Thanks to a recommendation from Martin Wolf, I’ve just read [amazon_link id=”0521094461″ target=”_blank” ]Theoretical Welfare Economics[/amazon_link] by J de V Graaff, a 1971 reprint of a 1957 primer on this subject – Martin told me it had been his university text. While in my undergraduate course we certainly covered welfare economics, it was as a settled body of knowledge, and I don’t recall reading anything like this on the earlier debates.

[amazon_image id=”0521094461″ link=”true” target=”_blank” size=”medium” ]Theoretical Welfare Economics[/amazon_image]

It’s a very clear and interesting discussion – succinct too, at about 150 pages. One of the most interesting aspects is how comprehensively the author demolishes the idea that questions about the size of the economic pie and its distribution can be separated. He lost the argument in later economics of course – economists assume, and often state, that these are separable. However, the strength of Graaff’s counter-argument is evident. At the formal level, he covers the inconsistency of the Kaldor/Hicks notion that losers from a change in allocation can (potentially) be compensated – Skitovsky originally showed that the (potential) Pareto optimality of any change depended on the initial distribution. (Here is a good Interfluidity explanation.) But the book also explains it far more intuitively:

“In a one commodity world some definite meaning could be attached to a phrase like ‘the size of national income’; and we could legitimately  say that welfare depended on the size and the distribution of this one commodity. But as soon as we leave a one-commodity world this ceases to be true. There is no unambiguous meaning we can attach to ‘the size of national income’ when we have a heterogeneous collection of goods and services. How can we combine the various goods into a single quantity that can be said to have a ‘size’? By weighting them and striking an average? This is certainly a possibility. But we can only get the relevant weights from a welfare function; and if we have the usual Paretian one … it will only tell us what weights to use when the distribution of goods among members of the community is given. Only in a very limited sense can welfare be said to depend on ‘size’ and ‘distribution’ – for the two elements are no longer independent and cannot be separated out.” [my italics]

He adds that the index numbers usd in constructing national income cannot be an indicator of change in welfare – they simply provide information relevant to a balanced judgement. “Index numbers of aggregate output or consumption should always be supplemented with information about the distribution of income and wealth – and also with separate indexes of investment, personal and collective savings, and expenditure on collective goods like defence. The more informatio made available, the more likely it is that a balanced judgement will be obtained.”

So here I think we have one of the earliest arguments for the ‘dashboard’ approach to measuring economic progress. But also an irrefutable case – with as many singing and dancing cross-partial derivatives as you like – for never leaving income distribution out of an assessment of how the economy is doing.