not have the smallest health inequalities: across twenty European countries, relative mortality inequality ran from 1.50 in Spain to 2.89 in Lithuania among men, with the Nordics standing out as larger than the rest of western Europe and Spain and Italy unexpectedly small. England ran the first systematic national strategy to close the gap and, after thirteen years, missed its own targets.

And behaviour does more work than my argument would like. In the Whitehall II civil-service cohort, adjusting for health behaviours measured repeatedly over the follow-up attenuated the link between employment grade and death by 72 per cent, with smoking alone about a third of it. The same team then ran both cohorts side by side in one paper and got 75 per cent for Whitehall against 19 per cent for the French. So how much of a health gap is "behaviour" is not even a fixed quantity across two European countries.

Finally, the person is not empty. In a British study of five hundred and nine fraternal twin pairs, the twin with worse self-control at five was, by twelve, more likely than their own co-twin to have started smoking and to be doing badly at school. They were raised in the same house on the same money by the same two people. Something person-shaped is doing real work, and I have no way to argue it out of the data.

What the same drink costs

Everything above is a subtraction. What is left is about price rather than appetite.

Take two people who drink identically and change nothing else about them except where they live. In a Scottish cohort of fifty thousand adults, In a Scottish cohort of fifty thousand adults, both numbers are measured against one reference group, light drinkers in advantaged areas. Heavy drinkers in those same advantaged areas ran 6.12 times their risk of alcohol harm. Heavy drinkers in deprived areas ran 10.22 times it. That is after adjusting for how much they drank, how often they binged, their weight and their smoking, all of it self-reported at a single interview. The review of this literature says plainly that nobody has tested whether poorer drinkers under-report more, and if they do, part of the extra bill is really extra drink. That objection stays open. What survives it is the direction of travel: the researchers most determined to explain the gap away are still left holding most of it.

Europe ran the experiment by accident, and not cleanly: the same spring also opened the borders to cheap imported drink. In March 2004 Finland cut alcohol taxes by about a third. Alcohol-related mortality rose sixteen per cent among men and thirty-one among women nationally, but the increase was not shared. Among men over thirty with only basic education it rose by 39.9 deaths per hundred thousand; among the university-educated, by 4.8. Among long-term unemployed men aged thirty to fifty-nine it rose by 166.1 per hundred thousand, a jump of twenty-one per cent, against 1.7 per hundred thousand and under three per cent for employed men. A price cut also changes how much people buy, and the poorest respond hardest to price, so read this as evidence about where the bill arrives when a country makes a cheap thing cheaper, rather than about an unchanged dose. One price change, one country, one moment, and the damage sorted itself by circumstance on the way down.

So the claim has to be narrower than I wanted it. Across Europe the poor demonstrably do not binge more. Smoking is the exception and it is the one that kills most, at 18.6 per cent daily among the least educated against 12.7 among graduates, and smoking is about a third of the Whitehall attenuation. Where the behaviour is the same, the same dose costs them more. Where it is not the same, it is the behaviour whose price a state sets by tax and by what it does to help people stop.

The outcome data across Europe reads to me exactly like that, with one thing said out loud first. Illness pushes people down the income scale as well as the other way round, and Insee does not claim to have separated the two: a man in the bottom five per cent of a single tax year may be there because he is already dying. That eats into these numbers. It does not eat thirteen years. In France the richest five per cent of men outlive the poorest five per cent by thirteen years, and the curve is steepest at the bottom: around twelve hundred euros a month, an extra hundred euros is associated with roughly another year of life, while at three thousand it buys about two months. In the Netherlands the education gap in life expectancy is about six years, but the gap in years lived in self-rated good health is nearly nineteen. Across the EU, healthy life years at birth run from 51.2 years for Latvian men to 71.7 for Maltese men, a twenty-year spread inside one internal market, and Eurostat notes the country differences are much wider for healthy years than for years as such.

Deaths that treatment should have prevented follow the same map: 51.7 per hundred thousand in Luxembourg against 199.7 in Romania. And the smaller, meaner version shows up in what people skip: going without medical care because it costs too much is reported by 2.1 per cent of the poorest fifth and 0.2 per cent of the richest.

An option you cannot afford to use is not really an option, which is the argument I made about courts and which turns out to be truer about lungs.

Defaults, and nothing else

Every technique this industry sells shrinks when somebody counts the studies that did not work, and I have watched the best-known one shrink furthest.

The best-known technique in behaviour change is the implementation intention, the if-then plan. A 2006 meta-analysis of ninety-four tests put it at d = 0.65, and that number has been on motivational slides ever since. In 2025 the same authors returned with six hundred and forty-two tests, unpublished ones included, and got 0.36, which they call small and highly heterogeneous; for behaviour change specifically, 0.27; and correcting for the bias of a literature that publishes what works, about 0.15. It works. It works about a quarter as well as it is sold.

Nudges took the same haircut: a reanalysis moved the pooled estimate from d = 0.43 to d = 0.04, with the informational ones, the kind that consist of telling people things, landing at zero. Defaults are nudges too, and I am about to lean on them, so the distinction had better hold: what follows is not a stack of small trials. It is whole populations, counted in registers, before and after a law changed. Cut the sentence. End the paragraph at "landing at zero." and go straight into "Now the other pile."

Now the other pile. Using forty-one million observations covering the entire Danish population, researchers found that about eighty-five per cent of people are passive savers who do not respond to tax subsidies at all, and that each dollar of subsidy raised total saving by one cent, while automatic employer contributions raised wealth substantially. Britain made workplace pensions opt-out instead of opt-in and participation among the employees the rule covers went from 55 per cent to 88. Britain announced a sugar levy and the sugar sold per person in soft drinks fell about thirty per cent in three years, almost entirely because manufacturers changed the recipes rather than because shoppers changed their minds.

With both piles on the table I find the pattern hard to miss. With both piles on the table I cannot miss the pattern. Paying people to be prudent does approximately nothing, and telling them things does nothing measurable at all. What works is changing what happens when nobody does anything, and it works hardest on people with the least attention to spare, because it asks them for none. Changing what happens when nobody does anything works enormously, and it works on the people who have the least attention to spare, because it asks them for none.

That is also why the Swedish lottery result and this essay are not in conflict, though it took me a while to see it. A prize is a stock. What the evening is short of is a flow: attention, an unbooked hour, the second attempt. Money handed to a person once leaves the structure of their week exactly as it was, which is why the food, heating and credit numbers earlier are not a claim that cash buys health. They are a claim about what running short costs in the currency the evening actually spends.

The levers that exist where you live

Principle is cheap, so here is the concrete version. Principle is cheap. What follows is concrete, though what actually exists changes at every border, so several of these are addresses rather than instructions.

The first three you set once and they then run without you. That property is what the evidence says matters.

Find out what your workplace pension is doing without you. If you are in Ireland you were enrolled on 1 January and the first opt-out window closes at the end of this month; if you leave, your own money comes back but the employer and state contributions stay invested for you, and you are automatically re-enrolled two years later. If you are in Poland, opting out is not permanent by design: employers must re-enrol everyone every four years, the next round falls in 2027, and staying out means filing a fresh declaration each cycle. If you are in Britain you are almost certainly in already, at the minimum rate. Everywhere else the question is smaller than a scheme name and you can settle it this week: ask payroll whether anything leaves your pay for a pension beyond the state one, at what rate, and whether your employer adds to it. Almost everyone who has such a thing is on the lowest rate it permits, because that is the rate it started at. Moving it up is a form, signed once, and then it happens without you.

Put a standing order on the day you are paid. This is the one I would do first. Not a budget, not an app, not a resolution: an instruction to your bank to move a fixed sum to a different account before you see it. It is free to set up online in every European bank I know of. Calling it the household version of the Danish finding would be too generous, since the Danish finding is that most people never make this kind of choice. It is the closest thing a person can arrange alone to the one mechanism that did work there, which is money moving before anybody decides anything. The mechanism is the whole point, which is that it works identically on your worst week and your best.

If your tax code subsidises long-term saving, find the deadline and put it in your calendar now. Most European systems refund income tax on some form of retirement contribution up to a ceiling, and most of them settle it at the turn of the year, which means the decision has a date attached and the date is the part people miss.

The next three cost nothing and still need you to turn up, repeatedly, which makes them the harder half. The state has removed the money barrier and left the Tuesday one exactly where it was.

Screening, and specifically screening in your own country rather than in the abstract. Since December 2022 the EU recommends breast screening at 50 to 69 and asks member states to consider widening it to 45 to 74, HPV-based cervical screening at 30 to 65, and colorectal screening at 50 to 74. What you actually get depends on where you live to a degree that has nothing to do with you: breast screening reaches 83.3 per cent of eligible women in Denmark and 14.5 per cent in Greece, cervical screening 77.7 per cent in Sweden and 6.2 in Romania, colorectal 74.4 in Finland and 9 in Hungary, and Bulgaria and Romania run no breast programme at all. If you live behind one of the low numbers, nobody is going to write to you. That is when you have to do the system's job for it.

If you smoke, find out what your country actually provides, because it is probably less than you assume. Of the fifty-three countries in the WHO European Region, only eleven provide sufficient support to quit and thirty-nine offer something. Set against the sixty-four per cent of European smokers who used nothing at all last year, the gap between what exists and what gets used is enormous. And it will probably not be the attempt that works, which is an argument for booking it rather than against: at nothing per attempt, it can be the fourth as easily as the first.

If the problem is debt, this is the year the ground moves. From 20 November 2026 every member state must ensure independent debt advisory services are available with only limited charges, and lenders must refer people in difficulty to them. That is Article 36 of the new Consumer Credit Directive, and it matters because the starting point is dismal: about eighteen million over-indebted households across the EU and the UK, of whom roughly 1.6 million get advice. Over ninety per cent get none.

And one for everybody, since this essay is partly about money left on tables: work out what you are entitled to and are not claiming. Somewhere between a third and a half of the people eligible for the big European income supports do not receive them, and the reason is almost never that they decided against it.

Self-employment did most of it for me

My own margin was handed to me, and I very nearly wrote otherwise.

My years in Britain were spent self-employed in IT, which means nobody else set my hours. I could move the working day around, take a Tuesday morning when I needed one, and stop when the work was finished rather than when a shift ended. That is the largest single input into whatever discipline I appear to have, and I earned none of it. It came attached to the job. Working for yourself is no free gift either, because work with no closing time takes the evening whenever you let it, and when it does, the same things go first: the cooking, the walk, the early night. Which is the point. The schedule moved and the habits moved with it. My character was not consulted.

That is also why I distrust the version of this argument that ends in a shrug. The evidence above does not put a share on it. What it says is narrower and more useful: circumstances set the price of every attempt, and the same attempt is dearer for some people than for others. It says almost nothing about the difference between you and you, which is where the whole of your actual life happens.

In Ireland the window closes on the thirty-first of this month. Something over eight hundred and thirty thousand people look set to let it pass without doing anything, and in twenty years most of them will be glad. The ones in Poland who are out of their scheme will get asked again in 2027, whether they think about it before then or not. Nobody in either country is going to become a different person in between. Ask which of the things you keep failing at could be made to happen whether or not you are having a good week. Then set up as many of them as you can before the month ends.