n.V1-6.04 | WHAT GOVERNMENTS DO WHEN THEY GOVERN
01| Positions on poverty are usually catalogued by faction — the right thinks this, the left thinks that. That framing obscures more than it shows, for two reasons. Factions divide internally more sharply than they divide from each other on several questions. And parties in government routinely do things their stated position does not predict.
02| So take the positions by what they answer rather than who holds them, then test the map against what governments have actually done. Latin America has run the experiment repeatedly over the past decade — alternating between left and right administrations while publishing the statistics that make comparison possible.
Three questions that divide
03| Why are people poor? Behaviour and incentives: Charles Murray's argument that welfare altered the returns to work and family formation and the poor responded rationally — a claim about how people respond to policy, not about their character, though it is usually read as the latter. Missing structures: opportunity unequally distributed through schooling, health, credit, networks and discrimination — the mainstream centre-left position and the operating assumption of most of the development economics profession. Missing legal infrastructure: Hernando de Soto's claim that the poor hold assets in forms the legal system does not recognise. Extraction: wealth flowing outward through colonial legacy, debt service, trade rules and corporate tax avoidance, of which Jason Hickel is the best-known exponent. And predatory domestic elites: George Ayittey's vampire state, which holds that African governments capture the state's resources and then blame external forces — an internal African critique that should not be collapsed into either Western position.
04| Evidence bears on this more directly than the debate suggests. The Mbeki High Level Panel, reporting after consultations from 2011 to 2014, estimated Africa loses at least $50 billion a year to illicit financial flows, and attributed roughly 65 percent to commercial activity — chiefly tax evasion by multinationals — 30 percent to criminal activity, and 5 percent to corruption by officials. The panel described the total as an underestimate.
05| That decomposition complicates both accounts as they are usually stated. Official theft is real and small as a share of the total. The dominant channel is corporate. And the money leaves through international financial infrastructure. The finding is compatible with a weak version of each argument and refutes the strong version of both.
06| What does assistance do to the recipient? That it deforms agency is the oldest claim — Peter Bauer's, restated by Dambisa Moyo. That it provides a floor that makes agency possible is the position underlying every social protection system: material security as a precondition for planning and risk-taking, not a substitute for them. That it depoliticises is James Ferguson's: development converts political questions about who holds power into technical questions about delivery, and the conversion is the function rather than a flaw.
07| The labour-supply version of the deformation claim has been tested directly. Banerjee, Hanna, Kreindler and Olken re-analysed seven randomised evaluations of government cash transfer programmes across six countries and found no systematic evidence that they discourage work, on either propensity or hours. The depoliticisation critique is normative and cannot be settled this way.
08| Who should act — market, state, community, or nobody — is the axis on which the spectrum is conventionally arranged. It is also the axis on which governing behaviour least reliably follows stated position.
The record
09| Mexico. López Obrador's administration terminated Prospera — descendant of Progresa, the world's first national conditional cash transfer and the most-evaluated anti-poverty programme in history — calling it clientelist, and replaced it with largely unconditional universal transfers. Social spending rose from 253 billion pesos in 2019 to 741 billion by 2024; the minimum wage from 88.36 pesos a day in 2018 to 207.44. Poverty fell from 41.9 percent to 29.6 percent, 51.9 million people to 38.5 million. Conditionality — school attendance and health visits in exchange for money — was a technocratic centrist innovation defended by the evaluation establishment. The left abolished it, on the grounds that conditions are paternalistic.
10| Brazil. Bolsonaro's government replaced Bolsa Família with Auxílio Brasil in December 2021, raising benefit values by roughly 18 to 20 percent and expanding coverage from 14.6 million to 16 million families ahead of an election. Lula's returning administration restored the name and expanded further — 13.2 million beneficiary families in December 2019 to 21.1 million by December 2023, monthly outlays from 2.1 billion to 14.2 billion reais. Both administrations expanded cash transfers. They differed on branding, on conditionality design and on scale. Not on the instrument.
11| Argentina. Milei's first year involved the largest social spending adjustment since 2002. Inside that contraction the universal child allowance rose roughly 47 percent in real terms while the food transfer lost real value. Poverty measured by INDEC rose to 52.9 percent in the first half of 2024, then fell to 38.1, then 31.6, then 28.2 percent by the second half of 2025 — a decline analysts attribute principally to disinflation rather than social policy. A self-described libertarian protected and expanded one universal child transfer while cutting most others, and the series moved violently in both directions inside eighteen months.
12| Colombia. Petro consolidated a fragmented system into Renta Ciudadana. Coverage fell from roughly 4 million households to about 3.3 million while per-household amounts rose substantially and were paid more often. That is the coverage-versus-depth trade-off made explicit, chosen by a left government in the direction of fewer households receiving more — the opposite of what the position usually implies.
13| Cuba and Venezuela. Cuba's universal model produced indicators cited for decades — literacy near 99 percent, life expectancy around 78, infant mortality below 5 per 1,000 — and its current position is different: acknowledged recession in 2024, rationing delivering less than its entitlements, prolonged blackouts. Venezuela's Misiones were large, oil-funded and state-led; the subsequent trajectory is documented not by the state but by a university household survey, which recorded income poverty at 94.5 percent in 2021 and 82.8 percent of households in 2023, with the Gini rising from 40.7 to 53.9 between 2014 and 2024. The cross-cutting finding in both is about measurement rather than policy. Where governments fear the numbers they stop producing them, and the substitutes carry their producers' politics. Any figure cited for either country should travel with its provenance: a university survey, a human rights monitor and a national statistics office are three different kinds of object, and a government with an interest in the number is a fourth.
What the record shows
14| Cash has won across the spectrum. Every government here — libertarian, left-populist, centre-right, socialist-adjacent — runs cash transfers as a central instrument. The argument about whether to transfer money is largely over.
15| Conditionality is not a left–right marker. A left government abolished the flagship conditional programme on anti-paternalist grounds; right governments have retained conditionality elsewhere. The position sorts by attitude toward state supervision of the poor, which cuts across the economic axis.
16| Coverage versus depth is the live operational axis — Colombia chose depth, Brazil chose breadth, Argentina chose depth for children and contraction elsewhere — and it is almost never how the debate is framed in public.
17| Attribution is weak in every single case. Argentina's decline is credited to disinflation, Mexico's is contested between transfers and the minimum wage, Brazil's is attributed to both assistance and employment growth. In no case does the published record permit clean attribution to the government's distinctive policy. Which means all five are cited by their supporters as vindication, and none of those citations is well founded.
