Synopsis
By the mid-1960s, the methods Charles Hitch had developed at the RAND Corporation for optimising nuclear deterrence had been carried, almost intact, into the welfare state. The Planning, Programming, and Budgeting System that Robert McNamara deployed at the Department of Defense was extended across the domestic agencies of the Johnson Great Society; the economists and systems analysts who had spent the 1950s calculating optimal bombing patterns were now calculating cost-effective allocations for Head Start, Model Cities, and the Office of Economic Opportunity. The transfer accomplished something more fundamental than an administrative update: it replaced the welfare state’s original question — what do these people need? — with a different question — given a defined objective and limited resources, what is the optimal allocation? The two questions sound adjacent; in practice, the second is a replacement for the first. The objective was now set by analysts, administrators, and elected officials with their own interests, and the analytical framework was not designed to challenge it.
1. The War Room Comes to Washington
The RAND Corporation’s analysts in Santa Monica had learned, under conditions of genuine urgency, that killing and destruction could be made more efficient by treating them as optimisation problems. Charles Hitch, who ran RAND’s Economics Division from 1948, argued in The Economics of Defense in the Nuclear Age (1960) that public-sector decision-making should be treated as an economic problem: specify the objective, enumerate the options, calculate a cost-effectiveness ratio for each, identify the optimal allocation. The framework treated the objective as a given — as a technical parameter determined before analysis begins, whose content the analysis would not question. For defence procurement, this division of labour was workable. The objectives of nuclear deterrence could be operationalised; the question of how much deterrence was enough admitted no clean answer but at least admitted measurement. But a framework that accepts objectives as given and optimises within them is not neutral with respect to whoever has the authority to specify what the objective is.
The migration of RAND’s methods into social policy was as much institutional as intellectual. Throughout the late 1950s, RAND’s board was actively seeking to reduce dependence on Air Force contracts. The Great Society programmes of the mid-1960s created a substantial new market. The Office of Economic Opportunity, HEW, the Department of Housing and Urban Development, and the Department of Labor all required analytical capacity they did not have in-house. RAND’s board pushed the organisation into non-military research from 1965 onward; the methods followed the money. The RAND that arrived in the social policy arena brought a complete analytical culture: quantitative modelling over qualitative case study, cost-effectiveness ratios as the primary programme evaluation criterion, and a confidence that expert analysis could resolve questions that democratic deliberation had been too slow to settle.
RAND’s work for New York City, which began in 1969 as the New York City–RAND Institute, provides the paradigm case. The Institute applied queuing theory and optimisation modelling to fire station placement and response capacity. Acting on RAND’s recommendations, the Lindsay administration reduced fire service capacity in the South Bronx, Brownsville, and Bushwick — communities experiencing extreme housing stress, overcrowding, and arson-for-insurance — because the model’s cost-effectiveness ratios made this appear optimal. The model could not see the relevant risk because fire risk of the accumulative kind — abandoned buildings, cascading neighbourhood departure, the systemic collapse of informal fire prevention — was not among its variables. The model measured response times and cost per call. It did not measure what would happen to a community if its fire stations closed.
2. McNamara’s Children
Robert McNamara brought PPBS from the Ford Motor Company to the Pentagon in 1961, and Lyndon Johnson mandated its extension across all federal agencies in 1965. The system required each agency to specify its objectives in measurable terms, enumerate programme alternatives for achieving them, calculate a cost-effectiveness ratio for each alternative, and allocate resources to the highest-ratio options. This was Hitch’s framework scaled to the entire federal government and applied to social programmes that had previously been evaluated in terms of coverage, eligibility, and political legitimacy rather than cost-effectiveness ratios.
The specific consequence for welfare was that the objective that proved most tractable was not “ensure that poor families have enough to eat” — which would have required measuring adequacy in household-specific terms and would have connected the analysis to Orshansky’s 124 thresholds. The tractable objective was “reduce welfare caseloads” and “increase labour market participation,” because caseloads and participation rates could be expressed as measurable outcome metrics, could be tracked over time, could anchor a cost-effectiveness ratio, and could be defended in a congressional appropriation vote. A welfare programme that reduced caseloads by ten per cent at a cost of 2X per caseload-reduction, regardless of whether either programme had done anything for the material conditions of the households it served.
3. The Moynihan Problem
In March 1965, Daniel Patrick Moynihan — then Assistant Secretary of Labor — circulated a classified report: The Negro Family: The Case for National Action. Its central statistical finding was that the number of Black families receiving Aid to Families with Dependent Children was rising even as Black unemployment was falling. Moynihan interpreted this as evidence of pathological family structure: the rise of female-headed households, the “tangle of pathology” he described, was self-perpetuating and had become, in his analysis, the proximate cause of persistent Black poverty, more significant than current discrimination or economic conditions. The report did not ask what AFDC benefit levels were relative to what families needed to survive, or what the unemployment rate for Black men had been in the preceding decade, or what the housing and labour market conditions were in the specific cities where caseload growth was occurring. It treated statistical outcomes as evidence of cultural deficiency.
The report was immediately and accurately criticised by Black scholars and civil rights leaders for its misuse of statistics, its pathologisation of Black family structure, and its diversion of policy attention from structural causes to behavioural ones. William Julius Wilson’s later analysis, and the extensive subsequent literature on concentrated disadvantage and structural unemployment, confirmed that the statistical patterns Moynihan had attributed to family pathology were the predictable consequences of deindustrialisation, residential segregation, and discriminatory employment. But the report’s influence on the subsequent policy conversation ensured that behavioural explanations for racial poverty disparities were institutionally available and analytically legitimate for the next three decades — and the cost-effectiveness framework had no mechanism for distinguishing between them and structural explanations, because it took objectives as given and did not question whether the objectives were tracking the right thing.
4. The New Jersey Experiment
In 1968, the Office of Economic Opportunity funded the New Jersey Income Maintenance Experiment: the first social policy randomised controlled trial in American history. The experiment randomly assigned low-income families to different guaranteed-income treatment levels and a control group, then tracked their employment and income over three years. Its primary measurement objective was labour-supply elasticity: did recipients work less when they received guaranteed income? The answer was yes, slightly — treatment families worked an average of five to seven per cent fewer hours than control families. This finding was used in the congressional debate to argue against Richard Nixon’s Family Assistance Plan, which proposed a guaranteed minimum income. The experiment’s other findings — improved health outcomes, improved school attendance and test scores among children in treatment families — received no equivalent weight in the policy conversation.
The New Jersey experiment inaugurated a methodology, the social policy RCT, that would shape welfare programme evaluation for the next half-century. Its specific design embedded the cost-effectiveness objective function at the moment of measurement: what mattered was whether recipients worked, because “reduce welfare caseloads” and “increase labour market participation” were the tractable metrics the PPBS framework had installed. What households needed to survive — the question Orshansky, Townsend, and Tillmon had each asked from different directions — was not a variable in the experiment’s primary analysis. The RCT’s authority derived from its experimental rigour. Its rigour was in service of the wrong question.
5. Legacy: From Cost-Effectiveness to Conditionality
The policy trajectory from PPBS through the War on Poverty’s analytical culture to TANF and Universal Credit is traceable as a direct institutional descent. TANF’s work requirements and time limits — enacted in the 1996 Personal Responsibility and Work Opportunity Reconciliation Act — are the cost-effectiveness framework’s logical endpoint: the welfare state’s objective is defined as labour market participation, the population that fails to achieve the objective is sanctioned, and the programme’s success is measured by caseload reduction. Universal Credit replicates the architecture in the UK context: a claimant commitment specifying weekly job-search requirements, automatic benefit sanctions for non-compliance, and a digital administrative system that enforces the objective function without requiring any caseworker to assess whether the objective is appropriate to the specific household’s circumstances.
Hitch’s original framework was explicit that the choice of objective was a political decision outside the analyst’s remit. The analyst served the decision-maker, who served the political process. But in transit from the Pentagon to the welfare agency, the political check did not travel with the method. The objective was set by the analytical culture that the method itself had produced — by economists trained in cost-effectiveness reasoning, by programme evaluators whose RCT designs measured the tractable metric, by administrators whose institutional incentives aligned with the objective of caseload reduction. The households whose welfare the programmes were nominally designed to serve had no equivalent role in specifying what the objective was. The result is welfare policy that is, in the PPBS sense, optimally efficient at achieving its objective, and whose objective has nothing to do with what poor families need.
Connection Forward
Chapter 7 traces where RAND’s analytical culture and the eugenic statistical tradition of Chapter 2 converged: in the meritocratic ideology of Silicon Valley, where the cognitive distribution became the market’s natural sorting mechanism, operations-research efficiency became disruption, and the welfare state became the obstacle to a self-correcting market whose fitness sorting the RAND framework and the Galtonian tradition had each prepared their adherents to accept.