Reoffending rates, Welsh taxes and the menopause

Episode Summary

Executive Summary: This episode of More or Less examines several claims through data: whether falling “re-offending” rates really show policy success, whether Wales has too few higher-rate taxpayers to fund the NHS via tax rises, why a previous nurses’ pay comparison was methodologically flawed, whether 13 million UK women are menopausal, and how English usage is shifting toward “data is” over “data are.” The common thread is that definitions, denominators, and measurement choices strongly shape public claims.

Main Topics: Re-offending rates vs. re-convictions (Priority: 5/5): The programme argues that the government’s falling re-offending rate is actually a measure of re-convictions/re-cautions, not direct offending, and may partly reflect reduced charging and cautioning rather than better rehabilitation. Declining charge and caution rates in crime outcomes (Priority: 5/5): Separate Home Office data show sharp falls in the share of recorded crimes leading to charges or cautions, raising concerns that lower re-offending rates could be influenced by weaker criminal justice throughput. Welsh tax base and funding the NHS (Priority: 4/5): A correction to a public claim about only 9,000 higher-rate taxpayers in Wales leads to a broader explanation of Welsh income tax structure and why meaningful new revenue would likely require raising the basic rate. Nurses’ pay comparison and mean vs median (Priority: 3/5): The episode revisits a previous item to explain that the OECD’s international comparison used mean pay, whereas median pay is usually the better statistic for salary comparisons because of skew from very high earners. The 13 million menopausal women claim (Priority: 4/5): The show investigates a widely repeated estimate and concludes it likely counts all women over 45, making it an imperfect proxy for peri-/menopausal women; the actual number experiencing symptoms is probably far lower. Grammar and the usage of 'data' (Priority: 4/5): Using corpus linguistics, the episode tests whether 'data' is more often treated as singular or plural and finds modern usage increasingly favors 'data is,' especially in general English, though academic and journalistic writing still often use plural agreement.

Key Arguments: The government’s “re-offending” rate is not a pure measure of offending behavior; it is a proxy for re-convictions and re-cautions within a fixed follow-up period. A falling conviction/caution rate can mechanically lower re-offending statistics even if underlying crime behavior has not improved. Recorded-crime outcome data show major declines in charge and caution rates, suggesting criminal justice system throughput may be part of the explanation. The Ministry of Justice’s response cites more police officers and anti-offending measures, but does not address the charge-rate critique. Vaughan Gething’s claim of only 9,000 higher-rate taxpayers in Wales was a slip; the intended figure was 9,000 additional-rate taxpayers. Because Wales has a relatively small higher-income tax base, large NHS funding gains would require raising taxes on basic-rate payers, not just higher earners. Using median rather than mean is generally better for pay comparisons; the OECD’s use of means reflects international data availability, not statistical ideal practice. The “13 million menopausal women” estimate is plausible only if it counts all women over 45, but menopause is a retrospective definition and many over-45 women are post-menopausal rather than menopausal. A rough symptom-based estimate suggests far fewer than 13 million women are experiencing peri-/menopausal symptoms at any one time. Corpus evidence suggests “data is” is now dominant in general usage, though “data are” remains more common in academia and traditional edited writing.

Data Points: Re-offending rate (October tweet claim): 23.1% - Dominic Raab said re-offending had fallen to this level from almost 31% in 2009-10. Re-offending rate in 2009-10: almost 31% - Baseline referenced in the minister’s tweet. Fall in re-offending rate: over 2 percentage points - Claimed in the tweet; programme notes the rate later ticked slightly higher. Charge rate: 15.5% to 5.5% - Home Office crime outcomes data; charge share fell by 10 percentage points from 2014-15 to the latest 12 months ending Sep 2022. Simple caution rate: 4.6% to 1% - Home Office data showing a 3.6 percentage point drop since 2014-15. Police officer headcount change: almost 144,000 to about 123,000 - Full-time equivalent police officers in England and Wales fell by around 21,000 from 2010 to 2019. Manifesto pledge on police numbers: 20,000 extra officers - Government pledge cited in response; programme notes it had not yet been fully achieved. Recruitment progress on police pledge: 16,700 extra officers - Most recent figures mentioned, as of January. Welsh higher-rate taxpayers: 166,000 - Correct figure for Wales, replacing the mistakenly cited 9,000. Welsh income taxpayers: 1.5 million - Used to show 11% pay the higher rate in Wales. English income taxpayers in higher-rate band: 16% - Comparison point for Wales. Welsh additional-rate taxpayers: 0.6% - Share of Welsh taxpayers in the top tax band. English additional-rate taxpayers: 2% - Share in England, over three times Wales’s proportion. Revenue from 1p increase in Welsh additional rate: £8 million - Best-case estimate for increasing 45p to 46p. Revenue from 1p increase in Welsh higher rate: £37 million - Best-case estimate for increasing 40p to 41p. Welsh NHS budget: about £10 billion - Used to show the limited impact of higher-rate tax increases. Revenue from 1p increase in Welsh basic rate: £219 million - Illustrates that raising revenue requires taxing the broad base. Median nurses’ pay UK: £37,000 - ONS figure used in the correction on nurses’ pay. Median UK wage: £33,000 - Used to show nurses earn 13% above the median wage. Women over 45 in the UK: just over 13 million - Cited as the likely source of the repeated menopause figure. Female population of the UK: around 34 million - Used to assess the plausibility of the 13 million claim. Average age of menopause: 51 - Reported mean age in UK, US, and Europe. Estimated perimenopause duration: 4 to 8 years - Range given by experts depending on definition. Average duration of hot flushes/night sweats: 7.4 years - From the SWAN study in the US. Median duration of symptoms for African-American women in SWAN: 10 years - Highlighted as longer than the overall average. Rough estimate of women experiencing perimenopause for 4 years: 1.9 million - Back-of-envelope calculation from ONS population estimates. Rough estimate of women experiencing perimenopause for 7 years: 3.3 million - Alternative rough calculation using longer symptom duration. Corpus usage: 'data is' vs 'data are': 80% vs 20% - General corpora findings from Dr Beth Mallory. Academic usage in the 1990s: 93% plural - Historical academic corpus finding for 'data' with plural verbs. Current academic usage: about 50:50 - Modern academic corpus trend described in the episode. Inbox preference on data usage: two-thirds 'data are' - Listener poll from the show’s inbox. Inbox preference opposing view: one-third 'data is' - Listener responses analyzed as an informal corpus.

Pivotal Quotes: "The re-offending rate isn't really the re-offending rate." — Danny Shaw: Explaining that the statistic measures re-convictions and re-cautions, not all offending. "The government hasn't presented any evidence, any cause and effect evidence, to say that the fall in re-offending rates is because of something that they are doing" — Danny Shaw: Critiquing the ministerial claim that policy interventions caused the fall. "The tax base for higher earners in Wales is actually really small." — Vaughan Gething: Used to explain why raising enough NHS funding through higher-earner taxes is difficult.

Implications: The episode warns listeners to scrutinize definitions and data sources before accepting political claims. It also shows how small methodological choices can reshape narratives about crime, taxes, health, and language change.