Issues in measurement of industry productivity

Accuracy

There is greater uncertainty about the accuracy of measurement of industry sector productivity than there is about the measurement of aggregate (market sector) productivity.

First, there is uncertainty about how accurately the industry of association is identified in underlying data. As a particular example, there is more uncertainty about industry hours worked than there is about aggregate hours worked. If survey respondents mis-identify their industry of attachment, it has consequence for the accuracy of the industry hours worked estimate, but not the accuracy of the aggregate hours worked estimate.

Second, productivity is better measured in some industries than in others. Typically, for example, it is easier to measure the output of goods than the output of services. Broadly speaking, goods have well-defined characteristics that can be objectively determined and quantified, whereas services have a number of quality elements that can be more subjective and difficult to quantify. An error in the measurement of output in an industry may be significant in the estimation of productivity for that industry, but it will lose significance in the estimation of aggregate productivity—especially if the industry is relatively small.

Industry estimates for years from the mid-1990s are likely to be more accurate than estimates for prior years. From the mid-1990s, the ABS has derived industry estimates from annual supply-use tables, which reconcile industry estimates from production, expenditure and income accounts. The estimates for years prior to 1984-85 may be subject to even more measurement error. Measures of hours worked in each industry are not available from the ABS, but have been estimated by Gretton and Fisher (1997).

The ABS provides a more detailed examination of industry level MFP and discusses the general concepts and methodology for measuring MFP in the information paper, Experimental Estimates of Industry Multifactor Productivity, (ABS Cat. No. 5260.0.55.001).

Determining trends

Year-to-year movements in industry productivity, and especially recent movements, should be interpreted with care. First, productivity can be volatile—especially in some industries such as agriculture—so that a movement from one year to the next might not necessarily reflect an underlying trend. Second, there are often revisions to estimates, particularly for the most recent years, as the underlying data are revised by the ABS.

In the aggregate estimates, the ABS measures underlying productivity trends by calculating annual average rates of growth between standard points (peaks) in productivity cycles. This is one way to insulate the detection of trends from spurious short-term effects. However, different industries do not conform to the same cycles. The mining industry, for example, has long cycles that go with investment and extraction phases. If the objective is to compare underlying productivity growth in different industries, the use of trend rates of growth has advantages over the peak-to-peak productivity cycle method. A trend productivity series has been formed for each industry by applying a Henderson 11-period moving average to the original series. Because short-term fluctuations have been smoothed out from the trend series, the start and end points of periods for comparison of growth rates can be selected flexibly and without introducing spurious effects.