^{page 34 note (2)} The next ten years: public expenditure and taxation in the 1990s, Cmnd 9198, HM Treasury, 1984.

^{page 34 note (3)} See T. P. Hill, The measurement of real product, OECD, 1971. The US measures output per head of two thirds of the Federal Government labour force, which grew at 1½ per cent per year between 1967 and 1982 (see ‘Federal Government Productivity 1967-82’, US Bureau of Labor Statistics, 1984), but largely because of problems concerning weighting the results and the difficulty of defining government net output these results are not used in the National Accounts.

^{page 34 note (4)} E. G. Green Paper, Cmnd 9189, chapter I, paragraphs 5 and 6. An alternative would be to deflate by the consumer price index on the grounds that ultimately the burden of financing public spending must fall in one way or another on the private consumer.

^{page 35 note (1)} These figures exclude capital grants in support of private sector investment and (in keeping with our concentration on general government) investment by nationalised industries. They are also net of council house sales.

^{page 35 note (2)} Details of annual growth in volume and cost terms, together with a discussion of public expenditure definitions, are also in V. Imber and P. Todd, ‘Public expenditure: definitions and trends’, Economic Trends, November 1983.

^{page 37 note (1)} Slope coefficients from regressions of the form log G = a + bT and log GDP = a + bT.

^{page 37 note (2)} See F. Gould, ‘The development of public expenditures in western industrialised countries’, Public Finance, no. 1, 1983 and J. P. Dunne, P. Pashardes and R. P. Smith, ‘Needs, costs and bureaucracy: the allocation of public consumption in the UK’, Economic Journal, March 1984.

^{page 37 note (3)} The OECD study of health (1977) provides some evidence on differences in the ratio of doctors' pay to industrial earnings in a sample of countries; the IMF study on government employment offers some data on civil service pay, but the accuracy of the basic data has been seriously criticised by the Treasury: see Hansard, 30 January 1984, col. 16.

^{page 37 note (4)} Suppose wage differences in the traded goods sectors of dif ferent countries reflect productivity differences, high productivity being associated with high incomes per head. Suppose also that labour markets tend to operate within countries so as to equalise the wages of similar types of labour, irrespective of the sector in which it is employed. Suppose also that productivity differences between countries' non-trading sectors are less than in the case of their trading sectors. Because wages in the non-trading sectors follow wages in the trading sectors within countries, it follows that the costs of non-traded goods and services will be higher in high income countries where productivity in the traded goods sector is high. Prices of non-traded goods do not enter (at least not directly) into the determination of market exchange rates.

Since government goods and services are not traded, their average price will tend to be higher in the higher per capita income countries, and unless this is allowed for their volume share in GDP will be biased upwards. (Of course it is always possible that the quality of the (expensive) public employees in the richer countries might be greater than that of their opposite numbers in poorer countries, but we have no evidence on that.) See also, for example, B. Balassa, ‘Purchasing power parity doctrine: a reappraisal’, Jour nal of Political Economy, 1964.

^{page 38 note (1)} This may, to some extent, reflect differences in national accounting conventions.

^{page 38 note (2)} ‘World Product and Income: International Comparison of Real Gross Product’, I. B. Kravis, A. Heston, R. Summers and John Hopkins, 1982.

^{page 38 note (3)} See Appendix for list of countries: government health expen diture is excluded.

^{page 38 note (4)} Data from Kravis et al., op. cit., tables 1-8.