robertson11sept.jpg
VoxEU Column Labour Markets Productivity and Innovation

Labour reallocation is not why China’s economy is slowing: Busting the myth of surplus peasants in China’s growth

The conventional wisdom is that labour reallocation has been a key driver of China’s growth miracle, and slowing migrant labour flows and rapid wage growth have raised concerns over whether this source of growth has run its course. This column argues that the literature on growth and labour reallocation in China has been dominated by a method that, relative to the now standard growth accounting model, substantially overstates the gains. Allowing for this and for human capital differences across sectors, sectoral labour reallocation has not been a key source of productivity growth in China.

China’s economy is slowing, but will it crash – and if so, when? As a middle-income country with rising wages, many have queried whether China’s old growth model can sustain its future growth (World Bank 2013). In particular, with over 200 million workers migrating from the countryside, the conventional wisdom is that labour reallocation has been a key driver of China’s growth miracle. But slowing migrant labour flows and rapid wage growth have raised concerns over whether this source of growth has run its course. As Krugman (2013) says, China may be about to “hit its Great Wall” as it runs out of “surplus peasants”.

Structural change in China

Krugman’s pessimism hinges on the relationship between labour misallocation and productivity growth. As described in the classic Lewis model of development, large sectoral wage gaps in China imply significant productivity gains from structural change. Figure 1 shows that employment in non-agriculture increased from 30% to 65% over three decades. Likewise, as shown in Figure 2, there has been a 6 to 9-fold difference in output per worker across sectors in China. Against this background most studies find that labour reallocation effects have added around 1.5-2.% per year to China’s growth in output per worker in recent decades.1

Figure 1 Employment share of non-agriculture in China, 1978-2011

Note: For detailed data sources see section 4.2 in Ye and Robertson (2017).

Figure 2 Productivity gap in China, human capital adjusted, 1978-2011

Thus, the data seem to support Krugman’s view. If, say, 2% were cut from China’s total factor productivity (TFP) growth it would be a large shock. Moreover, the impact on GDP growth could be twice as much due to associated contractions in investment and capital accumulation.

In a recent paper, however, we find that as a measure of the TFP gains from labour reallocation in China, these numbers are far too high (Ye and Robertson 2017). This is because (i) whereas most studies use the so-called ‘shift-share’ growth accounting decomposition, this overstates the conventionally defined TFP gains from labour reallocation; and (ii) the existing literature does not allow for sectoral differences in human capital when evaluating the actual sectoral productivity gap. The combination of these two adjustments dramatically reduces the benchmark value of the effect of labour reallocation on China’s growth.

Quantifying the gains from labour reallocation

Most studies that have tried to quantify the impact of labour reallocation on China’s growth have done so using the shift-share method.2 This was first used in the 1940s, but was superseded by the development of multi-sector growth accounting that incorporates neoclassical production functions, diminishing returns, and standard distribution theory that equates factor returns with the value of each factor’s marginal product (Dennison 1961, Kuznets 1967).

We show that the shift-share decomposition method generates ‘reallocation effects’ that are far larger than the reallocation effects generated by the Denison-Kuznets growth accounting. A simple calculation shows that the ratio of the size of labour reallocation effects measured by the Denison-Kuznets growth accounting method, to the shift-share method, is proportional to the labour cost share in non-agriculture. For example, with non-agricultural labour cost share in China of approximately 0.5, the shift-share method will give results that are around twice as large as the Denison-Kuznets growth accounting method.

Intuitively this difference arises because the shift-share method implicitly assumes that production functions in each sector that are linear functions of one input. Hence there is no allowance for diminishing returns to labour as factors move between sectors.

In our paper, we construct growth accounts for China and show that, using the same data, the gains from labour reallocation in China over 1978-2011 are reduced from a value of 1.76% per year if the shift-share method is used to a value of 0.77% using the Denison-Kuznets growth accounting. This also explains why, for example, the well-known paper by Bosworth and Collins (2008) reports reallocation effects that are around twice as large as a recent World Bank study by Bulman and Kraay (2011). The former uses shift-share while the latter uses Denison-Kuznets growth accounting.

Thus, the shift-share method has been used widely and arguably, also by the most prominent studies in the literature. In this sense, the literature on reallocation effects and China’s growth has overstated the amount of TFP growth generated by labour reallocation in China.3

Wage gaps and human capital

We also consider how differences in human capital across sectors might further reduce the estimated gains from labour reallocation. In particular, Li (2014) finds that the urban sector has human capital levels that are 2-3 times higher than the rural sector.

We use these data as an indicator of sectoral differences. Hence, whereas output per worker is 6-9 times higher in non-agriculture compared to agriculture, output per effective worker would only be 3-4 times higher.3 Thus, the productivity gap across sectors is significantly reduced once we account for the difference in skill levels of workers.

Intuitively, this should also affect the measure of growth gains generated by labour reallocation. We find that this reduces the estimated reallocation effects by approximately one-third. Hence the combined impact of using the standard growth accounting approach rather than the shift-share method, and allowing for sectoral human capital differences, is substantial. Over the period 1978–2011, these two adjustments reduce the estimated TFP gains from labour reallocation from 1.76% per year to just 0.25% per year.

Hitting the Great Wall

Thus, the literature on growth and labour reallocation in China has been dominated by a method that substantially overstates the gains relative to the now standard growth accounting model. Allowing for this, and for human capital differences across sectors, we conclude that the gains from labour reallocation have been a far less important source of China's growth than typically portrayed.

Unlike Young’s (2003) “gold to base metals” study of China and East Asia, our smaller productivity numbers don’t mean that productivity growth was unimportant. Rather they show that, according to the standard neoclassical model with competitive factor markets, sectoral labour reallocation has not been a key source of productivity growth in China.

This finding affects how we might think about the impact of slowing rural-urban migration on China’s growth prospects. An historical TFP contribution of 0.25% per year is only a very small fraction of China’s miracle growth rates of 8% per year. The slowing of migration and tightening urban labour market may throw up some policy challenges, but it seems unlikely that these changes are a Great Wall in the way of China’s growth.

References

Bosworth, B, and M S Collins (2008), “Accounting for Growth: Comparing China and India”, The Journal of Economic Perspectives, 22, 45-66.

Bulman, D, and A Kraay (2011), “Growth in China 1978-2008: Factor Accumulation, Factor Reallocation, and Improvements in Productivity”, World Bank Background Paper, 92552.

Denison, E F (1967), Why Growth Rates Differ: Post-war Experiences in Nine Western Countries, Brookings, Washington, DC.

Gollin, D, D Lagakos, and M E Waugh (2014), “The Agricultural Productivity Gap,” The Quarterly Journal of Economics, 129, 939-993.

Krugman, P (2013), “Hitting China's Wall,” The New York Times.

Kuznets, S (1961), “Economic Growth and the Contribution of Agriculture: Notes on Measurement”, International Journal of Agrarian Affairs, 3, 56-75.

Li, H (2014), “Human Capital in China - 2014”, China Human Capital Report Series, China Center for Human Capital and Labor Market Research.

Temple, J, and L Wößmann (2006), “Dualism and Cross-Country Growth Regressions”, Journal of Economic Growth, 11, 187-228.

World Bank (2013), China 2030: Building a Modern, Harmonious, and Creative Society, Washington, DC.

Ye, L, and P E Robertson (2017), “How Important was Labor Reallocation for China's Growth? A Sceptical Assessment”, Review of Income and Wealth, forthcoming.

Young, A (2003), “Gold into Base Metals: Productivity Growth in the People’s Republic of China during the Reform Period”, Journal of Political Economy, 111, 1220-1261.

Endnotes

[1] As discussed below some studies look at total factor reallocation while others focus on labour reallocation. With respect to agriculture and non-agriculture in China over the last 30 years the two numbers are very similar as there has been very little change in the share of capital in each sector.

[2] The shift-share method computes the difference between aggregate GDP per worker growth and the share weighted sum of the growth rate of value added labour productivity in each sector.

[3] The use of shift-share might be justified by a desire to move beyond a standard neoclassical framework. In surveying the broader literature on structural change for example Temple and Wößmann (2006) suggest the shift-share method may capture “broader effects”. We have not seen any such justification by the authors that use shift-share however, except as a “rough approximation”. Ye and Robertson (2017) also show that the shift-share method gives very different results to both the Denison-Kuznets measures irrespective of whether one interprets it as measuring labour reallocation or total factor reallocation effects.

[4] Li (2014) uses a standard Mincerian definition of human capital based on schooling rates. Gollin et al. (2014) find similar differences in schooling across sectors, across a wide range of countries.

2,310 Reads