Topic 27 o 64· Governmental performance an delivery metrics

Ootcomes-based accoontability (OBA)

Ootcomes-Based Accoontability, an aw cried Results-Based Accoontability (RBA), is Mark Friedman's framework for separatin twa questions that public-sector reportin habitually blurs taegither: "is the population daein weel?" (population accoontability) and "is this specific programme daein weel?" (performance accoontability). Mixin the twa up is, in Friedman's tellin, the single maist common reason weel-run programmes get blamed for population trends they never hed the pouer tae move.

Why it maitters

Friedman set the framework oot in Trying Hard Is Not Good Enough (2005), arguin that maist public reportin either droons decision-makers in population-level statistics nae single agency controls (teen pregnancy rate, unemployment rate, life expectancy) or droons them in programme-level activity coonts (clients seen, referrals made) that say naething aboot whether onybody's life improved. RBA's contribution is a wee, disciplined vocabulary that keeps the twa apairt: population results (conditions o wellbein for a hale population, like "children are born healthy") belang tae nae single agency and need mony pairtners movin taegither; performance measures (hou weel a specific programme serves its specific clients) belang tae ae agency and sud be judged only agin whit that agency can actually influence. Friedman's "three performance questions" — hou muckle did we dae, hou weel did we dae it, and is onybody better aff? — is noo embeddit across US state and county human-services contractin and, via the RBA-aligned consultancy and toolkit Clear Impact, widely uised in UK and Commonwealth local-government commissionin. The practical stakes are contractual: a housin programme sudnae be defunded because the city's homelessness rate rose frae macroeconomic causes ootwith its reach, but it absolutely sud be defunded if its ain clients arenae bein housed.

The maths

Population accountability (the "big picture" a community, region, or nation shares):
  Result       — a condition of wellbeing (e.g. "residents are economically secure")
  Indicator(s) — a measure of that condition (e.g. unemployment rate, median household income)
  → no single programme owns the indicator; movement requires many contributors

Performance accountability (what one programme is responsible for):
  How much did we do?      — activity volume (clients served, units delivered)
  How well did we do it?   — quality/efficiency (% completing programme, cost per client)
  Is anyone better off?    — the outcome that matters (% in employment 6 months after
                              programme, before/after or against a comparison group)

A programme is judged on the third performance question, never on the population
indicator directly, unless its scale and design could plausibly move it alone.

Wrocht example

City-funded employment support programme, 500 participants/year, contractit by a local authority unner an RBA-style performance framework:

Population indicator (context, not the programme's scorecard):
  City unemployment rate: 6.2% (up from 5.8% the prior year, driven by a factory closure
  outside the programme's control)

Performance measures (the programme's actual accountability):
  How much:   500 participants enrolled (target 480) — met
  How well:   78% completion rate; cost per completer = £340,000 / 390 completers ≈ £872
  Better off: of 390 completers, 260 in sustained employment at 6 months = 66.7%
              versus a matched comparison group's 41% (see counterfactual-analysis)

Unner a population-accoontability read, the programme looks like it is failin — the city's unemployment rate gaed up on its watch. Unner RBA's performance-accoontability read, the programme is succeedin: it hit its volume target, held quality steady, and produced an employment ootcome 25.7 percentage pynts abune a matched comparison group, while the population indicator muivit for reasons (a factory closure) entirely ootwith the programme's control.

RBA maps directly ontae a familiar SRE distinction: population indicators are like business-level North Star metrics that nae single ingineerin team awns end tae end (company revenue, market share), while performance measures are like a team's ain SLOs — the things that team's design decisions actually muive. A dashboard that reports baith wioot labellin which is which invites exactly the misattribution RBA wis biggit tae prevent: an on-call ingineer bein blamed for a metric a dependency team controls. Whan commissionin or biggin reportin tools for ootcomes contracts, big the "hou muckle / hou weel / better aff" triad as first-class, separately-filterable fields raither nor a single blended KPI — it is the same discipline as separatin leadin and laggin indicators in public sector KPIs. RBA is an aw the accoontability logic unnerneath payment by results and social impact bonds: a PbR contract can only fairly pay on the "better aff" performance measure, never on the population indicator, unless the intervention is genuinely the dominant driver o it.

Pitfaws

  • Payin or penalizin a programme agin a population indicator it cannae control: this is the single mistak RBA exists tae prevent; aye trace whether the programme is a major or minor contributor tae the population result afore attachin consequences tae it.
  • Reportin "hou muckle" as if it were "better aff": activity coonts (clients seen) are the easiest data tae collect and the least informative; insist the "is onybody better aff" question is answered wi real ootcome data, ideally agin a coonterfactual (see counterfactual-analysis).
  • Treatin RBA indicators as fixed forever: Friedman's method is explicitly iterative — a "data, story, what works, action plan" cycle — nae a yin-aff scorecard design exercise.
  • Nae comparison group for "better aff": a before/efter chynge wi nae coonterfactual mixes up programme effect wi the trend the population wad hae shown onywey.

Sources