Efficiency Wages and Work Incentives in Urban and Rural China 1

Eﬃciency Wages and Work Incentives

in Urban and Rural China 1

Belton M. Fleisher corresponding author Department of Economics, The Ohio State University 1945 North High Street Columbus, OH 43210 E-mail: ﬂ[email protected] and Xiaojun Wang Department of Economics, University of Hawaii at Manoa Honolulu, HI 96822 E-mail: [email protected]

Running Head: Wages and Work Incentives in China

1 This paper is forthcoming in Journal of Comparative Economics December 2001. Eﬃciency Wages and Work Incentives

in Urban and Rural China

Abstract

This paper examines incentive-wage effects for production and for managerial/technical workers in both urban and rural Chinese non-agricultural enterprises. We report strong evidence ofproductivity-enhancing wage behavior among enterprises in all ownership categories. There is also evidence that firms paying higher efficiency wages experience less shirking among their employees. We find that the profit-maximizing potential ofincentive-wage setting is not fullyexploited, although there is weak evidence that joint ventures come closer to profit-maximizing behavior at this intensive margin of wage/employment behavior than do collectives or state-owned enterprises.

Keywords: Eﬃciency Wages, Incentive Eﬀects, Productivity, Transition Economies, Chi- nese Economy

JEL-Codes: P23, J31, O15

2 1 Introduction and Literature Review

This paper is an empirical investigation ofthe efficiency-wage effect in Chinese labor markets. The market reforms that began in agriculture in the late 1970s spread unevenly and gradually to both rural and urban industrial enterprises and opened the door to profit-seeking behavior in product and input markets. State-owned, collective, and privately owned firms have come under increasing competitive pressure to reduce costs, ofwhich labor costs are among the most important. In the absence ofcostless information and in the presence ofcostly monitoring, indirect means ofcontrolling labor costs and reducing shirking are required. Efficiency-wage payments are recognized as an important means ofachieving these goals.

The idea that higher real wages can lead to greater productivity and even to higher profits is found in the writings of economists as far back as Adam Smith. The emer- gence ofefficiency-wage theory in the 1970s was a response to the persistence ofseveral phenomena that are inconsistent with neoclassical economic theory, including involun- tary unemployment, wage differences not explained by the theory ofcompetitive labor markets or models ofhuman capital, and wage rigidity in the presence ofcyclical unemployment. Presentations ofthe relevant models and citations to the literature are found

3 in Stiglitz (1986) and Katz (1986). Akerlofand Yellen (1986) contains reprints ofa number ofclassic articles on eﬃciency-wage models and presents a succinct introduction summarizing alternative underlying rationales for eﬃciency-wage schemes.

We study the incentive effects ofwages in Chinese enterprises in both rural and urban areas under various ownership arrangements. The uneven transition toward free markets in China involved different policies and enforcement in urban versus rural areas, state- owned versus collective- and privately-owned enterprises, and to special economic zones versus non-favored areas. Hence, we expect to observe different manifestations of market forces, including the application and effects of efficiency-wage policies, across firms. We assume that workers are living above subsistence levels, so that productivity enhance- ment does not operate primarily by making workers stronger and healthier (Leibenstein,

1986), but rather through enhanced incentives. We do not distinguish between wage policies in which bonuses or other forms of sharing are implicitly or explicitly linked to the firms’ profits (e.g., Weitzman, 1984), and those in which firms act up-front to pay workers higher than competitive-market wages in order to increase incentives to reduce quits, absenteeism, or shirking. Our major question is whether any policies that raise pay above a worker’s alternative wage increase incentives and thus contribute to higher productivity and/or profit.2

4 Empirical studies ofefficiency wages in non-agricultural firms have generally used data from major industrial countries. Frequently these studies have focused on intermediate targets ofwage policy, e.g. turnover or worker morale, rather than on the bottom line ofprofits or productivity. In a widely cited study, Krueger and Summers

(1988) report evidence that industry wage diﬀerentials that are uncorrelated with observable human capital characteristics in the United States are negatively related to worker turnover. Cahuc and Dormont (1997) use two panels ofFrench manufacturing

ﬁrms with observations from 1986 to 1989 to show a positive relationship between various proﬁt-sharing arrangements and productivity at the micro level. Huang, Hallam,

Orazem, and Paterno (1998) summarize research published in the 1990s that shows a negative relationship between relative wages and disciplinary layoffs; a positive effect on firm sales; and a positive effect on firm output. In their own research using United

States data at the two-digit industry level, these authors ﬁnd that wage premia that are unexplained by observable human capital characteristics and unemployment rates are positively related to productivity.

We are not aware ofresearch on China that looks explicitly forefficiency-wage effects on worker behavior, productivity, or profits. However, there is a considerable body ofresearch that explores the relationship between wage-payment schemes and productivity.3

5 In a recent paper, Zhuang and Xu (1996) show that the positive relationship between bonus payments and total-factor-productivity (TFP) growth in state-owned enterprises

(SOEs) reported by Groves, Hong, McMillan, and Naughton (1994) can be replicated in data that extends into the early 1990s. Moreover, they show that bonus incentives increase profits as well as productivity. Coady and Wang (2000a) establish evidence of a positive relationship between wage bonuses and profits in urban Liaoning province; however, they do not find direct evidence that bonus-sharing creates stronger work incentives. Dong and Putterman (2000) report a similar result for a sample of 1000 SOEs over the decade 1980 to 1990. Both papers conclude that bonus payments are underused because, iftheir results were produced by processes in which increased bonuses cause higher productivity, increases in bonus payments would increase profits in their samples ofenterprises. Yuen (2000) finds that enterprises performbetter financially ifemployees’ wage payments are closely related to their output.

In this paper, we depart from previous studies of the Chinese economy in three ways.

First, rather than focusing on the behavior of bonus payments, we estimate an eﬃciency- wage component ofthe reported wage payments indirectly and then estimate the impact ofthese excess wage payments on productivity. We also use data on reported shirking to show a negative relationship between our measure ofeﬃciency wages and shirking in

6 one sample ofdata. Second, we divide enterprises’ employees into skilled and unskilled workers and examine the incentive eﬀects ofwage payments separately foreach group.

Third, we exploit the panel featureofour sample ofrural enterprises to address the issue ofGranger causality between productivity and eﬃciency wage payments.

We find considerable support for the hypothesis that Chinese firms pay wages that have positive incentive effects. However, the degree to which this tool ofcost-minimization is exploited varies widely across ownership types and between rural and urban areas. We confirm Zhuang and Xu’s (1996) and Dong and Putterman’s (2000) finding that urban enterprises underuse efficiency-wage payments from a profit-maximization perspective.

However, we find that, among rural enterprises, efficiency-wage payments are higher than profit maximization would require.

The rest ofthe paper is organized as follows. In the next section, we describe briefly wage policies in China during the period studied; we then specify a simple efficiency- wage model ofproduction; finally, we discuss the two data sets (urban and rural) used in our empirical work. In section 3, we present estimation results for the urban sample; section 4 contains the estimation results for the rural sample. Section 5 concludes.

7 2 Enterprise Wage Policies, Eﬃciency Wages and Productivity, and the Data

Under central planning, particularly in SOEs, wages were set according to a wage grid that categorized workers into eight skill classes. Wage diﬀerentials were highly compressed; see, for example, Gordon and Li (1999). Bonus payments, which were deemed unacceptable during the Cultural Revolution, reappeared in the late 1970s as economic reform began. According to Naughton (1995), the share of bonus payments in total wage packages increased steadily from virtually nothing in 1978 to approximately

22% in 1992. He reports that enterprise managers were given more autonomy in hiring and wage decisions and encouraged to use bonus payment to improve workers’ incentive and thus productivity. Meng and Perkins (1998) conﬁrm that both state and non- state enterprises acquired more wage-setting autonomy with economic reform. Coady and Wang (2000b) note that, by the late 1980s, bonus payments were contributing to increased earnings inequality in both the state and collective sectors.

Gordon and Li (1999) provide evidence that wage compression among SOEs sub- jected them to increasing labor-market competition from the non-state sector as economic reform progressed in the 1980s. New ownership forms were not required to follow

8 the traditional wage grid, although they were not totally free of political pressure in setting rates ofpay. Thus, fast-growing collective and private enterprises became relatively attractive employers, particularly to underpaid managerial and technical workers. While there is evidence that wage diﬀerentials by level ofschooling have increased during the late 1980s and early 1990s (Li, 2001), substantial wage compression remained among all ownership forms.

2.1 Eﬃciency Wages in the Production Function

We postulate a simple eﬃciency-wage model following Stiglitz (1986), Shapiro and

Stiglitz (1984) and Krueger and Summers (1988) in which the excess ofwages paid over an estimate ofa spot-market competitive wage is included as a regressor in the following augmented Cobb-Douglas production function.

α βj Y = K ( (ej Lj ) )exp(φZ + ), (1) j where:

Y is gross output or value added,

K is net capital stock,

Lj is the labor ofthe jth group, e.g. skilled and unskilled,

ej is the eﬀort function for the jth group ofemployees,

9 Z is a vector ofenterprise characteristics, e.g. ownership type, region, and year of observation, and is an identically and independently distributed disturbance term.

The eﬀort function is deﬁned as:

W j ηj Z ej = Bj ( ) , (2) Waj

where B and η are parameters, Wj is the observed wage ofthe jth group ofemployees,

and Waj is the estimated spot-market competitive wage. This measure ofthe eﬃciency wage assumes that, ifa worker loses his current job, the alternative wage would be the average predicted wage for his group. Wherever possible, we include in-kind beneﬁts, particularly housing, in the wage measures we use.

2.2 Data for the the Urban Sample

The urban survey4 is a stratiﬁed random sample, collected in the second halfof1992 forthe year 1991, ofurban enterprises within a randomly selected sample oflocales.

This urban sample includes 442 enterprises in 24 cities of12 provinces. Two survey instruments were administered, one to an oﬃcial ofeach enterprise and the other to a random sample ofemployees ofthe enterprise. The sample size ofthe employee survey is

9397, which is about 1.5% oftotal employment. Enterprises come fromall major types ofownership, state owned, collective, joint venture, and private. The sample contains

10 almost the same number of enterprises from coastal areas as from non-coastal areas. See

Tables 1 and 2 for more detailed information on this sample. The names of the provinces are listed at the bottom ofTable 2.

TABLES 1 and 2 ABOUT HERE

2.3 The Rural Sample

The rural sample is a panel survey of200 large rural, mostly township/village enterprises (TVEs) for the years 1984 to 1990. To the extent that economic reform originated in agriculture and progressed gradually over time, spreading geographically from rural to urban areas, the time gap between the urban and rural samples may not be too serious.

Nevertheless, differences in the empirical results should be interpreted in light ofthe temporal mismatch ofthe two samples. The survey covers 20 enterprises in each of10 provinces. The survey includes not only data for the individual firms, but also important data describing the markets in which the firms operates, e.g., the total employment in the village where the firm is located. This allows us to test hypotheses on the impact of market structure on the behavior ofthe firm. 5 See Table 3 for sample statistics for this data set.

TABLE 3 ABOUT HERE

11 3 Estimation Results for the Urban Sample

When data on schooling are available, we use a Mincerian human-capital wage

equation to estimate Waj as the average wage for workers who have achieved a given level ofschooling and labor-market experience. Table 4 reports this regression, in which all

TABLE 4 ABOUT HERE right-hand variables are highly significant. Adding tenure to the right hand side has little effect on the estimates. The wage variable, W , is adjusted for housing benefits.

Ifthe employee lives in an employer-provided house, an estimated amount ofannual market rent for the same size house in the same province is added to the wage income.

This estimated annual rent is based on the average annual rent per square meter paid by those who rent an apartment in the same province multiplied by the size ofthe house.

One ofthe anonymous refereessuggests that, because oflabor-market segmentation, the wage equation would be better speciﬁed ifit included locational, e.g., provincial, dummy variables as in Gustafsson and Li (2000). This is a debatable point. Huang, Hallam,

Orazem, and Paterno (1998, p. 131) state that in their eﬃciency-wage paper, “The speciﬁcation ofthe earnings functionconcentrates on human capital variables only. The

12 intent is to estimate the wage premium as the portion ofthe wage uncorrelated with observable human capital.” These authors do not include variables for union status, industry or geographical region in the wage equation from which their efficiency-wage estimates are generated. On the other hand, iflabor markets in China are strictly segmented geographically, the relevant alternative wage in the efficiency-wage framework would be specific to the local or provincial labor-market. Indeed this is an implicit assumption in calculating the efficiency wage for the rural sample.

One could argue that urban workers are free to migrate to rural areas, e.g. back to their communities ofbirth, should they lose their urban jobs, but rural workers are not so mobile. Thus, assuming an economy-wide alternative wage for workers in the urban sample is more defensible than it would be for workers in the rural sample. However, we have re-estimated the wage equation reported in Table 4 with locational dummy variables included as regressors. The estimated coeﬃcients for high-school and college graduates fall from 0.084 to 0.049 and from 0.23 to 0.18, respectively, but both remain highly signiﬁcant. Given the a priori considerations discussed above, we choose the formulation of the wage equation without locational dummies.6

Workers’ rates ofreturn to schooling are relatively low, as is uniformly foundin studies ofthe Chinese economy, compared to both industrialized and other transition

13 economies.7 Workers who have completed secondary school (that is, upper middle school) earn about 8% more than workers without a high-school diploma. Assuming that the average non-high school graduate has six years ofschooling, this would amount to a private rate ofreturn ofapproximately 1.2% per year ofadditional schooling for high-school graduates. The private rate ofreturn forcollege graduates, similarly calculated, amounts to about 5% per year. Although higher than that for high-school graduates, this rate is still very low by international comparisons, especially compared to other nations at China’s level ofeconomic development. The coeﬃcients ofage and experience have signs and magnitudes found typically in other studies. Male workers receive on average about 14% higher earnings than do female workers; see Maurer-Fazio,

Rawski, and Zhang (1999).

In order to test the hypothesis that efficiency-wage considerations enter into firms’ wage policies, we estimate the production function defined in equation (1). The dependent variable is gross output because data on intermediate inputs are not reported in the urban sample.8 Econometric results for equation (1) are presented in Table 5, in which

Waj is taken to be the wage predicted for each worker from the equation estimated

TABLE 5 ABOUT HERE in Table 4.9 Column (1) shows the estimated coeﬃcients for the simplest speciﬁcation

14 ofthe production function, with both labor and the eﬃciency-wage variable pooled over both production and technical and administrative staﬀ (TAS) workers. The output elasticity oflabor, at 0.33, is estimated with considerable precision and, although low by comparison with estimates for most industrial economies, it is within the bounds of several studies ofthe Chinese economy, e.g., Chow (1994), Dollar (1990), and Fleisher,

Dong, and Liu (1996). However, the null ofconstant returns to scale is rejected, with a p-value ofno greater than 1% (not reported).

A critical test ofthe joint hypothesis that enterprises are both profit-maximizing and pay efficiency wages is that the elasticity ofwork effort with respect to the efficiency wage is unity; see Solow (1979) and Stiglitz (1986). By simple substitution ofequation (2) into equation (1), we see that the elasticity ofeffort with respect to the efficiency wage can be identified easily by dividing the coefficient ofthe variable MT by the coefficient of ln L. Ifthe effort-efficiency wage elasticity equals unity, the coefficient ofthe efficiency- wage variable will equal that ofthe labor variable. This condition is clearly violated by the estimates reported in column (1). The calculated effort-efficiency wage elasticity in column (1) is more than 3, implying that the impact ofpaying an efficiency wage on profit is not being fully exploited. A similar result for wage bonus payments is reported by Zhuang and Xu (1996) and Dong and Putterman (2000).

15 The same calculations for the estimates reported in column (2) imply that there is grossly insufficient exploitation ofthe profit-maximizing potential ofefficiency wages, more so by SOEs than by other ownership forms.10 While SOEs are expected to be less profit-motivated than firms ofother ownership forms,the other ownership classes appar- ently also fall far short of profit maximization with respect to the efficiency wage rate.

In column (3), we find that the implied effort-efficiency wage elasticity for production workers (the coefficient of MPWdivided by the coefficient of PW) is over 20. How- ever, the output elasticity for production workers is insignificant and estimated quite imprecisely. The effort-efficiency wage elasticity for technical/administrative workers

(the coefficent of MTAS divided by the coefficient of TAS) is about 1.4, which is rather close to the profit-maximizing value. Further calculations, using column (4), indicate that the effort-efficiency wage elasticity for technical/administrative workers is closest to profit-maximizing in joint-venture enterprises, which is reassuring, because ofthe likely degree ofprofit motivation in this ownership class. The calculation forTAS workers in joint-ventures is carried out using the estimated coefficient ofthe variables TAS

(0.47), MTAS (1.21), and MTAS 3 (-.50) to calculate the ratio (1.21 − 0.50)/0.47, which is equal to 1.51. The calculated ratio for collectives and SOEs is 1.65 and 2.58, respectively.11 Similar calculations for production workers yield eﬀort-eﬃciency wage

16 elasticities of26.2, 19.1, and 11.08, respectively, forjoint ventures, collectives, and SOEs, all based on imprecisely estimated coeﬃcients.

Another perspective on efficiency wages enables us to investigate more carefully whether the positive estimated relationship between the efficiency-wage variable and output reflects a true incentive effect. The urban sample includes answers to several questions about worker behavior. One ofthe questions asks workers how frequently they observe shirking among their co-workers.12 Table 6 relates reported observations ofshirking to ownership type, location (coastal or

TABLE 6 ABOUT HERE noncoastal province) and the eﬃciency-wage variable used to estimate equation (1). The dependent variable takes a value of1 forenterprises reporting below-average shirking and

0 otherwise. We include the coastal-province dummy because these provinces contain favored locations that set them apart, geographically, politically, socially, and econom- ically, from much of the rest of China (Démurger, Sachs, Woo, and Bao, 2001) . They have been the most productive, fastest growing, and wealthiest provinces, particularly since reform. We conjecture that effort and incentives are affected if workers live in favored locations and, therefore, we do not want to bias our results by excluding this variable from the estimated relationship. The results reported in Table 6 are not sensi-

17 tive to the inclusion ofthe coastal-province dummy variable. The estimated coefficients are all highly significant, except that the coefficient ofthe efficiency-wage variable has a p-value ofonly 0.23. The pseudo- R2 for the probit regression is approximately 0.5. The estimated coefficients ofthe ownership variables imply that shirking is most likely to be observed, or reported, among workers in SOEs than in the other ownership forms.13

Joint-venture enterprises are, on average, about 11% more likely to have below-average shirking than are SOEs. The comparable statistic for collectives is about 5% and for

firms located in the coastal provinces, about 2%. An increase in the efficiency-wage variable of1% is associated with a .016% decline in the probability that reported shirking is above average within ownership class groups, but this is a rather imprecise estimate, as indicated by the p-value ofthe corresponding coefficient in the probit equation. 14

4 Estimation Results for the Rural Sample: Panel Analysis

Table 7 presents the results ofestimating equation (1) forthe rural sample. The estimated production elasticities for aggregate labor and capital are close to the ordinary

TABLE 7 ABOUT HERE least-squares (OLS) results reported by Pitt and Putterman (1999) and to the generalized

18 least-squares (GLS) estimates reported by Dong and Putterman (1996) using the same data set. Compared to the results for the urban sample reported in Table 5, there is evidence ofunexploited scale economies. (Rejection ofthe null ofconstant returns and the important implications ofincreasing returns to scale forwages in rural collectives is discussed in Fleisher and Wang, 2001.) The principal diﬀerence in the estimated individual production elasticities between rural and urban samples is that the labor elasticity estimated for the rural sample is over twice the magnitude as that reported for the urban sample. The relative magnitudes of the labor and capital elasticities are about the same as those reported by Pitt and Putterman (1999).

The rural sample does not provide data on individual-worker schooling attainment.

Therefore, we have defined the efficiency wage for this sample to be the ratio of the deviation ofthe worker’s actual wage to the average wage paid to that type ofworker in the same province. The estimated coefficient ofthe efficiency-wage variable is highly significant, although smaller in magnitude than in the production functions estimated on the urban sample. Moreover, when the production function is estimated using the two categories ofworkers, their respective efficiency-wage coefficients are close in value to each other and ofapproximately equal statistical significance, which is in sharp contrast to the estimated coefficients for the urban sample. In contrast to the results for the urban

19 sample, the ratio ofthe estimated eﬃciency-wage coeﬃcients to their respective output- labor elasticities is smaller than unity. This result holds for all workers taken together as well as for production and TAS workers separately in column (3). Notice that column

(4) requires more complex calculations. Taken at face value, these coefficients suggests that wage rates are set higher than their profit-maximizing level based on an efficiency- wage calculation. The sharp contrast between the too-low wage rates estimated for the urban sample and the too-high wage rates estimated for the rural sample constitutes a puzzle yet to be solved. The result is even more surprising ifwe note that the wage data in the urban sample are augmented to reflect housing payments in kind, whereas there is no information in the rural sample that permits this adjustment.

The panel nature ofthe rural sample permits us to explore the efficiency-wage hypothesis in more depth. The fundamental question is whether the estimated relationship between our efficiency-wage measures and output represents a genuine productivity- enhancing effect or whether it reflects reverse causation in the form of bonus-sharing, as suggested by Coady and Wang (2000a). Table 8 reports the results ofGranger Causality tests ofthe hypothesis that the efficiency-wage measure

TABLE 8 ABOUT HERE used in the estimates reported for the rural sample cause, in the sense of lead, productiv-

20 ity growth. We deﬁne total factor productivity (TFP) in the usual way, as the residual from a regression of log output on log capital and log employment, in the form of equation (1), with the eﬃciency-wage variable omitted. The estimates reported in columns

(1) and (2) ofTable 8 show that the lagged efficiency-wage variable is a statistically significant predictor ofTFP, whereas the estimates reported in columns (3) and (4) indicate that lagged TFP is not a statistically significant predictor ofthe efficiency-wage.

Therefore, the hypothesis that paying an eﬃciency wage leads to higher productivity in the rural sample sample cannot be rejected.

5 Conclusion and Agenda for Future Research

Consistent with previous research, we find evidence ofproductivity-enhancing wage effects in Chinese enterprises. However, much variation exists in the degree to which the profit-maximizing potential ofincentive-wage payments is exploited forworkers of different skill levels and by enterprises both in different ownership categories and in different regions. First, productivity is higher in firms that pay wage rates above the norm as measured by the average paid to all workers with similar characteristics. Second, evidence from panel data reinforces the causal relationship between profit- or rent sharing and productivity. Third, evidence from cross-section data indicates that shirking is less

21 likely in enterprises that pay higher efficiency wages. Fourth, incentive wages do not appear to be set at profit-maximizing levels, although there is weak evidence that in the urban sample the productivity-enhancing potential ofefficiency wages is more fully exploited by joint ventures and collectives than by SOEs and also by enterprises located in the coastal provinces compared with those in the interior.

In ongoing research, we hope to learn why efficiency wages appear to be lower in urban enterprises and higher in rural enterprises than would be warranted by simple profit maximization. We also hope to learn why incentive-wage effects appear to be more fully exploited for managerial-level workers in urban enterprises. The research reported in this paper describes enterprises’ wage-setting behavior at the intensive margin, where the tradeoff is between higher wage costs and higher productivity per worker. In future research, we hope to reconcile these results with our understanding ofenterprises’ wage-employment choices at the extensive margin, where the tradeoff is between the marginal cost ofemploying additional workers and the value oflabor’s marginal product. Results from research in that area show that technical and administrative workers are consistently underpaid, or underemployed, relative to production workers, so that private returns to schooling are pervasively low (Fleisher and Wang, 2001). This re- mains one ofthe important unresolved puzzles ofwage and employment behavior in

22 the Chinese economy.

23 References

Akerlof, George and Yellen, Janet L., Eds., Efficiency Wage Models of the Labor Market.

Cambridge, UK: Cambridge University Press, 1986.

Byron, Raymond P. and Manaloto, Evelyn Q., "Returns to Education in China." Economic

Development and Cultural Change 38, 4:783-796, July 1990.

Cahuc, Pierre and Dormont, Brigitte, “Profit-Sharing: Does it Increase Productivity and

Employment? A Theoretical Model and Empirical Evidence on French Micro Data.”

Labour Economics 4, 3:293-319, September 1997.

Chow, Gregory C., Understanding China's Economy. Singapore: World Scientific, 1994.

Coady, David P. and Wang, Limin, "Incentives, Allocation and Labour-market Reforms During

Transition: The Case of Urban China, 1986-1990." Applied Economics 32, 4:511-526,

2000a.

Coady, David P. and Wang, Limin, "Equity, Efficiency, and Labor-market Reforms in Urban

China: The Impact of Bonus Wages on the Distribution of Earnings." China Economic

Review 11, 3:213-231, Winter 2000b.

Démurger, Sylvie, Sachs, Jeffrey D, Woo, Wing Thye, and Bao, Shuming, "Geography, Economic

Policy and Regional Development in China." Paper prepared for the Third International

Conference on the Chinese Economy, Clermont-Ferrand, France, 2001.

Dessi, Roberta, "Income, Occupation, and Education in China." DAE Working Paper No. 9107.

Cambridge, UK: University of Cambridge, Department of Applied Economics, March

1991.

Dollar, David, "Economic Reform and Allocative Efficiency in China's State-owned Industry."

Economic Development and Cultural Change 39, 1:89-106, Oct. 1990.

24 Dong, Xiao-Yuan and Putterman, Louis, "China's Rural Industry and Monopsony: An

Exploration." Pacific Economic Review 1, 1:59-78, June 1996.

Dong, Xiao-Yuan and Putterman, Louis, "China's State-owned Enterprises in the First Reform

Decade: An Analysis of a Declining Monopsony." Working paper. Providence, RI: Brown

University, Department of Economics, 2000.

Fleisher, Belton M., Dong, Keyong, and Liu, Yunhua, "Education, Enterprise Organization, and

Productivity in the Chinese Paper Industry." Economic Development and Cultural Change

44, 3:571-587, April 1996.

Fleisher, Belton M. and Wang, Xiaojun, "Skill Differentials, Return to Schooling, and Market

Segmentation in a Transition Economy: The Case of Mainland China." Working paper.

Columbus, OH: The Ohio State University, Department of Economics, 2001.

Gordon, Roger H. and Li, David D., "The Effects of Wage Distortions on the Transition: Theory

and Evidence from China." European Economic Review 43, 1:163-183, Jan. 1999.

Gregory, R. G. and Meng, Xin, "Wage Determination and Occupational Attainment in the Rural

Industrial Sector of China." J.Comp. Econ. 21, 3:353-374, Dec. 1995.

Groves, Theodore, Hong, Yongmiao, McMillan, John, and Naughton, Barry, "Autonomy and

Incentives in Chinese State Enterprises." Quarterly Journal of Economics 109, 1:183-209,

Feb. 1994.

Gustafsson, Bjorn and Li, Shi, "Economic Transformation and the Gender Earnings Gap in Urban

China." Journal of Population Economics 13, 2:305-329, May 2000.

Huang, Tzu-Ling, Hallam, Arne, Orazem, Peter, and Paterno, Elizabeth, "Empirical Tests of

Efficiency Wage Models." Economica 65, 257:125-43, Feb. 1998.

25 Jamison, Dean T. and Van der Gaag, Jacques, "Education and Earnings in the People's Republic of

China." Economics of Education Review 6, 2:161-166, Feb. 1987.

Katz, Lawrence F., "Efficiency Wage Theories: A Partial Evaluation." In Stanley Fischer, Ed.,

NBER Macroeconomics Annual 1986. Cambridge, MA: MIT Press, 1986.

Knight, John and Li, Shi, "Educational Attainment and the Rural-urban Divide in China." Oxford

Bulletin of Economics and Statistics 58, 1:83-117, Feb. 1996.

Krueger, Alan B. and Summers, Lawrence H., "Efficiency Wages and the Interindustry Wage

Structure." Econometrica 56, 2: 259-293, March 1988.

Leibenstein, Harvey, "The Theory of Unemployment in Densely Populated Backward Areas." In

George Akerlof, and Janet L. Yellen, Eds., Efficiency Wage Models of the Labor Market,

pp. 22-44. Cambridge, UK: Cambridge University Press, 1986.

Li, Tianyou and Zhang, Junsen, “Returns to Education under Collective and Household Farming in

China.” Journal of Development Economics 56, 2:307-335, August 1998.

Li, Haizheng, "Economic Transition and Returns to Education in China." Working paper. Atlanta,

GA: Georgia Institute of Technology, School of Economics, 2001.

Maurer-Fazio, Margaret, "Earnings and Education in China's Transition to a Market Economy:

Survey Evidence from 1989 and1992.” China Economic Review 10, 1:17-40, Spring 1999.

Maurer-Fazio, Margaret, Rawski, Thomas G., and Zhang, Wei, "Inequality in the Rewards for

Holding up Half the Sky: Gender Wage Gaps in China's Urban Labour Market, 1988-94."

The China Journal No.41:55-88, Jan. 1999.

Meng, Xin and Perkins, Frances, "Wage Determination Differences between Chinese State and

Non-state Firms." Asian Economic Journal 12, 3:295-316, Sept. 1998.

26 Naughton, Barry, Growing Out of the Plan: Chinese Economic Reform, 1978-1993. Cambridge,

UK: Cambridge University Press, 1995.

Pitt, Mark M. and Putterman, Louis, "Employment and Wages in Township, Village, and Other

Rural Enterprises." In Gary Jefferson and Inderjit Singh, Eds., Enterprise Reform in China:

Ownership, Transition, and Performance, pp.197-215. Oxford, UK: Oxford University

Press, 1999.

Psacharopoulos, George, "Returns to Education: A Further International Update and Implications.”

Journal of Human Resources 20, 4:583-604, Fall 1985.

Shapiro, Carl and Stiglitz, Joseph E, "Equilibrium Unemployment as a Worker Discipline Device."

American Economic Review 74, 3:433-444, June 1984.

Sheidvasser, Sofia and Benítez-Silva, Hugo, “The Educated Russian’s Curse: Returns to

Education in the Russian Federation.” Working Paper. Stonybrook: SUNY Stonybrook,

Department of Economics, 2000.

Solow, Robert M., "Another Possible Source of Wage Stickiness." Journal of Macroeconomics 1,

1:79-82, Winter 1979.

Stiglitz, Joseph, "Theories of Wage Rigidities." In James L. Butkiewicz, Kenneth J. Koford,

Jeffrey B. Miller, Eds., Keynes' Economic Legacy: Contemporary Economic Theories, pp.

153-206. New York: Praeger, 1986.

Wang, Boqing, Zhu, Jianfang and Stromsdorfer, Ernst, "The Rate of Return to Education and

Experience for Workers of State-owned Enterprises in the People's Republic of China."

Paper presented at the meetings of the Southern Economic Association New Orleans,

November 18-20, 1995. Working paper. Pullman, Washington: Washington State

University, Department of Economics, 1995.

27 Weitzman, Martin L., The Share Economy: Conquering Stagflation. Cambridge, MA: Harvard

University Press, 1984.

Yang, Dennis Tao and An, Mark Yuying, "Human Capital, Entrepreneurship, and Farm Household

Earnings." Journal of Development Economics (forthcoming).

Yuen, Terence, "The Evolution of the Link between Workers' Pay and Enterprise Performance in

China's State Sector 1980-1994." Paper presented at the annual meeting of the Association

of Comparative Economic Studies, New Orleans. Ottawa, Canada: Bank of Canada,

Research Department, 2000.

Zax, Jeffrey S., "Human Capital in a Worker's Paradise: Returns to Education in Urban China."

Discussion Paper in Economics 94-7. Boulder, CO: University of Colorado, Department of

Economics, 1994.

Zhuang, Juzhong and Xu, Chenggang, "Profit-Sharing and Financial Performance in the Chinese

State Enterprises: Evidence from Panel Data." Economics of Planning 29, 3:205-222, Fall

1996.

28 Table 1: Sample Statistics For Employees: Urban Sample

Variable Description Mean Std

Age years 34 10 Sex dummy, 1 if male 0.55 0.50 Experience years 15 10 Tenure years 10 9 Education –High School dummy, 1 if only high school graduate 0.44 0.50 –College dummy, 1 if at least college graduate 0.22 0.41 Job Title –Production Worker dummy, 1 if production worker 0.28 0.45 –Technical/Administrative Staﬀ dummy, 1 if tech./adm. staﬀ 0.48 0.50 Location –Coastal dummy, 1 if coastal 0.44 0.50 –Non-Coastal dummy, 1 if non-coastal 0.56 0.50 Income and Benefits –Wage yuan/yr 2470 1229 –Housing dummy, 1 if public housing 0.91 0.29 –Medical Expenses dummy, 1 if public medical care 0.84 0.36

Note: Sample size is 9397.

29 Table 2: Sample Statistics For Enterprises: Urban Sample

Variable Unit Mean Std

Gross Output mil.yuan 70.455 225.408 Total Employment person 1376 3733 Net Capital mil. yuan 29.300 110.520 Ownership Classification –State percent 0.40 0.49 –Collective percent 0.34 0.47 –Joint Venture percent 0.25 0.43 –Private percent 0.01 0.11 Location –Coastal percent 0.47 0.50 –Non-Coastal percent 0.53 0.50

Notes: (i)Sample size is 422. (ii) Ownership classification and location are dummy variables. For example, the variable state has a mean of0.40, which means that 40% ofthe surveyed ﬁrms are state-owned enterprises. (iii) Coastal provinces in the sample are Hebei, Jiangsu, Shandong, Fujian, Guangdong, and Hainan. Non-coastal provinces Shanxi, Jilin, Anhui, Hunan, Henan and Sichuan.

30 Table 3: Sample Statistics For Enterprises: Rural Sample

Gross Output Net Capital Employment UNIT 10,000 yuan 10,000 yuan person 1984 336.139 65.183 298.475 (593.218) (88.725) (309.381) 1985 486.576 96.718 369.968 (849.214) (132.202) (354.787) 1986 532.570 162.001 419.590 (1025.870) (289.581) (472.879) 1987 624.188 237.213ii 409.046 (1231.110) (509.598) (521.383) 1988 631.756 171.432 387.355 (1018.180) (277.068) (439.702) 1989 709.366 209.628 382.523 (1262.700) (358.718) (438.833) 1990 757.204 214.514 374.410 (1317.700) (353.842) (481.779)

Notes:(i) Sample size is 200.

(ii) Standard deviations are in parentheses.

(iii) In this year only, the original price of ﬁxed capital is used. Therefore, this year is not used in the estimations.

(iiv) The provinces in this sample are: Hebei, Guangdong, Liaoning, Jiangsu, Zhejiang (all deﬁned as coastal), and Anhui, Hubei, Sichuan, Shanxi, and Gansu.

31 Table 4: Wage Equation: Urban Sample

Variable Description Estimate Std P-Value

CONST. 6.9485 0.0745 0.0001 AGE 0.0211 0.0042 0.0001 AGE2 -0.0002 0.0001 0.0001 SEX 1 if male 0.1432 0.0107 0.0001 EDUH 1 if a high school graduate 0.0841 0.0122 0.0001 EDUC 1 if at least college graduate 0.2277 0.0147 0.0001 EXPER experience in year 0.0136 0.0018 0.0001 No. of Obs.=7318, Adj. R2=0.20

Notes: (i) Dependent variable is ln W

(ii) 7318 out of 9397 observations provided suﬃcient information for the estimation.

32 Table 5: Production Function : Urban Sample

Variable Description (1) (2) (3) (4) CONST. 7.3411 7.4902 8.4293 8.5810 (14.238) (14.128) (14.169) (13.736) K ln net capital stock 0.4761 0.4701 0.3878 0.3801 (11.494) (11.228) (8.117) (7.698) L ln total employment 0.3324 0.3283 - - (7.336) (7.145) PW ln production workers - - 0.0260 0.0233 (0.614) (0.541) TAS ln technical/administrative staff - - 0.467 0.469 (7.180) (7.008) MT efficiency wage 1.1497 1.5506 - - (7.541) (4.480) MPW efficiency wage for - - 0.4454 0.2546 production workers (2.221) (0.795) MTAS efficiency wage for - - 0.6587 1.212 technical/administrative staff (2.973) (2.533) MT 2 interaction between MT - -0.2214 - - andcollective dummy (-0.593) MT 3 interaction between MT - -0.3413 - - andjoint venture dummy (-0.942) MT C interaction between MT - -0.2632 - - andcoastal dummy (-0.712) MPW 2 interaction between MPW - - - 0.1855 andcollective dummy (0.371) MPW 3 interaction between MPW - - - 0.3556 andjoint venture dummy (0.708) MPW C interaction between MPW - - - 0.1244 andcoastal dummy (0.296) MTAS 2 interaction between MTAS - - - -0.4312 andcollective dummy (-0.675) MTAS 3 interaction between MTAS - - - -0.5027 andjoint venture dummy (-0.921) MTAS C interaction between MTAS - - - -0.3974 andcoastal dummy (-0.809) No. of Obs. 319 319 262 262 Adj. R2 0.69 0.69 0.73 0.72

Notes:(i) Dependent Variable is ln Y . (ii) Total sample size is 442. However, only 319 firms provided sufficient information on aggregated employment, while 262 provided sufficient information on disaggregated employment. (iii) t-statistics are in parentheses. (iv) Y is gross output. (v) The efficiency wages are the residuals from the regression reported in Table 4. 33 Table 6: Probit Estimates of the Probability of Not Shirking: Urban Sample

Variable Estimate Std χ2 P-Value

CONST. 0.162 0.009 320.288 0.000 Collective 0.0489 0.0118 16.844 0.000 JointV enture 0.1059 0.0156 45.923 0.000 Coastal 0.0236 0.0121 3.861 0.049 EW 0.0163 0.0137 1.435 0.231 No. of Obs. = 8238 Log Likelihood = -4575.766 Pseudo R2=0.498

Notes: (i) Dependent Variable: 0 if shirking or 1 if not-shirking.

(ii) The table is based on 9397 surveyed employees. 8238 employees provided sufficient information for the estimation. EW is the wage differential at the individual level and is different from MT in that the latter is simply the average of all interviewed employees in the same firm.

(iii) The estimated coeﬃcients and standard errors are measured in probability units calculated at the means of the respective regressors. Calculations are available from the authors on request.

34 Table 7: Production Function : Rural Sample

Variable Description (1) (2) (3) (4) CONST. -1.3289 -1.2780 -0.7898 -0.7420 (-6.346) (-6.141) (-3.962) (-3.764) K ln net capital stock 0.4221 0.4257 0.3998 0.4057 (13.728) (13.950) (12.677) (13.007) L ln total employment 0.7376 0.7294 - - (16.645) (16.576) PW ln production workers - - 0.4934 0.4851 (9.888) (9.833) TAS ln technical/administrative staff - - 0.3024 0.2967 (5.912) (5.870) MT efficiency wage 0.3636 0.6763 - - (6.213) (7.068) MPW efficiency wage for - - 0.2048 0.5020 production workers (3.168) (4.775) MTAS efficiency wage for - - 0.1724 0.3437 technical/administrative staff (2.533) (3.388) MT C interaction between MT - -0.4803 - - andcoastal dummy (-4.111) MPW C interaction between MPW - - - -0.4136 andcoastal dummy (-3.150) MTAS C interaction between MTAS - - - -0.2632 andcoastal dummy (-1.977) YR85 year 1985 dummy 0.2383 0.2205 0.2540 0.2557 (2.059) (1.919) (2.184) (2.224) YR86 year 1986 dummy 0.0288 0.0106 0.0545 0.0495 (0.248) (0.092) (0.470) (0.431) YR88 year 1988 dummy 0.4875 0.4707 0.4893 0.4822 (4.262) (4.145) (4.266) (4.252) YR89 year 1989 dummy 0.5178 0.4981 0.5081 0.4971 (4.461) (4.322) (4.361) (4.311) YR90 year 1990 dummy 0.6322 0.6053 0.6328 0.6175 (5.479) (5.280) (5.434) (5.364) No. of Obs. 988 988 979 979 Adj. R2 0.60 0.61 0.60 0.61

Notes:(i) Dependent Variable is ln Y . (ii) t-statistics are in parentheses. (iii) 1987 observations are dropped due to inadequate data. The maximum number of observations is 1200, but the discrepancies reﬂect missing values. (iv) Y is value added. (v) The eﬃciency wage is the ratio of the actual wage to the average wage paid to the group of workers in that province.

35 Table 8: Granger Causality Tests: Rural Sample

Variable Description (1) (2) (3) (4)

CONST. 0.1028 0.1261 -0.0517 -0.030 (3.374) (3.454) (-3.032) (-1.497)

TFP−1 ln TFP lagged 1 period 0.4777 0.3955 0.0183 0.0122 (16.503) (9.302) (1.159) (0.535)

TFP−2 ln TFP lagged 2 period - 0.1191 - -0.0128 - (2.890) - (-0.591)

EW−1 ln EW lagged 1 period 0.2196 0.1245 0.5964 0.4983 (3.548) (1.380) (17.457) (10.172)

EW−2 ln EW lagged 2 period - 0.2085 - 0.1788 - (2.356) - (3.730)

No. of Obs. 770 577 795 595 Adj. R2 0.29 0.27 0.29 0.30 F test statistics

EW−1 =0,EW−2 =0 - 8.3242 - - - (0.0003) - -

TFP−1 =0,TFP−2 =0 - - - 0.2114 - - - (0.8095)

Notes:(i) Dependent Variable: ln TFP (total factor productivity) for (1) and (2); ln EW (eﬃciency wage) for (3) and (4).

(ii) t-statistics are in parentheses for the top panel and p-values for the bottom panel.

(iii)lnTFP is the residual from the regression of log output on log capital and log employment.

(iiv) Eﬃciency wage is the ratio of actual wage and average wage paid to the same type of worker in the province.

36 Notes 1 We thank Xiao-yuan Dong, Guillaume Frechette, Hajime Miyazaki, Stephen Still- man, participants at the First EALE/SOLE World Conference, 2000, Milan, Italy and at the 2001 annual meeting ofthe American Economics Association fortheir comments. The paper has beneﬁted signiﬁcantly from the suggestions of two anonymous referees and the Editor.

2We note that ex-post profit sharing may diminish work incentives through free riding ifit is not accompanied by a perceived risk to workers ofbeing forcedto find less remunerative work in alternative employment should they be fired or laid off.

3We emphasize that bonus schemes and proﬁt sharing are neither necessary nor suf- ﬁcient conditions for workers to be paid more than they could earn in alternative employment.

4The urban survey was funded in part by a Ford Foundation grant to the Institute of Economics ofthe Chinese Academy ofSocial Sciences and in part by the Labor Science Research Institute ofthe Ministry ofLabor ofthe People’s Republic ofChina. We are grateful to Ernst Stromsdorfer, Elizabeth Li, and Jun Cao for making us aware of these data and for their help in using them.

5We are grateful to Dennis Yang, Yaohui Zhao, Xiao-yuan Dong, Isabelle Perrigne, and Gary Jeﬀerson for their help in obtaining and using these data.

6Although the value ofin-kind housing provided by employers is included in earnings, housing allocated by a spouse or a parent’s employer is excluded. In addition, individual speciﬁc dummy variables such as location and occupation dummies are not included because they are most likely to be related to the eﬃciency wage itself; see Krueger and Summers (1988).

7Papers on China include Jamison and Van Der Gaag (1987), Dessi (1991), Byron

37 and Manaloto (1990), Fleisher, Dong, and Liu (1996), Gregory and Xin (1995), Maurer- Fazio (1999), Maurer-Fazio, Rawski, and Zhang (1999), Psacharopoulos (1985), Wang, Zhu, and Stromsdorfer (1995), Knight and Li (1996), Li and Zhang (1998), Zax (1994), and Yang and An (forthcoming). Although returns to higher education in the Russian Republic are among the lowest in the world, this can be attributed to the extraordinarily high proportion ofcollege graduates, over 20% ofall individuals aged 25 to 64 in 1995. This percentage is nearly equal to that in the United States and higher than the average for OECD countries (Sheidvasser and Ben´itez-Silva, 2000).

8We experimented with alternate measures, including using provincial-level estimates ofintermediate inputs. The resulting estimates are insensitive to these alternative spec- iﬁcations.

9Re-estimation ofequation (1) with residuals calculated fromequation (2) that includes provincial dummy variables yields results quite similar to those reported in Table 5.

10The estimates reported in column (2) include interaction terms between the eﬃciency- wage variable and ownership-class dummies for collectives and joint ventures, so that SOEs occupy the residual ownership category.

11The numbers are (1.21 − 0.43)/0.47 and 1.21/0.47, respectively.

12The exact question is as follows:

“Sometimes, some workers employed in your enterprise might not work every hard, but they are not in obvious violations ofthe enterprise’s work rules. How oftendoes this problem arise in your enterprise?

1. Very often; 2. Average; 3. Infrequent.”

For simplicity, the ﬁrst two answers are grouped as 0 while answer 3 is treated as 1 in the probit estimation. For 9397 respondents, 22.7% chose answer 1, 36.5% chose answer

38 2, 31.2% chose answer 3, and 9.6% did not answer the question.

13SOEs are represented by the omitted dummy variable and thus their estimated eﬀect is less than for the other ownership forms.

14Re-estimation ofthe shirking probit equation by including residuals calculated from equation (2) with provincial dummy variables included yields results that diﬀer little from those reported in Table 6.