Office for Nuclear Regulation

Final comment on the J-value approach by Michael Spackman

I am grateful to the HSE and the ONR for the opportunity of a final, one-page comment.

Time and space preclude any item by item reply to the spirited response by Philip Thomas and Roger Jones ('the Response') to my Report for NERA ('the Report'). However, having reread the Response and the Report, I find that the criticisms in the Response all arise from misreadings or incomplete readings of the Report. Several criticisms refer to omissions of issues (such as that of 'delayed deaths') that are not needed to satisfy the Report's remit to 'review the potential contribution of the J-value approach to regulatory decision making'. Many criticisms reflect varying degrees of contextual or sometimes conceptual misunderstanding.

Engineering and hard sciences on the one hand and the social science of economics on the other are all analytical and often very mathematical disciplines. But they sometimes draw on different kinds of data and assumptions and language.

A conspicuous difference in language is in the term 'transparency'. In the usage of Philip Thomas and Niels Lind - invited by Philip Thomas to contribute as an originator of the Life Quality Index (LQI) methodology - any procedure expressed in mathematical terms is per se transparent. A distinguished engineer chairing a Hazards Forum presentation by Philip Thomas firmly supported this. But generally in Whitehall the word has a looser meaning. For example in the J-value presentation the implicit valuation of a fatality is in this looser sense not transparent!

With respect to data and assumptions, early approaches to valuing fatalities regarded people as machines and derived the cost of a death from the subsequent loss of output. This was superseded by two methods accepted as better reflecting people's preferences. One, adopted in the US, is to compare wage levels in occupations of different riskiness (after adjusting of course for other factors). The other method, adopted in the UK among other countries, is to seek people's stated preferences for changes in levels of risk (which is not straightforward, but workable). The J-value derivation is also based, ingeniously, on empirical data, notably on incomes and life expectancy and work/leisure trade offs. But relating these data to the value that people or society place on premature fatalities needs very strong assumptions about human perceptions and behaviour. It is unlikely that any ministry of finance or other body with central responsibility for CBA methodology would see it as a potentially better way forward.

The original LQI work by Niels Lind and others appears to have been an interesting contribution to a responsible initiative to encourage better regulation in some nuclear fields. But the J-value framework, though marketed with intense conviction, outstanding energy and great skill and persuasiveness, seems more of a distraction. As a valuation method it is much less plausible than those established and developed over recent decades. And its presentational framework is markedly less satisfactory for general use within government. As noted briefly in section 6.4 of the Report, there are ways in which resources might be better used to improve the analysis of fatality risks.

Michael Spackman
11 June 2014