trained, with the time and, comparatively speaking, the money to contemplate the plight of others. Living in Africa has opened my eyes to the Elastoplast effect of sporadic injections of money, which can lift even large numbers of indigent people for a time, but with little or no lasting effect. On a microscopic scale, I give every day to a man who sleeps in a bus shelter. The effect is superficial and transitory. He is back every night until the police move him on, because structural unemployment and political stasis leave him with no prospect of work. Effective altruists say that large- scale giving can have high expected value, but where poverty is politically and economically induced, well-meaning giving makes no substantive long- term difference. Without change at that level, nothing the effective altruists dream up will endure. D.A.L. COLDWELLProfessor of managementUniversity of the Witwatersrand, Johannesburg More letters on this topic can be found at: economist.com/lettersaltruism America’s gun violence will continue, you say, until voters follow Australians and Britons “and decide they have had enough” (“Under the gun”, October 10th). The truth is messier. From 2005 to 2008, nearly a decade after Britain banned most handguns, I was a detective superintendent in Lambeth, south London, where young gang members favoured machine pistols, firing them from passing cars. Local officers and the wider Metropolitan Police worked with community groups to stop the shootings. In December 2025, in Sydney where I now live, two gunmen killed 15 people at Bondi Beach. The older gunman held a firearms licence and owned six guns lawfully. Lambeth showed me the limits of a ban, and Bondi has shown Australians the limits of a licence. Strict laws have made both countries safer, yet each leaves a gap that falls to the police and the communities they serve, which is why Malcolm Gladwell’s point about the size of European police forces is

worth taking seriously. Crime fell in central London when almost a third more officers were put on its streets after the bombings of July 2005. In Philadelphia a randomised trial found that foot patrols on the most dangerous corners cut violent crime by almost a quarter, though the gains vanished when the officers left. Neither case isolated shootings, yet both show what visible policing adds to the law. America, with far more guns than either country, needs both. Jonathan TottmanSydney How should the law handle intent when robots act beyond their users’ control (“Unsuitable”, October 10th)? Consider the drunk driver. Getting blackout drunk doesn’t remove responsibility. The law traces culpability back to an earlier decision by an agent capable of making it; the pre-drunk self deciding to put the future self in that position. Why should autonomous systems be any different? A machine can’t bear moral responsibility. It sits with the human who exercises judgment in deciding to delegate authority to it. Liam TalbotBristol Your warning about relying on liability law to govern AI has a contractual dimension. OpenAI’s standard services agreement generally caps its contractual liability at the customer’s preceding 12 months’ fees, with exceptions including gross negligence, wilful misconduct and indemnification. Such caps can limit what an adopter recovers from its supplier even while the adopter faces claims from third parties. A small employer may control how it uses a model while having little power to inspect or repair it. Making the employer liable may encourage caution, but it need not give the developer the same incentive to remove a defect. Risk allocation should reflect who can prevent the particular harm, as well as who signs the purchase order. The bill can travel downstream. The power to fix the model may stay upstream.

Rohan SharmaUS delegate and technical expert for the International Organisation for Standardisation ISO/IEC JTC 1/SC

(AI standards)Temecula, California Pop and rock stars who demand that fans put away their phones at concerts are not considering the price of a ticket (“Crowd control”, September 19th). Musicians are charging extraordinary prices for the privilege of seeing them perform. Last year I paid more than $300 to see Sir Paul McCartney in Las Vegas. For that money I want to leave with more than a T-shirt and a programme (another $50 each). I want photo and video memories of what I experienced, captured on my phone and available to me long after the T-shirt has worn out. The Grateful Dead understood that recordings could extend the experience rather than diminish it, famously allowing, and even encouraging, fans to tape their concerts. Today’s artists should recognise the value of allowing fans to do the same. If I’m paying hundreds of dollars to be there, I’d like to decide how much of the experience to preserve. Jason RussellLehi, Utah Regarding the looming government-debt crisis (“Will bonds blow up?”, October 10th), politicians will not treat voters like grown-ups until voters

start behaving like grown-ups. Voters consistently want leaders who offer them candies and sweets, without understanding that this will only lead to the dentist’s drill. Vera HutzuliakMississauga, Canada This article was downloaded by zlibrary from https://www.economist.com/letters/2026/10/22/how-effective-is-effective-altruism

· By Invitation

Ignore the doomsayers. American democracy is in good shape A former ally asks why Donald Trump abandoned his own legacy

By Invitation · By Invitation | Integrity intact

Ignore the doomsayers. American democracy is in good shape Its resilience rests not on politicians’ restraint but on the robustness of institutional arrangements, writes Frances Lee Oct 22nd 2026 AMERICA’S TWO main parties do not agree on much, but they share a slogan. “We’ve got to be too big to rig,” Kamala Harris told an audience recently. Days later Donald Trump called for a midterm win “too big to rig” because “they rig elections in this country.” Both parties urge their own supporters to vote while warning that the other side cheats. Scholars and journalists, meanwhile, have settled on a diagnosis: democratic backsliding. Many voters themselves are anxious. Looking towards the midterms on November 3rd, only 49% of respondents expect a free and fair election. Under Mr Trump’s influence Republicans are especially cynical. Democrats worry about election interference from his administration. Minority voters

fear harassment from Immigration and Customs Enforcement agents at polling stations. Anxieties notwithstanding, American elections are healthier than the backsliding narrative allows. Indeed, there was also considerable anxiety in advance of both the 2022 and 2024 elections. Then, as now, Americans worried about election interference, voter fraud and political violence. But in the end turnout was high by historical standards and the vast majority of Americans expressed confidence in the outcomes. American democracy rests on durable institutions capable of checking concentrated power. Midterm elections are important in this regard. The “midterm law”—that the president’s party loses seats midway through his term—has held in 20 of the 22 votes since 1938. Mr Trump’s first term followed the pattern: Republicans lost their majority in the House of Representatives in 2018. Notably, the president’s party loses seats in midterms even when most Americans approve of the president’s performance. The penalty grows when the president is unpopular and the economy weak. Mr Trump’s approval rating languishes in the mid-30s. In September consumer confidence fell to its lowest level since 2014, partly owing to high fuel prices. Across midterms since 1992 the president’s party has lost an average of 25 House seats. The current Republican majority disappears with a net loss of three. In an effort to forestall losses, Mr Trump induced Republican-controlled states to redraw their electoral maps. The resulting gerrymanders probably netted some Republican seats, albeit fewer than they had hoped. But here again American institutions retain safeguards against a president’s effort at electoral deck-stacking. A president can demand new maps only in states where his party holds the pen, and Democrats in California answered a gerrymander in Texas with one of their own. The Cook Political Report projects a net Republican gain of four seats from redistricting, not nearly enough to protect the party’s House majority. Mr Trump floated and later disavowed the idea of postponing or cancelling the midterms. Regardless, that scenario was always wildly implausible. Election day is fixed by federal statute, and the constitution sets the date

when office-holders’ terms end. The decentralised administration of elections across thousands of local jurisdictions denies the president any easy lever to manipulate outcomes. Mr Trump’s attempt to turn the postal service into such a lever against mail-in ballots was thwarted by the courts. More likely than interference with the election is that unsuccessful office- seekers will refuse to concede, claiming their opponents cheated. Nothing is more basic to democratic legitimacy than the loser’s consent. Many feared that Mr Trump’s refusal to concede in 2020 would spread across the political system. It has not. Bright Line Watch, a research group, examined all 884 congressional races in 2020 and 2022. The overwhelming majority of losing candidates publicly acknowledged defeat. Explicit refusals were concentrated among first-time candidates who had lost by wide margins. And there were fewer such refusals in 2022 than in 2020. Candidates with political careers to protect still concede. When they do, complaints of election “rigging” lose their force. None of this is cause for complacency. My research with Geoffrey Layman and Christina Wolbrecht finds that party activists, long regarded as guardians of democratic norms, are now no more committed to losers’ consent than ordinary citizens. Any deterioration in basic democratic values is worrying. Nevertheless, the record of American elections since 2020 has been encouraging. Election administration has improved. Although many career administrators resigned amid increased harassment after 2020, the voters’ experience actually got better. In both 2022 and 2024 the share of people waiting more than 30 minutes to cast their ballot was lower than in 2020, and the long-documented racial gap narrowed: by 2024 black, white, Hispanic and Asian voters reported comparable waiting times. Overall, the evidence points to a resilient democracy. That resilience rests not on politicians’ restraint but on institutional arrangements that are difficult to manipulate: fixed election dates, decentralised administration and officials independently elected in their own constituencies, with their own interests to protect. James Madison defended the House’s two-year terms as giving Congress “an immediate dependence on, and an intimate sympathy with, the people”. That