Paradoxes of labor productivity

We continue to analyze the Total Economy database, TEDI, maintained by the Conference Board. This time we compare two similar estimates of labor productivity. The TEDI provides labor productivity estimates in 2011 EKS dollars per worker (person employed), Pw, and per hour, Ph,  i.e. the ratios of total GDP and the number of workers and the number of hours worked, respectively.  The former estimate depends on the evolution of employment/population ratio and the latter also depends on working hours per person (say, the length of average working week).  We also depict real GDP per capita, G, which formally is not a measure of productivity but provides a conservative reference.

Three figures below present the USA, France, and Australia – the countries with different behavior of labor productivity. To facilitate the analysis, all time series are normalized to their respective values in 1950. In the USA, all three curves had been evolving in sync between 1950 and 1970, when the Pw curve started to lag behind. This deviation is explained by the increasing employment/population ratio between 1965 and 2000.  Since 1978, the G curve has been growing faster than the Ph curve. This is a big surprise which cannot be explained by the increasing length of working week only. As we reported before, the GDP deflator and CPI also started to deviate in 1978. This is not a coincidence and the statistics of real GDP is likely biased. Another important issue is the trend of all variables. We show linear regression lines for all time series. As we discussed many times, real GDP per capita has a linear trend. Both productivity curves also oscillate around linear trends. Currently, the Ph and Pw curves are above the relevant trends and will likely to fall down.  When extrapolated back into the past, all trends intercept 0 between 1916 and 1930. What a surprise!

France has been demonstrating an outstanding growth in Ph; it has grown by a factor of 6.7 since 1950. This is a double US growth rate and twice as large as the overall growth in real GDP per capita in France. In other words, the hours worked to produce the same portion of real GDP per capita have fallen by a factor of 2.  This is an extraordinary performance.  It should be also noted that the G curve is below both productivity curves and all curves have reliable linear trends (regression lines). When extrapolated, the Ph curve intercepts 0 in 1944.  Currently, all curves are below their long term trends. They have to grow at an elevated rate to return to the trends.

In Australia, the G curve was similar to the Pw curve before 1982 and then jumped to the Ph curve.  The linear regression lines intercept 0 between 1924 and 1934 and the observed curves are very close to these lines over the whole period between 1950 and 2011.  The reason they do not follow the trends before 1950 is not clear. Currently, both productivity curves are on their long term trends and the G curve is far above its trend. One might expect real GDP per capita to grow slowly during the next five years.

This is a superficial analysis (more qualitative than quantitative) and we do not pretend to explain all suspicious and obvious features of the evolution of productivity in developed countries. We just illustrate general behavior and stress some problems with data. Quantitatively, the evolution of labor productivity in all developed countries is accurately described in our paper.