Munich Personal RePEc Archive

Accounting for Japan's Lost Score

Betts, Caroline (2021): Accounting for Japan's Lost Score.


Download (3MB) | Preview


This paper develops a quantitative framework to evaluate the sectoral origins of economic growth. First, I decompose growth in aggregate growth accounting variables–GDP per working age person, a capital factor, an hours’ worked factor, and an implied total factor productivity factor–into sectoral contributions. I decompose the TFP factor growth contribution of a sector into 1) sector-share weighted, within-sector TFP factor growth, and 2) several residual allocative effects. Second, I interpret structurally the observed sectoral contributions by comparing them to those predicted by a multi-sector neoclassical growth model. Using the framework to account for Japan’s economic growth slowdown I find that, empirically, two factors quantitatively dominated Japan’s slowing GDP per working age person in the 1990s. First, a large decline in aggregate TFP growth relative to the 1980s, driven by 1) slower within-industrial sector TFP growth, and 2) negative residual effects due to faster value-added reallocation towards services which mediated a larger impact of the sector for aggregate capital deepening. Second, a large fall in hours worked per working age person, originating mainly in smaller industrial sector contributions. In the 2000s, continued GDP per working age person and aggregate TFP growth decay were due largely to slower within-service sector TFP growth. In the 2010s, anemic aggregate TFP factor growth equal to just 18 percent of its 1980s value was depressed by zero service sector TFP growth; a modest growth rate recovery in GDP per working age person originated in rapid increases in hours worked per working age person, via roughly equal increases in industrial and service sector contributions. A calibrated three-sector growth model absent frictions, featuring sectoral TFP time series as inputs, reproduces closely the time-series from 1980–2018 of a) hours shares of sectors, b) GDP per working age person, and c) the aggregate TFP factor. It captures quite well a) sample-average aggregate TFP growth, b) aggregate TFP growth rate changes across decades, c) the decomposition of aggregate TFP factor growth into total “within-sector” TFP and total residual contributions of sectors, and d) “within-sector” TFP growth contributions of agriculture, industry, and services. The model cannot replicate the sources of, or sectoral contributions to, observed–albeit small–TFP growth residual effects. More importantly, the model’s predicted hours factor (hours per working age person): 1) captures only 46 percent of the decline in industry’s contribution to the fall in aggregate hours factor growth in the 1990s; 2) declines in the 2000s, while hours factor growth is positive in the data; 3) captures only 47 percent of observed average hours factor growth in the 2010s; and 4) allocates too much of the 2010s increase in aggregate hours factor growth to industry. A higher intertemporal elasticity of substitution, a higher Frisch elasticity, and an aggregate labor (policy) wedge resolve some, but exacerbate other, model failures.

Logo of the University Library LMU Munich
MPRA is a RePEc service hosted by
the University Library LMU Munich in Germany.