When Institutions Learn – and When They Don’t

on plumbing and feedback loops

The Chinese loop is tightly geared: experiments, metrics, promotion, promulgation. The other three loops leak energy—through partisan vetoes (US), fiscal dependency (India), or capacity gaps (Indonesia). None of this requires moral judgement about regime type; it is engineering. Tight gears deliver torque; loose belts slip.

Adam Tooze once described Beijing’s economic managers as surfers riding a succession of crises: the Party loosens the leash when growth stalls, yanks it tight when inflation or social unrest looms, then paddles out again for the next wave. Xu Chenggang peers deeper and says the board itself is special. China’s political “institutional genes” – a Party monopoly on top-level goals married to unusually wide fiscal elbow-room for provinces – create a built-in learning circuit. Local governments now handle about 85 per cent of all public spending, yet their officials live or die by centrally run promotion audits that reward hitting national targets and punish missing them, sometimes with a single one-vote veto on issues like pollution or debt. That combination invites thousands of pilot schemes – more than 20,000 since the 1980s, by one recent count – but funnels the results through a single data valve in Beijing. When Shanghai’s free-trade rules or Jiangsu’s river-chief model show better numbers, the State Council can stamp them into nationwide law within a few budget cycles; when something flops, it disappears without ever troubling the rest of the map.

The United States loves to call its fifty states “laboratories of democracy,” and in sheer variety the labs still hum. Massachusetts’ 2006 health-insurance mandate re-emerged as Obamacare four years later; California’s car-emissions rules keep nudging federal standards. Yet new work tracking 602 state laws since 2000 shows diffusion now follows party colour, not neighbouring success: red states copy red, blue copy blue, and cross-party borrowing has plummeted. Because Washington cannot compel uptake – it must coax with grants or wait for Congress – the vertical pipe that once carried proven ideas to the centre often clogs. Medicaid expansion, the best-evaluated path to closing America’s insurance gap, remains blocked in ten states a dozen years after the evidence was in. The laboratories still invent; the national system increasingly declines to read their lab notes.

India proclaims “co-operative competitive federalism”: 28 big states generate most frontline innovations, and the Union now ranks them on everything from school results to startup culture. But Delhi still collects about sixty per cent of national revenue while states shoulder roughly sixty per cent of spending, so the financial oxygen that powers bold policy remains thin outside the capital. When Telangana launched a cash transfer for farmers in 2018, New Delhi liked the optics, rebadged it as PM-Kisan and rolled it out country-wide – diffusion, yes, but by central appropriation rather than peer emulation. Other successes, from Kerala’s panchayat health model to Tamil Nadu’s school-meal strategy, crossed state lines only after Supreme Court orders or federal grants. NITI Aayog’s growing best-practice portal is nudging things faster, yet horizontal learning still moves at the speed of party rivalry and civil-service caution.

Indonesia, two decades into its “big-bang” decentralisation, has woven an elaborate scaffolding to capture the torrent of local ideas. Each year thousands of districts upload projects to the SINOVIK contest; the Top 99 are splashed across a public database, and since 2023 the JIPPNas portal has merged those files with Ministry data so any mayor can download a ready-made blueprint. The finance ministry even attaches bonus grants to winners to encourage replication. Yet diffusion still relies on voluntary uptake and local capacity. Poorer kabupaten often lack the engineers or payroll to copy a digital permit system that dazzled Jakarta. President Prabowo’s proposal to scrap direct regional elections in favour of appointments – floated last December on cost-saving grounds – has raised fears that the electorate’s spur to innovate could dull just as the scaling engine is beginning to work.

Across all four cases the lesson is institutional, not ideological. China’s circuit is tightly geared: centre sets ends, provinces try means, audits decide winners, scaling follows quickly. America, India and Indonesia each allow greater pluralism, but their feedback belts slip – on party mistrust in the US, on fiscal dependency in India, on uneven administrative muscle in Indonesia. Systems that insist they already know what works, or that disdain borrowing from rivals, are the ones that learn slowest. The trick is not authoritarian command or democratic virtue; it is plumbing that turns local discovery into national capability before the next crisis rolls in.

the most disturbing flower arrangement I’ve come across