Efficiency gains from reallocating human capital between China’s state and non-state sectors

ABSTRACT

This study is among the first to use household survey data from 1988 to 2019 to demonstrate how human capital misallocation between China’s state and non-state sectors has evolved and to estimate efficiency gains from reallocation over time. Our results show that human capital allocation between sectors has been converging to, diverging from, and re-converging to an optimal level, implying that the Chinese labour market has been gradually approaching efficiency during its market-oriented transition. Lower-educated labours had been more misallocated between the two sectors than higher-educated ones. Efficiency gains from reallocating human capital between sectors depend on the degree of misallocation at each stage and the gains are comparable to contemporaneous domestic gross research and development expenditures.

KEYWORDS:

• Human capital misallocation
• efficiency gains
• state sector
• non-state sector
• China

JEL CLASSIFICATION:

• J31
• J24
• P31
• O15

I. Introduction

China’s transition to a market-oriented economy has benefited from its rapidly developing private sector, which has allowed labour to flow from the inefficient state sector to highly efficient non-state sectors.¹ However, little is known about how human capital misallocation between state and non-state sectors evolved during China’s economic transition or the efficiency gains from reallocation. This study examines these two questions.

Previous studies primarily focus on labour misallocations across different industries and regions. Vollrath (2014) finds that reallocating human capital across industries produces less than 5% growth in total economic output for most developing countries. Evidence from China shows that efficiency gains from reallocating labour between the agriculture and non-agriculture sectors range from 0.25% to 1.76% annually between 1978 and 2011 (Ye and Robertson 2018). The gains from reallocating labour across industries have increased total output by 1.79% in 2007 and 1.63% in 2013 (Ma, He, and Li 2018).

Different from previous studies, we demonstrate the evolution of labour misallocation between state and non-state sectors from 1988 to 2019, which may directly reflect the effectiveness of Chinese reform from a planned to market-oriented economy.² We also extend the literature by discussing the different returns to heterogeneous labours and gains from their reallocation.³ This study is among the first to use long-run individual-level survey data to show how labour allocation has evolved over the most important years of China’s economic reform.⁴ Our findings imply that after more than 40 years of economic transition, China is becoming an effective labour market.

II. Evolution of human capital misallocation

The optimal allocation of human capital is achieved when homogenous labour obtains the same wage rate across sectors; that is, there is no wage premium in any sector. However, when homogenous labour is misallocated due to institutional or policy distortions like labour market segmentation or rigid household registration systems, wage-wedges occur across sectors. Following Vollrath (2014), wage-wedges, ${\tau }_j$, can be estimated using (1(1) $ln\left(1+{\tau }_j\right)={\beta }_j-{\sum }_{j=0}^1{\beta }_j{s}_j$(1) ),

(1) $ln\left(1+{\tau }_j\right)={\beta }_j-{\sum }_{j=0}^1{\beta }_j{s}_j$(1)

where j = 1 for the state sector and 0 for the non-state sector. ${\beta }_j$ is the wage in sector j, which will be estimated from a Mincer (2(2) $\mathrm{l}\mathrm{n}{w_i}_j=\alpha +\beta secto{r}_j+X_i^{\prime }\gamma +{\delta }_i+{\theta }_i+{\mu }_i+{\epsilon }_{ij}$(2) ) below. ${\sum }_{j=0}^1{\beta }_j{s}_j$ denotes the social average wage (i.e. the equilibrium wage if labour moves freely), equal to the sum of sector returns ${\beta }_j$ weighted by the share of labour in sector j (s_j). If there is no misallocation between the sectors, ${\beta }_j$ will be zero and so ${\tau }_j$ is zero. Empirically, it is difficult to determine an optimal wage rate for homogenous labour. However, if labour moves across sectors freely and approaches the optimal allocation, the wage of homogenous labour may be close to the social average wage. Therefore, the right-hand-side item of (1(1) $ln\left(1+{\tau }_j\right)={\beta }_j-{\sum }_{j=0}^1{\beta }_j{s}_j$(1) ) implies the actual sector j’s wage deviates from the potential optimum and suggests relative misallocation of human capital. ${\tau }_j$ thus measures the degree of relative misallocation, which is larger than zero if the sector wage has a premium and lower than zero otherwise.

We estimate the following Mincer wage equation to estimate ${\beta }_j$ in (1(1) $ln\left(1+{\tau }_j\right)={\beta }_j-{\sum }_{j=0}^1{\beta }_j{s}_j$(1) ):

(2) $\mathrm{l}\mathrm{n}{w_i}_j=\alpha +\beta secto{r}_j+X_i^{\prime }\gamma +{\delta }_i+{\theta }_i+{\mu }_i+{\epsilon }_{ij}$(2)

where $w_{ij}$ denotes the annual wage of individual i working in sector j. $secto{r}_j$ is a dummy variable that equals 1 for the state sector and 0 otherwise. X_i is a vector of conventional individual characteristics, including gender, years of schooling, and working experience and its quadratic term. We use dummies to further control for occupation (${\delta }_i$), industry (${\theta }_i$), and province (${\mu }_i)$; the final term is an error term. We use Heckman’s two stage approach to alleviate the endogeneities in estimating (2(2) $\mathrm{l}\mathrm{n}{w_i}_j=\alpha +\beta secto{r}_j+X_i^{\prime }\gamma +{\delta }_i+{\theta }_i+{\mu }_i+{\epsilon }_{ij}$(2) ),⁵ and employ whether the spouse works in the state sector as the identification variable in the first stage.⁶ We calculate the share of labour in sector ($s_j$) using our individual survey data. Substituting the estimated values of ${\hat{\beta }}_j$ and $s_j$ into (1(1) $ln\left(1+{\tau }_j\right)={\beta }_j-{\sum }_{j=0}^1{\beta }_j{s}_j$(1) ), we obtain the estimates for ${\hat{\tau }}_j$.

Data used in this study were derived from two sets of representative household survey data – the Chinese Urban Household Survey (1988–2009) and China Household Finance Survey (2011–2019). Urban male employees aged 16–60 and urban female employees aged 16–55 were selected as our sample. After excluding re-employment after retirement, retired persons, household workers, and students, we have 209,957 observations with complete information.

Panel A, Figure 1 demonstrates the evolution in the degree of human capital misallocation (${\hat{\tau }}_j$) in state and non-state sectors during China’s transition. Overall, for labour with the same human capital, before the mid-2010s, the state sector’s annual wage was higher than the social average wage, while the non-state sector’s wage was lower. The evolution of human capital misallocation occurred in three stages. First, the allocation slightly converged to an optimal level between 1988 and 2000 when China began establishing a market-oriented economic system. Second, the state sector allocation began diverging from the optimum in 2001 and reconverged in 2008, while that in the non-state sector also diverged in 2001 but recovered more quickly and significantly. Two major events characterizing this stage are China’s accession to the World Trade Organization and the privatization of state-owned enterprises (SOEs). At that time, SOEs had the privilege to operate import and export businesses, and most private enterprises could not engage in international trade. Therefore, the revenue of SOEs increased rapidly and the wage premium rose accordingly. During the privatization of SOEs, many workers were laid off and became unemployed; therefore, the employment proportion in SOEs declined. However, the non-state sector rapidly expanded, which allowed it to absorb most of those laid off by the state-owned sector; thus, its employment share increased rapidly at this stage. The higher wage premium, lower employment share of the state sector, and higher employment share of the non-state sector were the important developments in the second stage. Third, after 2013, when China officially acknowledged the central role of markets in allocating production factors, non-state sector returns become higher than state sector returns, and human capital allocation began approaching the optimal level.

Figure 1. Evolution of human capital misallocation in state and non-state sectors in China, 1988–2019.

We assume that each human capital unit is substitutable in the above analyses. To extend previous studies, we relax this assumption and assume that sector returns to labour with high (tertiary education) and low (secondary education and below) human capital stock differ; thus, they are not perfectly substitutable. We calculate the degree of misallocation for higher- and lower-educated labour by re-estimating (2(2) $\mathrm{l}\mathrm{n}{w_i}_j=\alpha +\beta secto{r}_j+X_i^{\prime }\gamma +{\delta }_i+{\theta }_i+{\mu }_i+{\epsilon }_{ij}$(2) ) and (1(1) $ln\left(1+{\tau }_j\right)={\beta }_j-{\sum }_{j=0}^1{\beta }_j{s}_j$(1) ), respectively. The results in Panel B, Figure 1, suggest the misallocation of human capital is primarily attributable to misallocating lower-educated labour. A higher wage premium for lower- than higher-educated labour is especially obvious in the state sector. The wage gap between higher- and lower-educated labour is lower in the state than the non-state sector, implying the state sector’s misallocation may be due to redundant labour with low productivity. By contrast, even in the non-state sector’s early development, its labour market performance was much more market-oriented: the returns on higher-educated labour were higher than those on lower-educated labour. The wage premium in the non-state sector since the mid-2010s is largely due to higher returns on higher-educated than lower-educated labour.

III. Efficiency gains

What would the efficiency gains be if human capital were reallocated between the sectors? To answer this question, we follow Vollrath (2014) methodology to construct a ratio, R ((3(3) $R=\frac{Y^{optional}}{Y^{observed}}={\frac{\left[\sum _j{\left(\frac{w\left(1+{\hat{\tau }}_j\right)}{1+\alpha }\cdot H_j^{\alpha }\right)}^{\frac{1}{\alpha }}\right]}{{\left[\sum _j\frac{{\left(\frac{w\left(1+{\hat{\tau }}_j\right)}{1+\alpha }\cdot H_j^{\alpha }\right)}^{\frac{1}{\alpha }}}{{\left(1+{\hat{\tau }}_j\right)}^{\frac{1}{\alpha }}}\right]}^{\alpha }}}^{\alpha }=\frac{{\left[\sum _j{\left(\frac{w\left(1+{\hat{\tau }}_j\right)}{1-\alpha }\cdot {\left[\sum _{i\in j^{exp\left(ln{\hat{w}}_{ij}-\hat{\alpha }-\hat{\beta }secto{r}_j\right)}}\right]}^{\alpha }\right)}^{\frac{1}{\alpha }}\right]}^{\alpha }}{{\left[\sum _j\frac{{\left(\frac{w\left(1+{\hat{\tau }}_j\right)}{1-\alpha }\cdot {\left[\sum _{i\in j^{exp\left(ln{\hat{w}}_{ij}-\hat{\alpha }-\hat{\beta }secto{r}_j\right)}}\right]}^{\alpha }\right)}^{\frac{1}{\alpha }}}{{\left(1+{\hat{\tau }}_j\right)}^{\frac{1}{\alpha }}}\right]}^{\alpha }}$(3) ) below), which is the optimal total output (when there is no wage wedge between the two sectors) divided by actual total output. Specifically, we use a Cobb-Douglas production function and assume all wedges between the sectors come from the misallocation of human capital. That is, we evaluate the efficiency gains from reallocating human capital, holding technology and capital stock constant.⁷ Therefore, most parameters in the production function except the stock of human capital (H_j) cancel out in the R ratio. The divergence of the actual total output from the optimum is derived from the wage wedge (${\hat{\tau }}_j$). The stock of human capital (H_j) in sector j is the sum of the human capital stock of each individual working in sector j, which is the residual of the actual wages excluding sector wages shown in (3(3) $R=\frac{Y^{optional}}{Y^{observed}}={\frac{\left[\sum _j{\left(\frac{w\left(1+{\hat{\tau }}_j\right)}{1+\alpha }\cdot H_j^{\alpha }\right)}^{\frac{1}{\alpha }}\right]}{{\left[\sum _j\frac{{\left(\frac{w\left(1+{\hat{\tau }}_j\right)}{1+\alpha }\cdot H_j^{\alpha }\right)}^{\frac{1}{\alpha }}}{{\left(1+{\hat{\tau }}_j\right)}^{\frac{1}{\alpha }}}\right]}^{\alpha }}}^{\alpha }=\frac{{\left[\sum _j{\left(\frac{w\left(1+{\hat{\tau }}_j\right)}{1-\alpha }\cdot {\left[\sum _{i\in j^{exp\left(ln{\hat{w}}_{ij}-\hat{\alpha }-\hat{\beta }secto{r}_j\right)}}\right]}^{\alpha }\right)}^{\frac{1}{\alpha }}\right]}^{\alpha }}{{\left[\sum _j\frac{{\left(\frac{w\left(1+{\hat{\tau }}_j\right)}{1-\alpha }\cdot {\left[\sum _{i\in j^{exp\left(ln{\hat{w}}_{ij}-\hat{\alpha }-\hat{\beta }secto{r}_j\right)}}\right]}^{\alpha }\right)}^{\frac{1}{\alpha }}}{{\left(1+{\hat{\tau }}_j\right)}^{\frac{1}{\alpha }}}\right]}^{\alpha }}$(3) ). R is 1 if there is no misallocation, larger than 1 if the actual output is less than optimal and the state sector has a wage premium, and less than 1 if the non-state sector has a wage premium. The more R deviates from 1, the greater the human capital misallocation and thus, the higher the efficiency gains from reallocation. α is the elasticity of labour demand, which is commonly set at 0.3 (Hamermesh 1993).⁸

(3) $R=\frac{Y^{optional}}{Y^{observed}}={\frac{\left[\sum _j{\left(\frac{w\left(1+{\hat{\tau }}_j\right)}{1+\alpha }\cdot H_j^{\alpha }\right)}^{\frac{1}{\alpha }}\right]}{{\left[\sum _j\frac{{\left(\frac{w\left(1+{\hat{\tau }}_j\right)}{1+\alpha }\cdot H_j^{\alpha }\right)}^{\frac{1}{\alpha }}}{{\left(1+{\hat{\tau }}_j\right)}^{\frac{1}{\alpha }}}\right]}^{\alpha }}}^{\alpha }=\frac{{\left[\sum _j{\left(\frac{w\left(1+{\hat{\tau }}_j\right)}{1-\alpha }\cdot {\left[\sum _{i\in j^{exp\left(ln{\hat{w}}_{ij}-\hat{\alpha }-\hat{\beta }secto{r}_j\right)}}\right]}^{\alpha }\right)}^{\frac{1}{\alpha }}\right]}^{\alpha }}{{\left[\sum _j\frac{{\left(\frac{w\left(1+{\hat{\tau }}_j\right)}{1-\alpha }\cdot {\left[\sum _{i\in j^{exp\left(ln{\hat{w}}_{ij}-\hat{\alpha }-\hat{\beta }secto{r}_j\right)}}\right]}^{\alpha }\right)}^{\frac{1}{\alpha }}}{{\left(1+{\hat{\tau }}_j\right)}^{\frac{1}{\alpha }}}\right]}^{\alpha }}$(3)

Table 1 presents the percentage changes in R. The economic gains shown in column 1 indicate that reallocating homogenous labour between the state and non-state sectors would increase total output by around 1% from the late 1980s to the early 2000s and by 2% or more during most of the next decade. The gains decrease following that and would approach zero recently since the labour market has become increasingly more efficient. Intuitively, the percentage of output gained from reallocation is larger than or comparable to the contemporaneous share of R&D expenditures to GDP (column 4) before the gains gradually become trivial.

Table 1. Efficiency gains from human capital reallocation between state and non-state sectors (percentage of total output).

Year	Percentage Changes in R			R&D Expenditure/GDP (%)
	Total Labour	Higher-educated Labour	Lower-educated Labour
1988	1.12	0.05	1.23	0.88
1989	0.85	0.09	1.11	0.83
1990	1.26	0.06	1.55	0.82
1991	0.97	0.12	1.06	0.81
1992	0.89	0.18	0.77	0.77
1993	1.06	0.09	1.12	0.72
1994	1.39	0.21	1.45	0.64
1995	1.38	0.31	1.25	0.57
1996	1.24	0.13	1.15	0.56
1997	1.40	−0.05	1.50	0.64
1998	1.23	0.00	1.17	0.65
1999	1.51	0.28	1.14	0.75
2000	0.97	0.04	1.26	0.89
2001	1.26	0.10	1.54	0.94
2002	2.26	0.41	3.24	1.06
2003	2.04	0.53	2.88	1.12
2004	2.19	0.61	3.65	1.21
2005	2.24	0.47	3.89	1.31
2006	2.53	0.57	4.09	1.37
2007	2.24	0.94	3.06	1.37
2008	2.58	0.51	4.28	1.45
2009	1.62	0.43	3.09	1.66
2011	1.04	0.02	3.15	1.78
2013	0.09	0.02	−0.05	2.00
2015	−0.03	0.00	0.00	2.06
2017	−0.09	0.30	0.00	2.12
2019	0.01	0.02	0.17	2.24

Note: Percentage changes in R are authors’ own calculations. Data for R&D expenditures and GDP were collected from Chinese Statistical Yearbooks, various years.

To further consider the heterogeneous gains from reallocating more and less productive labour (column 2 and 3), we find that reallocating less productive labour would yield much higher total output, consistent with the observation from Figure 1 that the allocation of lower-educated labour deviates more from the optimal level. Economic gains were highest between 2002 and 2008—almost three times R&D expenditures.

IV. Conclusions

To the best of our knowledge, this study is the first to demonstrate the evolution of misallocation of human capital between state and non-state sectors and illustrate the economic gains from reallocation after China began its market-oriented reforms. Our results suggest that after 40 years of economic transition, China’s labour market has been gradually approaching efficiency. The economic gains from reallocation are comparable to gross R&D expenditures in China. Our findings improve the understanding of what Chinese labour market reforms have achieved.

Acknowledgement

This work was supported by the National Social Science Fund of China (18BJY051) and the Fundamental Research Funds for the Central Universities. The authors thank David Peel and the anonymous reviewers for their valuable comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The work was supported by the National Office for Philosophy and Social Sciences [18BJY051]; Fundamental Research Funds for the Central Universities.

Notes

¹ The total factor productivity (TFP) of China’s state sector has been increasing during the transition but is still lower than the non-state sector’s (Brandt, Biesebroeck, and Zhang 2012).

² Brandt, Biesebroeck, and Zhang (2012) report that the TFP loss from within-province distortions due to misallocation of employment between state and non-state sectors are less than 3% from 1988 to 2006. Ge and Li (2019) find that the TFP of industrial enterprises have increased by 37% in 1998 and 30% in 2007 after reallocating human capital across enterprises with different ownerships.

³ Most studies estimate TFP increases by reallocating the number of employment (i.e. labour quantity) rather than the units of human capital (i.e. labour quality) across industries, firms, or regions (Brandt, Tombe, and Zhu 2013; Gong and Hu 2013; Gai et al. 2015; Li and Wang 2021).

⁴ Previous studies employing macro-level data show the labour misallocation across sectors and regions over two decades (Brandt, Biesebroeck, and Zhang 2012; Ye and Robertson 2018); although, they do not reflect the heterogeneities at micro-level. Some studies further use firm-level data (Gong and Hu 2013; Gai et al. 2015; Ge and Li 2019); although, they only focus on a short Chinese transition period due to limited data availability (mostly from 1998 to 2007). Vollrath (2014) and Ma, He, and Li (2018) use the individual-level data but they focus on estimating the misallocation across industries. In addition, Vollrath (2014) does not include China and Ma, He, and Li (2018) only include data in 2007 and 2013.

⁵ If the more capable individuals are more likely to work in the state sector and the capability is hard to control in estimations, then endogeneities arise.

⁶ Following Zhao (2002), we use the same identification variable and estimate the probability of individuals choosing to work in the state sector. The inverse Mills Ratio estimated in the first stage is then put into (2(2) $\mathrm{l}\mathrm{n}{w_i}_j=\alpha +\beta secto{r}_j+X_i^{\prime }\gamma +{\delta }_i+{\theta }_i+{\mu }_i+{\epsilon }_{ij}$(2) ) for the second stage regression.

⁷ The economy gains would be higher if we examine the misallocation of both human capitals and physical capitals.

⁸ We also estimate efficiency gains by setting α equal to 0.2 and 0.45, respectively, and the results are robust.