There has been much economic and political comment on the "Japanese style' productivity growth in British manufacturing beginning in 1981. These figures have been cited as evidence of a productivity breakthrough. They have also been claimed as a success for the economic policies of Mrs. Thatcher's first administration. Using published productivity figures to determine whether this shift is permanent is difficult. Such data contain short-term cyclical variations in labour utilization that can easily be misinterpreted as long-term improvements.

CEPR Research Fellow John Muellbauer presented estimates of the underlying trend in UK manufacturing productivity growth at a lunchtime discussion meeting on 3 December. He estimated annual productivity growth of 3.4% for 1955-73, 1.4% for 1973-79, -0.3% for 1979-80 and 2.9% for 1980-83. These figures, he argued, suggested that the 1973 deceleration and the 1980 acceleration in productivity were genuine . There were also indications of a slightly faster underlying rate of productivity growth in 1983, though this was not statistically significant.

Muellbauer outlined research done jointly with Lionel Mendis, and reported in CEPR Discussion Papers No. 32 and 34. They estimated an aggregate production function for British manufacturing using quarterly data for 1956-83. They had encountered a number of problems in the measurement of labour utilization, of output and of capital. For the first two they were able to find economic variables to solve the measurement problems. Measurement problems in capital are, however, by nature more trend-like. In the absence of appropriate economic indicators, they used shifting time trends to proxy capital measurement biases. Muellbauer conceded that the procedure was necessarily more speculative, given the other factors that trend shifts could be reflecting. With these techniques, they were able to construct a measure of productivity growth corrected for the very substantial variations that took place in labour utilization and for the relatively less important measurement biases in output. This gives a very different picture of short-run movements from the conventional measure of growth in output per head. It is also much more informative about underlying trends.

Their labour utilization measure is based on overtime hours relative to normal hours. Muellbauer argued that this is a good indicator of above-normal utilization levels. Since most workers receive payment for a standard week even when underutilized, no corresponding direct measure of underutilization exists. A statistical aggregation argument suggests, however, that the average degree of utilization can be deduced as a non-linear function of the proportionate overtime data. Comparing their measure with a capacity utilization index derived from the CBI Industrial Trends survey and with Bennett and Smith-Gavine's "Percentage Utilization of Labour' index reveals a broadly similar cyclical pattern. Muellbauer argued, however, that his measure produces more satisfactory econometric results in modelling output.

Economists prefer to measure productivity by value added per worker. The increase in raw material prices after 1973 would have caused substitution away from raw materials, making gross output grow more slowly than value added. The Central Statistical Office (CSO) uses gross output data to measure value added. Productivity figures based on gross output data may therefore have a downward "gross output bias' after 1973. According to Muellbauer's estimates, however, this bias appears to be small. There is only a 2% understatement of 1974 output relative to 1970.

Other output measurement biases considered by Muellbauer stem from potential measurement errors in the price deflators used by the CSO to deflate the output measure to obtain "real' output.

Unfortunately, no reliable indices of export prices exist to deflate output sold abroad. In 1981 the Rayner cuts in the Civil Service brought to an end a promising pilot project to construct accurate export price indices. The CSO use domestic price indices instead. These can diverge significantly from export prices when the real exchange rate changes sharply, and this can lead to what Muellbauer called the "domestic price bias'.

Domestic price indices are used by the CSO to deflate output, but they may not fully reflect the prices at which goods are actually traded. Muellbauer argued that list prices are likely to respond less quickly to changes in costs and in competitive pressures. This can give rise to a "list price bias' in measured output.

The "gross output bias' depends on the ratio of raw materials prices to output prices. The "domestic price' and "list price biases' can be proxied (represented) by the same price ratio; by competitiveness, as measured by the ratio of foreign to domestic output prices; and by rates of change of each ratio.

In the production function relating output to inputs, capital is a key ingredient: obviously output per head is higher when more capital is available. The CSO measure of the gross capital stock assumes that the efficiency of assets in use is constant as they age and that the ages at which capital assets are scrapped are independent of economic conditions. The implausibility of these assumptions, Muellbauer suggested, meant that the capital stock figures are subject to large measurement errors. In Muellbauer's estimated production function these errors are proxied by shifts in time trends.

Muellbauer argued that the "domestic price' and "list price biases' can significantly distort the output figures in the short run. For example, in the 1st quarter of 1983 following the sharp but temporary depreciation of Sterling, he estimated that measured output grew 2.0% while true output grew only about 1%. More recently, official figures on growth between the 3rd quarters of 1983 and 1984 overstate true growth in manufacturing by between 0.2 and 0.3 percentage points.

Output per head in British manufacturing certainly grew more slowly in 1973-80 than it had in the previous two decades. Muellbauer asked: Was the slowdown genuine? If so, what caused it? He found it unlikely that a decline in labour utilization alone could explain such an extended slowdown.

Muellbauer's "' productivity grew 3.4% p.a. between the first quarters of 1955 and 1973 but slowed to 1.4% p.a. between the first quarter of 1973 and the third quarter of 1979. Of this 2% slowdown, 0.4% can be explained by slower growth of measured capital. Muellbauer had two explanations of the remaining deceleration. It was possible that, in their struggle to maintain real wage growth in difficult conditions, workers increased the effectiveness of restrictive practices, thus slowing productivity growth. Muellbauer thought a more plausible explanation was that the true capital stock grew by much less than the measured stock. Capital may have been scrapped in response to changed relative prices and slack demand conditions. If so, this "unrecorded' scrapping could have been as high as 27% of the 1973 capital stock.

Output per head fell sharply from 1979 to 1980. Muellbauer attributed this in part to the 16% collapse of manufactured output in 1979-80. This decline was the second sharpest in British history, exceeded only by the 1920-21 collapse which ushered in the long period of inter-war depression. Not surprisingly, firms in 1979-80 were unable to shed labour as rapidly as output fell. As a result, the labour force on their books was severely underutilized in 1980, and conventionally measured productivity fell. In the aftermath, as firms caught up in their shedding of labour, utilization and with it output per head increased. Muellbauer commented that if a government aim had been to reduce overmanning, in the short run its policies paradoxically contributed in 1980-81 to the most severe level of overmanning in living memory.

Muellbauer's measure of "corrected productivity' fell by 0.3% between the 3rd quarters of 1979 and 1980. The most plausible explanation according to Muellbauer was again unrecorded scrapping of perhaps 14% of the 1979 capital stock, coinciding with a drop in output of over 16%.

From 1980 quarter 3 to the end of 1983, Muellbauer's "corrected productivity' measure grew at a 2.9% annual rate, which was less than for 1955-73. Since the recorded (and probably the true) capital stock grew more slowly than in 1955-73 because of low rates of investment, however, productivity growth is liable to increase with the investment recovery now under way. Muellbauer suggested that the acceleration in productivity already seen might be due to the end of the high levels of unrecorded capital scrapping which began in 1973. In 1983 and 1984 scrapping may have been lower than indicated by the CSO capital stock figures. The weakened power of unions to enforce restrictive practices or the microchip revolution and the spread of computer-controlled machine tools may have played a role as well.