Information about Asia and the Pacific Asia y el Pacífico

Chapter 9. Chronicle of a Decline Foretold: Has China Reached the Lewis Turning Point?

Anoop Singh, Malhar Nabar, and Papa N'Diaye
Published Date:
November 2013
  • ShareShare
Information about Asia and the Pacific Asia y el Pacífico
Show Summary Details
Mitali Das and Papa N’Diaye 

The final section of the book examines what needs to be done to complete the process of domestic rebalancing. Impending demographic changes make reform ever more urgent. China is on the eve of a demographic shift that will have profound consequences on its economic and social landscape. Within a few years the working-age population will reach a historical peak, and then begin a precipitous decline. This fact, along with anecdotes of rapidly rising migrant wages and episodic labor shortages, has raised questions about whether China is poised to cross the Lewis Turning Point, a juncture at which it would move from a vast supply of low-cost workers to a labor shortage economy. Crossing this threshold will have far-reaching implications for both China and the rest of the world. This chapter empirically assesses when the transition to a labor shortage economy is likely to occur. The central result is that on current trends, the Lewis Turning Point will emerge between 2020 and 2025. Alternative scenarios—with higher fertility, greater labor force participation rates, financial reform, or higher productivity—may peripherally delay or accelerate the onset of the turning point, but demographics will be the dominant force driving the depletion of surplus labor.


China’s large pool of surplus rural labor has played a key role in maintaining low inflation and supporting China’s extensive growth model. In many ways, China’s economic development echoes Sir Arthur Lewis’s model, which argues that in an economy with excess labor in a low-productivity sector (agriculture in China’s case), wage increases in the industrial sector are limited by wages in agriculture, as labor moves from the farms to industry (Lewis, 1954). Productivity gains in the industrial sector, achieved through more investment, raise employment in the industrial sector and the overall economy. Productivity running ahead of wages in the industrial sector makes the industrial sector more profitable than if the economy were at full employment, and promotes higher investment. As surplus labor in agriculture is exhausted, industrial wages rise faster, industrial profits are squeezed, and investment falls. At that point, the economy is said to have crossed the Lewis Turning Point (LTP) (Figure 9.1).

Figure 9.1.The Lewis Turning Point and Its Implications

Anecdotal evidence of rapid nominal wage increases and episodic labor shortages have raised questions about whether the era of cheap Chinese labor is coming to an end and whether China has reached the LTP. China’s crossing the LTP would have consequences for both China and the rest of the world. For China, it would mean that the current extensive growth model that relies so heavily on factor input accumulation could not be sustained and that China would need to invest less, but in better, capital. China would thus need to switch to a more “intensive” growth model with a greater reliance on improving total factor productivity (TFP), which means accelerating implementation of the government’s agenda to rebalance growth away from investment toward private consumption.

Successfully rebalancing China’s growth pattern would yield significant positive external spillovers to the rest of the world, potentially raising output, particularly in those economies within the supply chain (mainly emerging Asia) and commodity exporters, and somewhat more limited spillovers to advanced economies (IMF, 2011). Moreover, rising labor costs—the impact of which will be felt in prices and corporate profit margins in China—will have implications for trade, employment, and price developments in key trading partners.

Against this backdrop, this chapter presents estimates of China’s excess labor supply using a general procedure developed in Rosen and Quandt (1978), Quandt and Rosen (1986), and Rudebusch (1986). The empirical model attributes an explicit role to population composition, labor force participation, and productivity—characteristics of the Chinese labor market most likely to be relevant in an analysis of excess supply. The remainder of the chapter is organized as follows: The next section provides a brief overview of recent trends in China’s labor market. The subsequent section presents the empirical framework and is followed by a section that presents baseline results. A scenario analysis around a central baseline forecast of future trends in the labor market is presented in the penultimate section, and the last section concludes.

Recent Developments

China’s labor markets first came into focus in 2004, when reports of migrant labor shortages in the export-oriented coastal regions began to emerge. Anecdotal evidence of large increases in migrant wages in this period was viewed by some as an indication that China had depleted a previously vast supply of surplus labor, and the transition to a mature economy had begun (Garnaut, 2006; Cai and Wang, 2008). This view was bolstered by the demographic transition, in particular, a slowdown in the growth of the working-age population and a rising proportion of elderly to young Chinese workers. An alternative perspective, however, viewed rising migrant wages as a consequence of labor market segmentation (the result, for instance, of the household registration system, limited portability of benefits, and rising rural reservation wages), thus consistent with the coexistence of shortages on the coast and surplus labor inland (Chan, 2010; Zhang, Yang, and Wang, 2010; Knight, Deng, and Li, 2011).

Developments in the Chinese labor market since 2009 are reminiscent of the episode in the middle of the first decade of the 2000s—after a lull, wage increases have again been rapid, especially in inland provinces as companies relocate from the coast. However, the authors’ reading of recent developments in Chinese labor markets suggests no conclusion about whether surplus Chinese labor has been exhausted. Despite localized signs of labor market pressures, aggregate nominal wage growth has ranged between 12 and 15 percent annually since 2000. Corporate profits have remained high, and even rose during 2009–11, as wage growth has trailed productivity gains. These developments appear inconsistent with the basic premise of the LTP depicted in Figure 9.2 (panel a). Anecdotally, urban employers point to frictions such as skill and geographic mismatches, not labor scarcity itself, as causes of intermittent, localized wage escalation. Systematic labor shortages on the industrial coast would be reflected in divergence between coastal and industrial wage growth, but wage developments are to the contrary (Figure 9.2, panel c).

Figure 9.2Labor, Wage, and Demographic Changes in China

(panels a and b)
(panels a and b)
(panels a and b)

Source: CEIC; and IMF staff estimates.

Source: IMF staff estimates.

Other evidence, however, suggests tightening labor market conditions. Urban registration records indicate that the margin between city demand and supply of labor has progressively narrowed and is now effectively closed (Figure 9.2, panel b).1 The pace of industry relocation to the interior provinces—where wages are lower and the large reserve of rural labor resides—has picked up since the 2008–09 global financial crisis.2 Parallel developments, such as an uptick in labor activism since the financial crisis, are also consistent with the strengthened bargaining power that accompanies a shrinking pool of labor. The rise in wages also reflects the government’s decision to increase minimum wages as a means to support household income and promote consumption.

Thus, overall, labor market developments paint a mixed picture of excess labor: Wage developments do not suggest exhaustion of surplus labor, whereas employment, industrial relocation, and some policies signal tightening conditions.

However, demographics more forcefully suggest an imminent transition to a labor shortage economy. China is poised to undergo a profound demographic shift within the next decade, driven by the mutually reinforcing phenomena of declining fertility and aging. The UN projects that growth of the working-age (ages 15–64) population will turn negative about 2020 (Figure 9.2, panel d). This forecast potentially understates prospects for a labor shortage because industry employees are predominantly young (Garnaut, 2006); the growth rate of the core 20–39-year-old subpopulation, for example, shrank to zero in 2010 (Figure 9.3), and is projected to decline faster than the overall working-age population through 2035. Population data also show that after a protracted period of “demographic dividend,” the share of dependents—those ages younger than 15 and older than 64 in China’s population—bottomed out in 2010, and will rise to nearly 50 percent by 2035 (Figure 9.2, panel e).

Figure 9.3Demographic Pressures

Sources: UN Population Database; and IMF staff estimates.

Because the looming demographic changes are large, irreversible, and inevitable in the medium term, they will be key to the evolution of excess labor in China. Other factors could, however, be pivotal in accelerating or slowing this process. More progress in hukou reform (easing the ability of workers in provinces to move to cities) could spur rural labor to move to the city. Training rural workers to meet the skill requirements of industrial jobs could decongest urban labor bottlenecks. Although the Chinese primary sector employs nearly half of the labor force, agricultural value added was only about one-fifth of 2011 GDP. Raising agricultural productivity—by raising mechanization to comparators’ levels, for instance—could result in a sizable release of rural workers that could partially offset shortfalls in urban labor demand.

In summary, the combined implications of demographics, labor developments, and policies suggest that China likely is on the eve of the LTP, but give little indication about when the transition will occur. Against this background, the next section attempts to gauge China’s excess labor supply and project its likely evolution.

Empirical Framework

To quantify excess labor supply in China, a first step is to identify the appropriate analytical framework. One approach (e.g., Lucas and Rapping, 1969; Barro and Grossman, 1971) is the conventional simultaneous equation model. This approach assumes the labor market is in equilibrium, that is, that real wages clear the labor market and unemployment results from labor market frictions, such as mismatch, preferences (i.e., intertemporal substitution) and government policies. An alternative framework is the disequilibrium approach, which assumes that the observed real wage does not clear the labor market (Rosen and Quandt, 1978). Instead, in this approach the observed quantity of employment is the minimum of the notional supply of and demand for labor, and unemployment results from excess labor supply, that is, the supply of labor exceeds demand at the observed real wage (see, e.g., Quandt and Rosen, 1986; Hajivassiliou, 1993). In practice, this means that the excess supply of labor includes both the actual unemployed and underemployed, which would encompass part of China’s large pool of migrant workers. Survey data from the National Bureau of Statistics show that about 169 million migrant workers were seeking jobs outside their home provinces as of end-September 2012. The number of unemployed people stood at 21.5 million as of end-2011, amounting to about 3 percent of the total labor force (the unemployment rate in urban areas is 4.1 percent).

Stylized facts of the Chinese labor market described above, in particular, wage growth that has trailed productivity growth for more than a decade, in addition to a rural share of the labor force near 50 percent and a labor share of income that is both low and has declined some 20 percentage points since 1980 (Aziz and Cui, 2007), suggest that the disequilibrium framework is well suited for analyzing the Chinese labor market. Moreover, this framework is sufficiently general to nest the equilibrium approach as a limiting case. The main equations used in the analysis are described in Annex 9A; details of the approach are in Rosen and Quandt (1978), Quandt and Rosen (1986), and Rudebusch (1986).


The model is estimated using annual observations from 1992 through 2010. The dependent variable, L, is the natural logarithm of the total number of employees in urban and rural areas, including those in the government sector, state-owned enterprises, and the private sector. The wage variable, W, is nominal aggregate wages in billion renminbi deflated by the consumer price index (CPI). Wealth is measured as households’ nominal net financial assets, while TFP is residually calculated from a standard growth accounting model with fixed labor and capital shares. Population, labor force, and working-age data are from the United Nations population database. A full list of the sources and definitions of the variables is given in Table 9.1.

Table 9.1Variable Definitions and Sources
LLn of total employment; millionsCEIC Data
WLn of aggregate annual nominal wages deflated by CPI; billion renminbiCEIC Data
GDPpLn of weighted average of real GDP growth in Chinese trade partners; percentWEO, DOTS, IMF staff calculations
TFPTotal factor productivity, residually calculated from growth accountingCEIC Data, WEO, IMF staff calculations
HParticipation rate times population; millionsUN, WDI
WealthNet household financial wealth; 100 million renminbiHaver Analytics
UUnemployment rate; percentCEIC Data
Note: CPI = consumer price index; DOTS = Direction of Trade Statistics; Ln = natural log; WDI = World Development Indicators; WEO = World Economic Outlook.
Note: CPI = consumer price index; DOTS = Direction of Trade Statistics; Ln = natural log; WDI = World Development Indicators; WEO = World Economic Outlook.

Results are reported in Table 9.2 with estimated standard errors in parentheses.3 Overall, the results are of the expected signs and of plausible magnitudes. The wage elasticity of labor demand is negative. Furthermore, the absolute value of the elasticity of labor demand with respect to real wages is less than 1, consistent with the stylized fact that labor costs in China represent a low fraction of total firm costs. The effect of TFP, which is assumed to raise labor demand by increasing profitability, is positive, as expected, while trading partner GDP also has the expected positive sign but is statistically insignificant.4

Table 9.2Estimated Regression Coefficients for Excess Labor
Labor demand
Labor supply.050**
Participation Rate × population.0005**
Excess supply indicator.018**
Number of observations36
Pseudo R-squared.82
Note: Dependent variable: Employment (second stage). Standard errors in parentheses. See footnotes 3 and 4 in the text for additional details about the estimator.

denotes statistical significance at the 10 percent error level;

significance at the 5 percent error level.

Note: Dependent variable: Employment (second stage). Standard errors in parentheses. See footnotes 3 and 4 in the text for additional details about the estimator.

denotes statistical significance at the 10 percent error level;

significance at the 5 percent error level.

Estimated supply-side coefficients are also consistent with theory. The wage elasticity of labor supply is positive and smaller in absolute size than the wage elasticity of labor demand, conforming to the stylized fact of a large pool of surplus labor in China. The wealth variable is negative, suggesting that an increase in households’ net worth induces an increase in their consumption of leisure and a decline in hours supplied; conversely, the unemployment rate has the expected positive association with the supply of labor, supporting the presence of the “added-worker” effects. Finally, the elasticity of labor supply with respect to the scale variable, H, is positive, consistent with the idea that an increase in the size of the potential labor force is associated with an increase in aggregate labor supply.

The estimated coefficient on the excess supply indicator variable, δ, is positive and statistically significant at the 10 percent error level, suggesting that labor market tightness, measured as the deviation of the unemployment rate from its nonaccelerating inflation level, is statistically informative about the presence of excess labor. Furthermore, significance of δ is rejection of the null hypothesis in an equilibrium model (see Annex 9A).

The excess labor supply implied by these results is presented in Figure 9.4. There are several features to note: (1) China has had an excess supply of labor continuously since at least 1991; (2) the estimated surge in excess supply around 2000–04 coincides with the end of the reforms of state-owned enterprises, which resulted in job-shedding, underemployment, and unemployment; (3) after falling continuously between 2004 and 2008, surplus labor rises abruptly in 2009–10, reflecting the impact the global financial crisis has had on labor demand; and (4) the model generates an excess supply estimate of about 160 million in 2007, which encompasses the frequently cited 150–200 million estimate of the National Population Development Strategy Research Report (2007).

Figure 9.4Estimated Excess Supply Levels

Source: IMF staff estimates.

Scenario Analysis

The empirical estimates presented above point to a long period of excess labor supply, notwithstanding significant job creation since about 1980 (which amounted to about 350 million jobs). Because the key question in this chapter is about the onset of the LTP, this section considers several scenarios to forecast the evolution of excess labor supply.

Baseline Scenario

The central forecast of the path of excess labor is the baseline scenario. Forecasts under the baseline are derived by making the following assumptions about the paths of variables that have been identified as key determinants of the notional supply of and demand for labor.

Real wage adjustment. Real wages at time t depend on real wages in the past two periods, on contemporaneous inflation, inflation in the past period, and the nonaccelerating inflation rate of unemployment (NAIRU) in the current period:

The presence of lagged real wages captures potential sluggishness in real wage adjustments. Inflation, p, is included to reflect that workers’ (nominal) wage demands rise with inflation. Finally, U *, the NAIRU, is expected to affect wages because of the assumption that tighter labor market conditions (i.e., lower NAIRU) are likely to increase workers’ bargaining power and result in higher wages. Inflation and NAIRU forecasts through 2017 are drawn from the IMF World Economic Outlook (WEO), and both variables are assumed to grow at the 2017 rate thereafter.

Household net wealth. The evolution of net financial wealth (NFW) is derived from a standard wealth accumulation equation in which net wealth increases each period as the result of interest payments on the stock, and the new flow from household saving:

in which i is the nominal deposit rate and α is a constant fraction of household saving that is assumed to flow into household wealth every period. The parameter α is estimated from a time series regression of NFW on household saving. The forecast of household saving is, in turn, derived by assuming that the household-saving-to-GDP ratio follows WEO projections of China’s private-saving-to-GDP ratio.

Demographics. Population and working-age population (15–64 years) are obtained from the “constant fertility” variant of the UN population database. The constant fertility projections assume that the fertility rate through 2050 remains at the average rate of 2005–10. Forecasts of the labor force are derived from a nonlinear regression of the time series of labor force on a constant, the stock of working-age population, and its square (Figure 9.5).5

Figure 9.5Labor Force, Actual and Regression Estimates

TFP. The TFP level is assumed to increase annually at the average of its 2005–10 growth rate (3.9 percent) until 2017, and remain at its 2017 level thereafter.

Unemployment rate. Forecasts are from the WEO through 2017. From 2017 onward, the unemployment rate is assumed to stay fixed at the 2017 rate (4 percent).

Partner GDP growth. Real GDP projections of China’s eight largest trading partners, weighted by export shares, are obtained from the WEO and Direction of Trade Statistics (DOTS).6 Real growth rates after 2017 are assumed to stay at the 2017 level; export shares are fixed at their 2011 level.

Under these assumptions, the projected path of excess supply in the baseline scenario indicates that China’s excess supply of labor peaked in 2010 and is on the verge of a sharp decline: from 151 million in 2010, to 57 million in 2015, and 33 million in 2020 (Figure 9.6). The LTP is projected to emerge between 2020 and 2025, when excess supply turns negative (i.e., the labor market moves into excess demand). The rapid rate of decline in excess supply closely follows the projected path of the dependency ratio, which bottomed out in 2010 and is projected to rise rapidly.7 The projected path of excess labor supply also reflects the expected evolution of wealth, which reduces labor supply, and TFP, which raises labor demand, in the baseline.

Figure 9.6Baseline Scenario: Surplus Labor

Sources: IMF staff estimates.

Baseline results are derived under the assumption that market conditions and economic policies remain unchanged. However, the significant demographic transition, the changing external environment, and rising social needs may well spur an endogenous policy response that could alter the structure of the economy, and an endogenous market response (e.g., higher wages, greater bargaining power) is also likely to emerge. In addition, the government’s plans to change the technological mix of industry, urbanization, and income distribution outlined in the 12th Five-Year Plan will likely impact the LTP by shifting the supply of or the demand for labor (or both). Therefore, the following section considers the consequences of assuming a higher fertility rate, a higher labor force participation rate, financial sector reform, and product market reform. The results of these scenarios are given in Table 9.3 and Figure 9.7.

Table 9.3Excess Labor Supply Scenarios(millions)
Higher fertility51.936.0−16.8−126.3
Higher labor force participation92.368.15.31−114.1
Financial sector reform57.010.5−70.4−220.5
Product market reform54.911.6−44.0−153.7

Figure 9.7Alternative Scenarios: Surplus Labor

Source: IMF staff estimates.

Increase in the Fertility Rate

The first scenario considers the effects of a one-time permanent increase in the fertility rate, for example, by selective relaxation of the One Child Policy. This scenario is simulated using the UN’s “high-fertility” variant to forecast working-age population,8 and using these new variables to forecast the labor force. All other variables are left as in the baseline scenario.9

Results are given in Figure 9.8. The estimates indicate that in a higher-fertility scenario, the LTP is delayed relative to the baseline. This result is consistent with the priors because higher fertility will result in a larger working-age population and larger potential labor force. However, the increase in excess supply relative to the baseline is small, rising from 33 million to 36 million in 2020 and from −27.8 million to −16.8 million in 2025. There are two possible explanations. First, higher fertility increases the potential labor force with a delay because it takes time for new larger cohorts to join the workforce. Second, the UN high-fertility variant is a modest increase in fertility that lifts China’s fertility rate of about 1.6 to just around the replacement-level fertility rate of 2.1, with a correspondingly small induced increase in the working-age population relative to the baseline.10

Figure 9.8Working-Age Population: Constant Versus High Fertility

Sources: UN Population Database; and IMF staff calculations.

Higher Labor Force Participation Rates

This scenario analyzes the impact of greater labor force participation on the LTP. Participation rates in China are high relative to comparators, but have fallen since the mid-1990s, from 0.87 in 1995 to 0.82 in 2010 (Figure 9.9). Given the disposition toward hiring younger workers and a relatively low retirement age, this decline reflects the growing share of older workers in the labor force. The stability of the pension system has been suggested as another reason for declining participation (Yang and Wang, 2010). Although this chapter does not identify a specific mechanism for raising participation, one path is through greater interprovince labor mobility, for example, accelerated progress in hukou reform. The specific scenario is a one-time increase in the participation rate from 0.82 to 0.85, which amounts to the average rate of the past two decades.

Figure 9.9Labor Force Participation Rate

Source: IMF staff estimates.

The impact of higher participation rates on excess labor supply is significant. With higher participation, the analysis projects that an excess supply of labor persists beyond 2025 and the LTP emerges between 2025 and 2030. Unlike fertility, higher participation has an immediate impact on the size of the labor force and thus on the supply of labor. Therefore, higher participation causes a longer delay in the LTP relative to higher fertility.

Financial Sector Reform

The financial reform scenario considered here is a very specific one: interest rate deregulation that lifts deposit rates. The channel by which interest rates is assumed to affect excess supply is the wealth effect: Higher deposit rates raise the return on the stock of wealth, but decrease the flow into wealth as households meet saving targets more easily, which, in turn, reduces the supply of labor. The scenario is simulated for a 5 percentage point increase in nominal deposit rates, using estimates of the saving response to interest rates in Nabar, 2011 (also see Chapter 5 in this volume).

The results indicate that financial reform accelerates the crossing of the LTP. Whereas in the baseline the excess supply of labor in 2020 was in the range of 30 million, interest rate liberalization would reduce this excess to about 10 million, and the LTP would occur shortly after 2020. Because the depletion of excess labor is likely to be associated with higher wages, potentially raising labor’s share of income, financial reform along the lines of this scenario is broadly consistent with Chinese authorities’ objectives of raising household income in the medium term.

Product Market Reform

The final scenario considers product market reform that lifts TFP. Although TFP contribution to output growth in China has been positive since 1990, its growth has slowed in recent years. Raising TFP is consistent with a wide variety of policies announced in the 12th Five-Year Plan, such as greater competition in the services sector and investment in higher value added activities. Unlike in the other scenarios, higher TFP works through the labor demand side in this framework, raising firm profitability and thus the demand for labor (Freeman, 1980). This scenario is simulated through a one-time permanent increase in the growth rate of TFP to 4.5 percent, the average TFP growth since 1990.

The impact on the LTP from higher TFP is qualitatively similar to financial reform: a faster decline of excess labor supply, and a faster emergence of the LTP, relative to the baseline. However, this result is, in part, a consequence of the model specification in which TFP does not directly affect the supply of labor. In an alternative setup (e.g., Pissarides and Valenti, 2007) in which gains in productivity translate into lower unemployment—and thus a lower notional supply of labor—a smaller decrease in excess labor supply could result.


China is on the cusp of a demographic shift that will have profound consequences on its economic and social landscape. Within a few years the working-age population will reach a historical peak, and will then begin a precipitous decline. The core of the working-age population, those ages 20–39 years, has already begun to shrink. With this, the vast supply of low-cost workers—a core engine of China’s growth model—will dissipate, with potentially far-reaching domestic and external implications.

This chapter empirically assesses when labor shortages might emerge. The central result is that, barring an endogenous market or policy response, the excess supply of labor—the reserve of unemployed and underemployed workers (currently in the range of 150 million)—will fall to about 30 million by 2020 and the LTP will be crossed between 2020 and 2025.

An endogenous policy response to potential labor shortages is, however, likely as the government tries to slow the transition to the LTP. Market mechanisms that result in higher wages as labor markets tighten, or induce a transition to more capital-intensive production, may also offset the shrinking labor pool. Scenario analysis reveals that higher fertility through relaxation of the one-child policy and greater labor force participation through hukou reform will delay depletion of excess labor. Financial reform and higher TFP, conversely, accelerate the transition to a labor shortage economy, through wealth effects and greater profitability of firms, respectively.

Quantitative estimates of the excess supply of labor and the timing of the LTP presented in this chapter are inherently uncertain. In addition, alternative scenario exercises are analyzed for specific reforms of particular magnitudes and do not take into account potential inter-scenario effects (e.g., the offsetting impacts of higher fertility and financial sector reform). That said, the main point of the analysis is that market and policy responses to the declining labor surplus will be largely peripheral; demographic forces play a dominant role in the imminent transition to a labor shortage economy.

Annex 9A. The Disequilibrium Model of the Labor Market

Labor Demand. Aggregate demand for labor is specified as a function of the endogenous real wage, TFP—reflecting the standard Cobb-Douglas production function in which profit-maximizing firms demand more labor with technological progress—and partner GDP growth, a proxy for demand conditions given China’s high dependence on exports. The model is log-linear with an additive stochastic error:

in which LD denotes the natural logarithm of the notional aggregate demand for labor; W is the natural log of gross real wages (gross nominal wages deflated by the consumer price index); GDPp is the natural log of the real GDP-weighted growth rate of trading partners; and TFP is total factor productivity, calculated as the residual of a growth accounting framework with capital and labor shares assumed in the literature.11

Labor Supply. Aggregate labor supply depends on real wages, net household wealth, the scale of the potential labor force (approximated by the participation rate interacted with population), and the unemployment rate. The unemployment rate is included to capture “added-worker effects,” that is, the notion that under weak labor demand conditions, households may send additional individuals to look for work, resulting in a positive observed association between the supply of labor and the unemployment rate (Basu, Genicot, and Stiglitz, 2000):

in which LS is notional aggregate supply of labor, H denotes the natural log of the scale variable, and U is the unemployment rate.

In the equilibrium model, the observed quantity of labor equates the notional supply of labor, LS, with the notional demand for labor, LD. In the disequilibrium model, however, it is assumed that the observed quantity, L, is the minimum of the notional labor supplied and demanded:

This assumption implies that if LS > LD, then L = LD and the observed quantity lies only on the demand curve, whereas LS < LD indicates L = LS and the observed quantity lies only on the supply curve. This is the key contrast with the equilibrium model (L = LS = LD), because in the disequilibrium case, the demand for and supply of labor are unobservable unless they are the minimum in equation (9A.3). The deterministic function is defined as

in which I denotes an indicator of excess supply, I * denotes the equilibrium value of I, and δ is an unobserved parameter (δ > 0 under the null hypothesis). In the empirical analysis, I is the unemployment rate and I* is the NAIRU. Alternatives for I are wage inflation, the layoff rate, and the quit rate (Baily, 1982). Two additional variables are defined as

Substituting equations (9A.4) and (9A.5) into equations (9A.1) and (9A.2) and rearranging yields the model to be estimated:

Identification follows immediately because both system equations are overidentified. Because δ enters both equations, the model is estimated by three-stage least squares.

Note, when δ = 0, equation (9A.6) reduces to

which is identical to the standard equilibrium model with L= LS = LD. Thus, statistical rejection of δ = 0 is evidence in support of disequilibrium.

Annex 9B. Derivation of the Empirical Framework

Derivation of equation (9A.6) is as follows:

Switching model. The presence of the min condition in equation (9A.3) introduces computational difficulties caused by the nonlinearity of the resulting reduced-form equations for labor supply and demand. This issue is addressed by invoking a switching model in the spirit of Fair and Jaffe (1972). Specifically, the derivation assumes δ > 0. Then, when I > I*, it is inferred that the market is in excess supply, LS > LD and L = LD. The switching model overcomes the computational difficulties of the min condition because it exactly partitions the data into periods of excess supply and excess demand, effectively making the min condition redundant in the estimation (see below). The switching model is deterministic, but does not necessarily describe a causal relationship.

Disequilibrium simultaneous equations model. The key step in deriving equation (9A.6) is replacing the difference of the unobservable variable (LSLD) with the deterministic indicator and partitioning the data into regimes of excess supply or excess demand, which together yield a simultaneous equation model. Specifically, note the following:


    Aziz, J., and L.Cui, 2007, “Explaining China’s Low Consumption: The Neglected Role of Household Income,” IMF Working Paper 7/181 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    Baily, M., 1982, Workers, Jobs and Inflation (Washington: The Brookings Institution).

    Barro, R.J., and H.I.Grossman, 1971, “A General Disequilibrium Model of Income and Employment,” American Economic Review, Vol. 61, No. 1, pp. 82–93.

    • Search Google Scholar
    • Export Citation

    Basu, K., G.Genicot, and J.E. Stiglitz, 2000, “Unemployment and Wage Rigidity When Labor Supply Is a Household Decision,” Working Paper 00-10 (New York: Cornell University, Center for Analytic Economics).

    • Search Google Scholar
    • Export Citation

    Cai, F., and M.Wang, 2008, “A Counterfactual Analysis on Unlimited Surplus Labor in Rural China,” China and the World Economy, Vol. 16, No. 1, pp. 51–65.

    • Search Google Scholar
    • Export Citation

    Chan, K.W., 2010, “A China Paradox: Migrant Labor Shortage amidst Rural Labor Supply Abundance,” Eurasian Geography and Economics, Vol. 51, No. 4, pp. 513–30.

    • Search Google Scholar
    • Export Citation

    Fair, R., and A.Jaffe, 1972, “Methods of Estimation for Markets in Disequilibrium,” Econometrica, Vol. 40, pp. 497–514.

    Freeman, R., 1980, “The Evolution of the American Labor Market, 1948–80,” in The American Economy in Transition, ed. byM.Feldstein (Chicago: University of Chicago Press).

    • Search Google Scholar
    • Export Citation

    Garnaut, R., 2006, “The Turning Point in China’s Economic Development,” in The Turning Point in China’s Economic Development, ed. byRossGarnaut and LigangSong (Canberra: Asia Pacific Press).

    • Search Google Scholar
    • Export Citation

    Golley, J., and R.Tyers, 2006, “China’s Growth to 2030: Demographic Change and the Labour Supply Constraint,” Australian National University Working Paper (Canberra).

    • Search Google Scholar
    • Export Citation

    Hajivassiliou, V., 1993, “Macroeconomic Shocks in an Aggregative Disequilibrium Model,” Cowles Foundation Discussion Paper No. 2063 (New Haven, Connecticut: Cowles Foundation, Yale University).

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF), 2011, People’s Republic of China: Spillover Report for the 2011 Article IV Consultation and Selected Issues (Washington).

    • Search Google Scholar
    • Export Citation

    Knight, J., Q.Deng, and S. Li, 2011, “The Puzzle of Migrant Labour Shortage and Rural Labour Surplus in China,” China Economic Review, Vol. 22, No. 4, pp. 585–600.

    • Search Google Scholar
    • Export Citation

    Kwan, C.H., 2007, “China Shifts from Labor Surplus to Labor Shortage: Challenges and Opportunities in a New Stage of Development” (Tokyo: Research Institute of Economy Trade and Industry).

    • Search Google Scholar
    • Export Citation

    Lewis, W.A., 1954, “Economic Development with Unlimited Supplies of Labour,” Manchester School, Vol. 22, No. 2, pp. 139–92.

    Lucas, R.E., and L.Rapping, 1969, “Real Wages, Employment, and Inflation,” Journal of Political Economy, Vol. 77, pp. 721–54.

    Nabar, M., 2011, “Targets, Interest Rates and Household Saving in Urban China,” IMF Working Paper 11/223 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

    National Population Development Strategy Research Report, 2007.

    Pissarides, M., and G.Valenti, 2007, “The Impact of TFP Growth on Steady-State Unemployment,” International Economic Review, Vol. 48, pp. 607–40.

    • Search Google Scholar
    • Export Citation

    Quandt, R.E., and H.Rosen, 1986, “Unemployment, Disequilibrium and the Short Run Phillips Curve: An Econometric Approach,” Journal of Applied Econometrics, Vol. 1, No. 3, pp. 235–53.

    • Search Google Scholar
    • Export Citation

    Rosen, H., and R.E.Quandt, 1978, “Estimation of a Disequilibrium Aggregate Labor Market,” Review of Economics and Statistics, Vol. 60, No. 3, pp. 371–79.

    • Search Google Scholar
    • Export Citation

    Rudebusch, G.D., 1986, “Testing for Labor Market Equilibrium with an Exact Excess Demand Disequilibrium Model,” Review of Economics and Statistics, Vol. 68, No. 3, pp.468–76.

    • Search Google Scholar
    • Export Citation

    Yang, D.U., and M.Wang, 2010, “Demographic Aging and Employment in China,” Institute of Population and Labour Economics, Chinese Academy of Social Sciences.

    • Search Google Scholar
    • Export Citation

    Zhang, G., and Z.Zhao, 2006, “Reexamining China’s Fertility Puzzle: Data Collection and Quality over the Last Two Decades,” Population and Development Review, Vol. 32, pp. 293–321.

    • Search Google Scholar
    • Export Citation

    Zhang, X, J.Yang, and S.Wang, 2010, “China Has Reached the Lewis Turning Point,” IFPRI Discussion Paper No. 977 (Washington: International Food Policy Research Network).

    • Search Google Scholar
    • Export Citation

This measure comes from the National Bureau of Statistics of China. However, it is unclear whether all firms are required to record demand and whether both informal and formal workers are recorded in the measure of labor supply.


Industrial relocation to the inland may not necessarily be a reflection of coastal cost pressures and instead reflect, for instance, expansion into the vast consumer base in inland provinces. In this scenario, labor bottlenecks in coastal areas may simply be a result of migrants’ rising reservation wages for coastal employment, given greater opportunities inland.


Note, because the model is derived by assuming that δ ≥ 0, critical test values for the estimated δ are obtained from one-sided t-tables. This restriction is not imposed in estimation. Nonnegativity of δ is a testable assumption, and statistically significant nonnegativity of δ is a rejection of equilibrium in the labor market.


Because the parameter δ appears in both the supply and demand equations, estimation is done by three-stage least squares of (9A.6). This approach yields the added benefit of doubling the observations available for estimation. In estimating the model, it is assumed that the errors ε1 and ε2 are each serially uncorrelated and (ei1, ei2) ~ N(0, Σ) where Σ is possibly nondiagonal, that is, the errors may be contemporaneously correlated. Furthermore, it is assumed that only the wage variable is endogenous.


Although the labor force could potentially include those outside the working-age population, this regression has high predictive power (adjusted R2 = .998).


These are the euro area, Hong Kong SAR, Japan, Korea, Singapore, the United Kingdom, the United States, and emerging and developing economies. This group received more than 92 percent of all Chinese exports in each year of the sample.


The dependency ratio (less than 15 plus greater than 64 years of age, in fraction of the population) is not in the model; however, the working-age ratio is captured in the scale variable given that the participation rate is measured as working-age population normalized by the size of the labor force.


The UN’s high-fertility variant assumes 0.5 children more than the constant-fertility variant.


In each subsequent scenario, with the exception of the variable whose impact on the LTP is examined, all other forecasts are left as in the baseline.


Fertility rates are from Golley and Tyers (2006), replacement rate estimates from Zhang and Zhao (2006). The working-age population in the high-fertility variant is larger than the baseline working-age population by 0 percent, 1.4 percent, and 4 percent in 2025, 2030, and 2035, respectively.


TFP is in level terms.

    Other Resources Citing This Publication