Scottish referendum polls

Episode Summary

Executive Summary: This episode of More or Less examines how numbers can mislead or clarify public debate. It scrutinizes volatile Scottish independence polls, rebuts misleading claims about Scottish welfare dependency, debunks the idea of a Strictly Come Dancing divorce curse, and challenges alarmist reporting on food poverty and malnutrition. The common thread is that statistics need context, methodology, and careful interpretation before they drive political or media conclusions.

Main Topics: Scottish referendum polling volatility (Priority: 5/5): The programme explains why a series of conflicting referendum polls caused political and market turmoil, focusing on the limitations of internet polling and the uncertainty of predicting a first-of-its-kind referendum. How internet panels differ from random polling (Priority: 5/5): Pollsters describe building representative online panels by quota and weighting rather than using purely random samples, and why this can work imperfectly when there is limited historical precedent. Misleading claim about Scots 'living on benefits' (Priority: 4/5): The episode fact-checks Nigel Farage’s claim about welfare dependency in Scotland, showing that the figure includes pensions and in-work benefits and is similar across the UK. Food poverty, malnutrition, and hospital coding (Priority: 5/5): The show examines whether rising malnutrition admissions reflect worsening poverty or improved detection and coding, concluding that the increase is likely driven largely by awareness and administrative practices. Debunking the 'Strictly curse' (Priority: 4/5): Using divorce-rate calculations, the programme argues that celebrity relationship breakdowns on Strictly Come Dancing are not unusual once the professions involved and baseline risk are considered. Countdown and mental arithmetic (Priority: 3/5): A celebratory segment revisits a famous Countdown numbers-round moment, explaining the shortcut logic behind the solution and highlighting the difference between pattern recognition and full calculation.

Key Arguments: Polling shocks in the Scottish referendum were driven by closely fought, method-dependent estimates rather than clear evidence of a stable lead. Internet polls are not simple random samples; they rely on building a modeled panel and then adjusting it using historical experience and weighting. Because the Scottish referendum had no prior equivalent, pollsters lacked historical data to calibrate biases with confidence. Claims that more than half of Scots are 'living on benefits' are misleading because the category includes pensions, child benefit, and tax credits, not just unemployment support. Reported rises in malnutrition may reflect better awareness, screening, and coding in hospitals rather than a true surge in severe food poverty. The Strictly Come Dancing 'curse' disappears once divorce rates are compared with expected rates for a group that includes dancers, who have higher relationship breakdown rates on average. The programme repeatedly stresses that raw numbers without denominator, context, and definition can create false impressions. In the Countdown example, the apparent complexity of the solution hid a simpler algebraic identity that avoided needing the intermediate multiplication to be computed explicitly.

Data Points: Yes vote in Sunday Times/YouGov internet poll: 51% - Poll result excluding don't-knows that triggered major political and market reaction No vote in Sunday Times/YouGov internet poll: 49% - Complementary figure in the same poll excluding don't-knows Later YouGov poll lead for Better Together: 6 points - A subsequent poll showed the No/Better Together side ahead by a solid margin Guardian telephone poll lead for Better Together: 2 points - A closer result within the margin of error Scottish people receiving some state support: 60% - DWP figure cited in response to Nigel Farage; includes pensions and in-work benefits People receiving state support in England: 60% - Same DWP measure as Scotland, showing the statistic is not uniquely Scottish People receiving state support in Wales: 64% - Comparable UK-wide benchmark People receiving state support in Northern Ireland: 66% - Comparable UK-wide benchmark Benefit spending per person in Scotland (2011-12): over £3,200 - A more useful measure than headcount of benefit recipients Gap in benefit spending versus UK average (2011-12): 2% - Scottish per-person spending was higher but only slightly Gap in benefit spending versus UK average five years earlier: 7% - Shows the difference had been shrinking Countdown series count: 70 series - Part of the show’s Guinness World Records celebration Countdown episode count: 6,000 episodes - Illustrates the programme’s longevity Rickets diagnosis trend: down by a third - Hospital diagnoses in the last couple of years contradicted anecdotal claims of increase Malnutrition admissions (2009-10): just under 4,000 - Recorded hospital admissions with malnutrition as primary or secondary diagnosis Malnutrition admissions (year ending April, latest cited): nearly 7,000 - Shows a substantial rise in recorded cases Primary malnutrition diagnoses: a few hundred / 612 cited later - Only a small share were main reasons for admission Malnutrition cases coded over time: five-fold increase - Professor Elia argued coding/awareness changes explain much of the rise Malnutrition cases in three most deprived deciles: 40% - Area-based deprivation data TB and measles cases in three most deprived deciles: nearly 60% - Used as a comparison showing stronger deprivation gradients for other diseases Malnutrition primary diagnoses in people over 60: more than half - Shows malnutrition is especially common among older adults Malnutrition primary diagnoses in children/teenagers: 35 - Very small share of primary diagnoses Strictly contestants with marital split in same year: 2 out of 150 celebrities - Used in the divorce-rate calculation UK national divorce rate: 0.3% per year - Baseline expectation used to estimate likely splits Expected divorces among Strictly contestants/dancers: just under 1 - Initial calculation using general population rates Adjusted expected divorces accounting for dancers: about 2.3 - Revised estimate after factoring higher divorce rates for dancers Historical occupational divorce-rate example (1900 US census): actors and showmen top the list - Supports the idea that some professions have higher relationship breakdown rates Historical occupational divorce-rate example (2000 US census): dancers and choreographers top the list - Supports the revised Strictly calculation

Pivotal Quotes: "It's almost building a model of the British population." — Anthony Wells: Explaining how YouGov internet panels differ from random-sample polling "I think this referendum, it has the potential to be a poll in Waterloo, the biggest since 1992, when the polls got it wrong." — Martin Boone: Expressing concern that Scottish referendum polling could prove badly inaccurate "What you can't control for is, for example, they also like to be more au fait and more comfortable with technology. And those things are just inescapable." — Anthony Wells: Describing the limits of weighting and adjustment in internet polls

Implications: Listeners should treat headline statistics cautiously: definitions, sampling methods, and baselines matter more than raw figures. For pollsters, hospitals, and media outlets, better context and transparency are essential to avoid overreaction and false narratives.