Efficiency wage hypothesis--the case of Pakistan.

Abbas, Syed Kanwar ; Zaman, Asad

I. PROBLEMS WITH THE CLASSICAL THEORY OF LABOUR MARKETS

The goal of this section is to point out the observed difficulties with the classical/neoclassical theory of labour markets. According to classical and neoclassical economics the labour market is a market like any other market. The equilibrium wage is determined by the intersection of the supply and demand for labour.

[ILLUSTRATION OMITTED]

It is important to note that the labour supply and demand are determined by real as opposed to nominal wages. As depicted, the equilibrium wage is real [(w/p).sup.*] and the equilibrium labour supply is [L.sup.*]. If this classical theory is correct, then it has several important observable consequences. We list some of them, which are relevant to our study below:

(1) Theory. There is no involuntary unemployment. All unemployment is voluntary since those who are unemployed are on the portion of the labour supply curve above the equilibrium real wage [(w/p).sup.*]. This means that these people are willing to work only if the real wage is higher than the equilibrium wage. In particular, anybody who wishes to work at the going wage rate can find a job.

Observation. We find many people who are involuntarily unemployed. These people want to work at the going wage rate, but cannot find jobs. A recent Islamic University ad for Naib-Qasids produced a 1000 applicants for l0 positions. All these people were willing to work for the going wage rate, but they could not find jobs at that rate. This is in contradiction to the supply-demand equilibrium theory of classical/neoclassical economics.

(2) Theory. Classical theory offers some possible explanations for the observed involuntary unemployment. One is transitional or frictional unemployment. People change jobs for various reasons; firms may expand or contract due to changes in demand for their products, As people transit from one job to another, they may he temporarily unemployed. The unemployment we observe is of this kind--people who are temporarily unemployed and will soon find jobs (as our theory predicts) in equilibrium.

Observation. This means that the condition of involuntary unemployment is only temporary and should not persist for a long period of time. The Great Depression, and many other historical episodes, show that this is not true, Large numbers of people want jobs at existing and prevailing real wage rates, but cannot do so for long periods of time.

(3) Theory. Another possibility, closely related to (2) above, is that we may have temporary disequilibrium. The real wage may be higher than the equilibrium wage. In this case we see people who wish to work at the going wage, hut cannot find jobs. Such disequilibrium must be temporary according to classical/neoclassical theory. The mechanism which will eliminate disequilibrium is the following. Those who are unemployed will offer to work for less than the going wage rate. The firm will have every incentive to hire new workers at lesser wage, and fire/retire other workers who are currently working at a wage above equilibrium. This process will lower the real wage until it reaches equilibrium.

Observation. This process does not seem to work in practice. Even when there were a thousand applicants for the Naib-Qasid job, the university did not reduce the prescribed real wage. Nor did the applicants offer to work for less. Similarly, there are many historical situations where, despite large and widespread unemployment, the wages of those who are employed are not reduced. Workers equivalent in skill to existing workforce and willing to work for substantially lower wages are not hired.

(4) Theory. Real wages depend only on worker characteristics, and not on firm/industry characteristics. There is an aggregate demand for labour (of any type) which is the sum of the demand functions over all the firms. Similarly, there is an aggregate supply curve. The intersection determines the equilibrium wage. A less technical way to make this point is to say that all firms compete in the same market. If one firm offers higher wages than other firms (for unskilled workers, say) all workers will flock to this firm. By the equilibrating mechanism described in (3) above, the excess supply of labour available to this firm will drive down the wage until it reaches equilibrium. Similarly, if one firm offers less than what other firms are offering for unskilled labour, no one will be available to work for the firm and the firm will be forced to increase its wages to hire anyone.

Observation. There are large and persistent wage differentials across industries. Unskilled labourers with equivalent qualifications have different wages in hospitals, construction sector, government sector, etc. Such differences could arise in temporary disequilibrium, but should get eliminated in the long run as wages move towards equilibrium. However, the data shows no such tendency for wage differentials to become less over time.

(5) Theory. A large pool of involuntarily unemployed labourers will exert a downward pressure on real wages, but will not have an effect on the productivity of firms.

Observation. While we find that wages are resistant to change even in presence of large pool of unemployed labourers, we find that productivity of the firms increase in presence of such a pool. This is in conflict with the classical/neoclassical view of labour markets.

The above observations show that there are several difficulties with the classical and neoclassical view of how labour markets work. The efficiency wage theory provides an alternative model for labour markets which seems to be more compatible with the observations described above.

II. HISTORICAL BACKGROUND FOR THE EFFICIENCY WAGE HYPOTHESIS

It was the Great Depression which led to the downfall of classical economics. The large-scale unemployment which persisted for a long period of time was impossible to explain by classical theories. One of the key ideas of the classical economists was that the market regulates itself and provides the best possible outcomes without any interference from the government. It was plain for all to see that the Great Depression was not the best possible outcome--the economy had been capable of functioning in a much better way, providing more goods and services to all of the population. One of the main contributions of Keynes (1936) was to say that the market forces did not guarantee full employment. He argued that the labour market was peculiar and different from other markets. One could have large scale and persistent unemployment. Suitable government policy was needed when the effective demand for products was not sufficient to generate full employment. In such situations, appropriate government policy would increase demand and lead to full employment. Keynesian views were dominant in economics until the 70s when classical theories made a comeback. This was possible mainly because the Great Depression had faded from memories of most of the population. Furthermore, problems of stagflation created by the oil crisis showed up some weaknesses in Keynesian theories.

The neoclassicals argued that the labour market was just like other markets and did reach equilibrium rapidly. It was necessary for them to find an alternative to labour market failure to explain the Great Depression. Friedman and Schwartz (1963), argued that government mismanagement of the money supply led to the Great Depression. Other authors have offered other explanations. See for example Bernanke (2004) who argues that the rigid Gold Standard was the main contribution factor in the Great Depression. The causes for the Great Depression have been hotly disputed and it is not our intent to go into this controversy here. Rather, we will focus on arguments developed in support of Keynesian ideas, sometimes labeled "new Keynesian" economics. One of the key weaknesses of the Keynesian position was the idea that the labour market can stay in disequilibrium for a long period of time--it requires government intervention to fix this problem. Why should this be? Keynes himself did not provide an explanation. He said that wage bargains were conducted in nominal terms rather than real terms, and also that this was not rational, but this was how the world worked. The main justification offered for failure of equilibrium in labour market was "sticky wages". Real wages could not be pushed downwards. Observations from the Great Depression and other episodes of long term high unemployment provided empirical support for this idea, but there was no theoretical explanation of why this should be the case.

Under pressure from the neoclassical attack in the 70s neo-Keynesians tried to defend the idea of sticky wages. They wanted to find an explanation of why the labour market fails to function like other markets. One of the main arguments that has been developed in this context is the "'Efficiency Wage" hypothesis. According to this hypothesis, higher wages lead to more efficient performance by the workers. If true, this would explain a lot of the observed phenomena discussed in the previous section. The classical and neoclassical have a strong ideological commitment to the idea the free markets work and provide best possible outcomes for society.

Efficiency wages support Keynesian ideas that government interference is required to fix problems arising from free markets. Therefore neoclassical have strongly resisted the idea of efficiency wages and have attempted to find alternative explanations for these phenomena. They have also attacked the idea of efficiency wages on many different grounds. Our goal in this paper is to review the evidence for and against efficiency wages in the literature, and also especially in the context of the labour market in Pakistan.

III. VARIANTS OF THE EFFICIENCY WAGE HYPOTHESIS

Before proceeding to examine the evidence, we provide some more detail about the efficiency wage hypothesis. Why does paying higher wages increase efficiency? There are many possible mechanisms which have been postulated for this purpose. Each of these leads to a different variant of the efficiency wage hypothesis.

(1) "Nutritional Efficiency". An early idea due to Leibenstein (1957) is that the equilibrium wage is so low that workers cannot feed himself and his family properly. In this case he will not have enough energy to work well. Giving him a higher wage will allow him to feed himself and will increase his output at work. If correct, this effect would operate only for low wage earners--white collar workers and other high wage earners should not be subject to this effect. Substantial rise in real wages in the developed countries has reduced or eliminated the number of labourers working at or near the subsistence level, so this hypothesis is no longer seen in the literature. Efficiency wages are seen at higher wage levels as well, so that some other effect must be responsible. Nonetheless, the hypothesis may still have some validity in LDC's where many wage earners earn very low wages. Some empirical evidence for this "nutritional effect" may be available by looking at sick leaves and/or medical insurance payments for low wage earners and comparing them with the same for high wage earners.

(2) The Adverse Selection Model. This model, due to Weiss (1980), assumes that better workers have better alternative offers. Firms set higher wages to attract a large "hiring pool" of the applicants who are heterogeneous in their ability to work and, in this way, they select the best workers from large pool. Firms have an incentive to pay higher wages if there is positive correlation between the average quality of the worker and wage rate. Simply, firms like to have a good pool of applicants for their jobs so that they may select among them. If the firm does a good job at selection via its tests and interviews, it will be able to pick up workers of better quality than otherwise. This gives the firm incentive to offer higher than equilibrium wages.

(3) The Gift Exchange Model. Partial gift exchange hypothesis by Akerlof (1982, 1984) is an efficiency wage theory based on sociological factors. This model takes into account 'non economic variables', Akerlof argues that people will work hard with higher wages when there is even no threat of dismissal from job. He interprets the model as a "gift-exchange" between the firm and its workers. Simply, when firms pay higher wages in excess of the competitive wage, the workers feel obliged and reciprocate with repaying in the form of the gift of higher effort level. According to the basic idea of the "'labour market as partial gift exchange", the loyalty of workers is exchanged for high wages, and this loyalty results in high productivity of the firm.

(4) Fair Wages. This model was developed by Akerloff and Yellen (1990). Workers have some fair-reference wage, and firms have an incentive to pay wages that are closer to worker's fair reference wage. Firms which pay less than the fair wage create dissatisfaction, low morale, high quit rates, shirking and absenteeism on the job, as therefore receive less productivity from their workers. Fair reference wage depends upon a number of factors as given below:

(i) Fair reference wage may correlate with firm's profit opportunities and hence high profit firms are forced to pay higher wages to draw out the required level of effort.

(ii) If higher profit opportunities are associated with higher marginal product of effort, firms have an incentive to exploit higher profits by paying higher wages more than competitive wages.

(iii) Fair reference wage may depend upon the previous wage periods and wages paid to the workers across different firms with similar human characteristics like age, education etc.

(5) The Shirking Model. This version of Efficiency Wage Theory has been developed by Shapiro and Stiglitz (1984), Bowles (1985), Fehr (1986) and others. The problem confronting the employers is to minimise shirking because employees shirk on their jobs whenever they find opportunity. Monitoring is imperfect and costly for the firms so the payment of wages to the workers in excess of the current competitive market wage is an effective way to discourage shirking. At the competitive wage, workers fired for shirking can easily find other jobs at the going wage rate. In equilibrium, all workers are paid above the market-clearing wage and, as a result, the consequent unemployment acts as a 'worker discipline device'. In this way, cost of job loss will increase the firm's output. The firm can hire a worker at low wage but it knows that it is in favour of worker to shirk on the job. Another hypothesis associated with the Shirking Model is that firms should pay high wages to the workers in the occupation where poor work performance can cause larger damage to the firm. [Romaguera (1991)].

III. WAGE DIFFERENTIALS ACROSS INDUSTRIES

An important piece of evidence in favour of the EWH is that there are stable and persistent differentials in real wages across industries. Early studies which established the existence of such differentials were Slitcher (1950) and Dunlop (1957). More sophisticated recent studies will be discussed later. If labour is homogenous, and job characteristics are the same, then this observation contradicts the neoclassical theory which maintains that there should be only one equilibrium wage in the labour market as a whole. This section examines the data on wage differentials in the context of Pakistan.

For Pakistan, Nasir (2000) has calculated wage differentials taking into account personal and structural factors that determine compensation package for public and private sector. He concluded that private sector pays higher wages for the identical characteristics.

Based on annual industry level wage data given in government publication 50 Years of Pakistan, we analysed wage differential in eight industries for which a complete time series from 1964 to 1994 was available. These were Textile, Engineering, Minerals and Metals, Chemical and Dyes, Paper and Printings, Wood Stone and Glass, Skin and Hides and Miscellaneous. The data is given in the table below:

 Mineral Chemical
 and and
Year All Textile Eng. Metals Dyes

1964 1564 1387 1381 1377 1491
1965 1452 1473 1502 1645 1280
1966 1572 1730 1660 1583 1835
1967 1131 1822 1582 1660 1721
1968 1644 1897 1625 1553 1966
1969 1887 1811 1885 1656 2033
1970 1833 1716 1859 1711 2012
1971 1807 1855 1594 1864 2533
1972 1955 1969 1786 2091 2153
1973 3036 2766 2378 3199 6578
1974 3630 3458 3577 3138 4521
1975 4243 4394 3911 3616 4667
1976 4661 4325 4731 4619 5669
1977 5860 5284 6807 4842 7217
1978 7346 6910 7747 6183 7874
1979 7600 6976 8185 6239 8532
1980 7099 5420 9106 8866 8657
1981 7527 5940 9313 8437 8273
1982 8436 7157 10806 12363 8853
1983 9023 7165 10787 11484 11217
1984 8561 8193 6879 8095 14715
1985 10579 8758 13054 13516 14319
1986 11138 9542 12005 13491 14557
1987 13905 10313 17889 24176 15860
1988 13571 9930 16153 16032 17127
1989 15477 11866 18250 20014 22547
1990 20820 19008 22687 2969 27889
1991 16314 14365 17330 22976 18776
1992 24259 17042 22234 21564 51179
1993 18019 17828 15197 21280 26837
1994 23469 17330 26560 25037 33082

 Wood
 Paper Stone Skin
 and and and
Year Printing Glass Hides Misc.

1964 1999 1217 1390 1113
1965 1665 1283 1801 1529
1966 1733 1401 1818 1641
1967 1836 1547 1825 1553
1968 2166 1727 2027 1432
1969 2046 1864 1802 1998
1970 2027 1525 1934 2584
1971 1161 1882 1754 1544
1972 1994 2043 2322 2141
1973 4421 2802 2824 6773
1974 6752 3392 3749 6185
1975 5135 3967 4483 8981
1976 6122 3095 5895 9971
1977 5837 5913 6276 11185
1978 5665 7046 7075 16190
1979 6005 7647 7115 7334
1980 8893 9668 6574 1902
1981 13767 6367 6387 9899
1982 13800 6822 9281 9726
1983 12375 6799 12389 11284
1984 10471 8594 4303 9525
1985 7843 13479 10478 8602
1986 14623 13867 II501 11324
1987 20186 19568 I4G56 14499
1988 16592 17860 13596 14137
1989 20625 19126 12567 14100
1990 22344 20299 21749 19517
1991 16254 19967 6851 19746
1992 25719 23249 26801 48217
1993 16495 17691 21578 16961
1994 37368 28776 23187 19801

A graph of the differences between the sectoral wage and the overall average across all industries is presented below. While there is widespread fluctuation in these wages, there does not seem to be any overall stability or persistence in the differences across industries. This is in contrast to the evidence from USA, where there are stable and persistent differences in wages across industries. A number of different methods were tried to assess the existence of a wage differential. In all cases, the textile industry appeared to offer significantly different wages from the rest, while in all other industries, there was no significant differential between the industry specific wage, and the overall wage in all industries. Below we indicate two methods, both of which led to this same conclusion.

[ILLUSTRATION OMITTED]

Method 1. The graph of the wages clearly shows that there is substantial and increasing heteroskedasticity with time. Let Wi(t) be the wage in the ith industry in year t, and let W(t) be the overall wage in all industries. According to conventional theory, any difference between the wage in the ith industry and the overall wage Di(t) = Wi(t)-W(t) can only be due to chance and random fluctuations. In particular, there should be no relation between Di(t) and Di(t-1). Thus, in a regression of Di(t) on Di(t-1), the coefficient on Di(t-l) should not be significant. In running this regression, it is crucial to take care of the heteroskedasticity which is evident from the graphs of the wage data. A number of ways of estimating the standard deviation and adjusting the data for heteroskedasticity were tried, all of which led to the same result. In all cases, the regression coefficient of Di(t-I) was not significant except for the textile industry. The table below presents the coefficient estimates for the eight regressions of Di(t) on Di(t-1). For heteroskedasticity, we partitioned the data in three time periods: (i) 1964-72, (ii) 1973-86, and (iii) 1987-96. For the three periods, we estimated the standard errors to be std(l)=220, std(2) = 2200, std(3) = 5500. The data was divided by these estimated standard errors prior to running the regressions reported in Table 1.

Except in the case of Textiles, the coefficients on last periods wage differential are not significant, showing that fluctuations away from overall average wage do not persist, and are temporary only. However, the differential between the textile wage and the overall industry wage is significant and also persistent across time.

Method 2. Another way of taking care of heteroskedasticity is to look at the rate of change. Define di(t) = log(Wi(t)/W(t)) to be the log of the ratio of the wage in the ith industry to the overall industry average. If wages across industries conform to the competitive labour market theory, then di(t) should be a purely random fluctuation, unrelated to di(t-1). If there are significant differences in wages across industries, then the regression of di(t) on a constant and di(t-1) should yield a significant coefficient for di(t-1). Running these regressions led to the same result as before--only the textile industry had a significant coefficient on lagged wage differential, while the other industries conformed to the competitive model.

Overall, we may summarise our findings for Pakistan by concluding that the wage differential for the textile industry appears to be stable and persistent across time, contrary to the neoclassical theories of the labour market. Other industries appear to conform to the competitive labour market structure, with wage differing by random and non-persistent amounts from the overall wage average.

Since the finding of wage differentials has been well established on US data, there has been substantial work on explaining why these wage differentials arise. Several explanations which conform to the neoclassical theory have been offered--these allow one to defend the neoclassical idea of efficiency of markets. Other explanations consistent with the efficiency wage hypothesis have also been offered. For a more complete discussion of the strengths and weaknesses of these alternative explanations for persistent and stable differentials in industry wages, [see Abbas (2006)]. it would be worth exploring in future research why the wages in the textile sector differ significantly from overall wages in over the time period examined in Pakistan, and also why there are no significant differences between wages in the other industries.

V. EFFICIENCY WAGES IN THE TEXTILE SECTOR IN PAKISTAN

From our preliminary investigation of the wage differential, it appears that the textile industry offers efficiency wages, while the others are competitive. In this section, we do a direct test of the efficiency wage hypothesis for the textile sector, replicating similar studies by Wadhwani and Wall (1991), and Levine (1992). We test the efficiency wage hypothesis at industry level rather than at the firm level because data is not available at firm level in Pakistan. A regression of log (Q), output of the textile industry, on capital, labour, and the relative wage variable log (W/[W.sup.*]), W is wages and [W.sup.*] is the Average Wage level in textile industry yields the following estimates:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (8)

The numbers in parentheses are the standard errors. All coeffients are significant at levels above 99 percent except the coefficient of Log(K) which has a p-value of 6.9 percent.

Data and Variables. The data used in this regression is obtained from various issues of Censes of Manufacturing Industries (CMI) for 19 years. The variables are:

Q: the value of production of Textile industry (Rs million) at the end of the year.

K: We use value of fixed assets (Rs million) at the end of the year.

E: the reported number of workers.

Wi: is calculated by dividing sum of total employment costs (Rs million) by Average number of workers employed during the year for textile industry.

W: is the Average Wage level in the manufacturing industry.

We use manufacturing price index with 1980-1981 as base year to deflate.

Discussion of Results. Our coefficients on all the variables are positive, significant and plausible in magnitude. The sum of the coefficients on capital and labour is nearly unity, so that constant returns to scale is observed. The key observation is that the coefficient on log (Wi/W) is positive and significant. According to neoclassical theory, the inputs of Capital and Labour determine the output, and coefficient of the wage ratio should not be significant. Indeed, the wage ratio Wi/W should be one in equilibrium and differences from 1 only represent temporary disequilibrium which should not impact on production. The significant positive coefficient corresponds to the prediction of efficiency wage hypothesis, according to which higher than equilibrium wages will result in increased output. Our estimated coefficient for log (Wi/W) is (.60) [t-ratio, 14.89]. In comparison, Levine reports (.46), Wadhwani and Wall (.39), Huang, Hallam, Orazem and Pater (1998) estimate ranged between (.19) to (.61) and Seref Saygili (.15). Our high coefficient shows that there is a significant impact of increased wages on productivity.

Solow Condition. Profit maximisation in an efficiency wage setting requires that the productivity gain from increasing wages should exactly offset the loss due to increased wage bill. It can be shown that this requires the coefficients of E and Wi/[W.sup.*] to be the same. This well-known result of the standard efficiency wage model is due to Solow (1979) and is known as the Solow condition. The efficiency wage theory suggests that Solow condition holds which implies that percent change in wage should lead to percentage change in effort level to such an extent it will be unity. However, if it does not hold, equilibrium is not achieved with unemployment in Efficiency wage models [Akerlof and Yellen (1986)]. (1) Imposing the Solow condition, the results for the constrained regression are as follows:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

The F-statistic for the constraint is 1.04 with p-value 0.37, comfortably for from rejection.

Thus it appears that the Solow condition holds for efficiency wages in the textile industry. This contrasts with findings of Saygili (1998) for the Turkish cement industry, where the estimated coefficient on wages is significantly smaller than the estimated coefficient on labour input. Similarly, Wadhwani and Wall (1991) also report that for selected UK manufacturing industries, the coefficient on relative wage (.39) is significantly less than the estimated coefficient on labour input (.65). Thus, while efficiency wages are present, the Solow equilibrium condition for efficiency wage models does not hold in the Turkish cement industry or in the UK manufacturing industries tested.

Effects of Unemployment. Conventional propositions of the standard neoclassical theory hold that the outside changes in the cluster of unemployment do not affect productivity of the firm. Conversely, the efficiency wage hypothesis suggests that the outside rates of unemployment have an impact on productivity of the firms. In this context, in earlier nineties, the effect of rate of unemployment on firm's productivity was analysed. Wadhwani and Wall (1991) using OLS and GMM techniques test the impact of unemployment on productivity with Cobb-Douglas production function with the data from published accounts of 219 UK manufacturing companies over the period 1972-1982. Efficiency wage hypothesis requires that the outside cluster of unemployment positively affects the output of the firm. Their findings show that the coefficient of unemployment is positively signed and statistically significant (.05) (2.12). On the other hand, when Huang, et al. (1998) add unemployment rate to the regression, the unemployment output elasticity is positive and ranges from (.06) to (.1l) which are also consistent with efficiency wage hypothesis.

In contrast to these typical findings reported above, unemployment, when added to the equation for textile industry in Pakistan, does not have significant coefficients. Thus, in contrast to predictions of the efficiency wage theories, outside pool of unemployment does not increase productivity in the textile industry in Pakistan. This finding supports "Fair Wage" and "Gift Exchange" models, but not the "Shirking Model" since large pools of unemployment would increase losses from shirking. Alternatively, at sufficiently high levels of unemployment, the loss from being fired may be so high that no one shirks. In such cases, further increases in unemployment would not change productivity. More investigation is needed to discover exactly why unemployment fails to affect productivity in the textile sector in Pakistan.

VI. CONCLUSIONS

Our main findings are that the textile sector in Pakistan offers efficiency wages, which differ significantly from overall average industrial wages. Other industries appear to be competitive, in that their wages do not differ significantly from overall industrial wages. Further investigation is needed to discover factors which result in efficiency wages in the textile sector but not in the other sectors. A direct estimate of the production function shows that the ratio of wages in textile industry to average wage level significantly affects textile production. This is impossible according to neoclassical theory, since their should be no significant differences between wages in textile sector and overall industrial wage. Efficiency wage theory predicts a positive coefficient for this variable, and also suggest that the coefficient on relative wage should equal that of log(labour)---the Solow equilibrium condition. While typical estimates in literature reject the Solow condition, our estimates for Pakistan accept the Solow condition. We also find that the outside pool of unemployed labour does not affect productivity in the textile sector. Again this last result is in contrast with typical findings in the literature. Thus, our investigation of efficiency wages in Pakistan show strong empirical support for the hypothesis, together with interesting local variations from results reported elsewhere. Further research is needed to determine the reasons for the variations.

Comments

The paper under discussion tests the Efficiency Wage Hypothesis (EWH) for the textile industry of Pakistan and concludes that for the period under study there is empirical evidence to support the EWH. In view of strengthening the overall theoretical and empirical analysis presented in the paper, the following points need to be taken into consideration by the authors:

(1) It is not clearly mentioned in the paper as to what is the period under study. Which year's Census of Manufacturing Industries (CM1) has or CMIs have been used. Table 2 gives summary statistics for the variables used but does not mention the source of data. What is U? (Mean U=1.03 ??) Is source of data on unemployment, the Labour Force Survey and if so, for which year?

(2) Regarding stable and persistent differentials in real wages across industries, the Table on Page 7 gives correlation coefficients for wages across industries in Pakistan. Given that the EWH is being tested specifically for the textile industry in Pakistan, how valid is the inference drawn from the correlation coefficients for the aggregate industry wage data.

(3) On Page 8, the authors claim that their research confirms the basic idea that workers with similar characteristics receive different wages in different industries. How are they controlling for similar characteristics, especially if this result is being inferred from the simple correlation coefficients in the Table on Page 7.

(4) How realistic is the perceived linkage from higher wages to higher productivity, (according to EWH), given that the thrust of current macroeconomic and socio-economic policy and planning framework in Pakistan is on promoting skill development linked with higher productivity leading to higher wages and lower unit costs of production!

(5) According to the LFS 2003-2004, the overall labour force profile of the Pakistani labour force reveals that only 13.73 percent of the employed persons 10 years and older are employed in the manufacturing industry (11.25 percent male + 2.49 percent female) in contrast with 43.05 percent employed in the agriculture sector. The authors should contextualise their results for the textile industry with the distribution of the employed labour force in Pakistan.

(6) Even though the paper does not contain any policy recommendations, it is worth noting that it draws attention to the importance of a Wage Monitoring Mechanism both at the aggregate as well as at a disaggregated level, especially in the context of tracking the wage-productivity relationship.

(7) As part of the on-going labour policy reform, the Government intends to build upon the institutional arrangements under the labour policy framework by establishing a National Wage Commission to work on a range of wage-related issues, including minimum wages. The purpose, specific functions, and operational arrangements for the National Wage Commission, including its technical and secretariat support requirements, will be elaborated in a separate and detailed policy paper to be prepared in close consultation with employers' and workers' organisations.

Aliya H. Khan

Quaid-i-Azam University, Islamabad.

REFERENCES

Abbas, Syed Kanwar (2006) Efficiency Wage Hypothesis in Pakistan. M. Phil Dissertation. International Islamic University, Islamabad, Pakistan.

Akerlot, George (1982) Labour Contracts as Partial Gift Exchange. Quarterly Journal of Economics 97:4, 543-63.

Akerlof, George (1984) Gift Exchange and Efficiency-Wage Theory: Four Views. American Economic Review 74, 79-83.

Akerlof, George, and Janet Yellen (1986) Efficiency Wage Models of the Labour Market. Cambridge: Cambridge University Press.

Akerlof, George, and Janet Yellen (1990) The Fair Wage-Effort Hypothesis and Unemployment. Quarterly Journal of Economics 105, 255-283.

Bernanke, B. S. (2004) Essays on the Great Depression. Princeton University Press.

Bowles, Samuel (1985) The Production Process in a Competitive Economy: Walrasian, Neo-Hobbesian and Marxian Models. American Economic Review 75, 16-36.

Dunlop, J. (1957) The Task of Contemporary Wage Theory. In J. Dunlop (ed.) The Theory of Wage Determination. London: Macmillan.

Fehr, Ernst (1986) A Theory of Involuntary Equilibrium Unemployment. Journal of Institutional and Theoretical Economics 142, 405--430.

Friedman, M., and A. Schwartz (1963) A Monetary History of the United States.

Huang, Tzu-Ling, Arne Hallam, Peter F. Orazem, and Elizabeth M. Peter (1998) Empirical Tests of Efficiency Wage Models. Economica 65, 125-43.

Keynes, J. M. (1936) General Theory of Employment, Interest, and Money.

Leibenstein, H. (1957) The Theory of Underdevelopment in Densely Populated Backward Area. In H. Leibenstein (ed.) Economic Backwardness and Economic Growth. New York: Wiley.

Levine, David (1992) Can Wage Increase Pay for Themselves? Test with a Production Function. The Economic Journal 102, 1102-15.

Nasir, Z. M. (2000) Earnings Differential between Public and Private Sectors in Pakistan. The Pakistan Development Review 39:2, 111-130.

Pakistan, Government of (Various Issues) Census of Manufacturing Industries. Islamabad: Federal Bureau of Statistics.

Pakistan, Government of (Various Issues) Economic Survey. lslamabad: Economic Advisor's Wing, Ministry of Finance.

Romaguera, Pilar (1991) Wage Differentials and Efficiency Wage Models: Evidence from the Chilean Economy. University of Notre Dame. (Kellog Institute Discussion Working Paper No. 153.) <http://kellogg.nd.edu/publications/workingpapers/>

Saygili, Seref (1998) Is the Efficiency Wage Hypothesis Valid for Developing Countries? Evidence from the Turkish Cement Industry. Studies in Economics, University of Kent at Canterbury, Kent. (Discussion Paper No. 9810.)

Shapiro, and Joseph Stiglitz (1984) Equilibrium Unemployment as a Worker Discipline Device. American Economic Review 74, 433-44.

Slitcher, S. (1950) Notes On the Structure of Wages. Review of Economics and Statistics 32, 80-91.

Solow (1979) Another Possible Source of Wage Stickiness. Journal of Macroeconomics 79-82.

Wadhwani and Wall (1991) A Direct Test of the Efficiency Wage Model Using U.K Micro-Data. Oxford Economic Papers 43, 529-48.

Weiss (1980) Job Queues and Layoffs in Labour Markets with Flexible Wages. Journal of Political Economy 88, 526-38.

(1) For details see Akedof and Yellen (1986) who discuss that there can be situations where equilibrium effort-wage elasticity' can be lower.

Syed Kanwar Abbas, lives and works in Kasur. Asad Zaman is Director-General, International Institute of Islamic Economics. Islamabad.

Table 1

 Standard
 Coefficients Error t-stat

Intercept -0.07 0.14 -0.55
Textile 0.51 0.16 3.09
Intercept 0.19 0.13 1.52
Engineering 0.23 0.18 1.27
Intercept 0.28 0.21 1.34
Mineral Metals -0.07 0.19 -0.36
Intercept 0.97 0.29 3.34
Chemical and Dyes 0.16 0.19 0.86
Intercept 0.44 0.26 1.69
Paper and Printing 0.31 0.18 1.70
Intercept 0.09 0.14 0.66
Wood Stone and Glass 0.27 0.17 1.60
Intercept 0.25 0.19 1.30
Skin and Hides 0.11 0.19 0.57
Intercept 0.77 0.32 2.41
Misc. 0.05 0.18 0.29

 Eight
 P-value Regressions

Intercept 0.59 Let TW*(t) be the
Textile 0.00 Wage in Textiles
Intercept 0.14 in year t divided
Engineering 0.21 by Std. Error. We
Intercept 0.19 regress TW*(t) on
Mineral Metals 0.72 constant and
Intercept 0.00 TW*(t-1). The
Chemical and Dyes 0.40 constant is
Intercept 0.10 estimated to be
Paper and Printing 0.10 -0.07, with t-slat
Intercept 0.51 -0.55. The
Wood Stone and Glass 0.12 coefficient on
Intercept 0.20 TW*(t-1) is 0.51
Skin and Hides 0.57 with the only
Intercept 0.02 significant t-slat
Misc. 0.77 (3.01) in the table.

 Standard
 Coefficients Error t-stat

Intercept -0.05 0.03 -1.58
Textile 0.53 0.17 3.14
Intercept 0.04 0.03 1.60
Engineering 0.27 0.18 1.52
Intercept 0.01 0.08 0.13
Mineral Metals -0.15 0.19 -0.81
Intercept 0.21 0.05 3.81
Chemical and Dyes 0.14 0.18 0.80
Intercept 0.11 0.05 2.11
Paper and Printing 0.31 0.18 1.67
Intercept 0.02 0.03 0.58
Wood Stone and Glass 0.33 0.17 1.91
Intercept 0.01 0.05 0.15
Skin and Hides -0.06 0.19 -0.34
Intercept 0.11 0.08 1.42
Misc. 0.28 0.18 1.58

 Eight
 P-value Regressions

Intercept 0.13 This table gives the
Textile 0.00 results for
Intercept 0.12 regressions of di(t)
Engineering 0.14 on di(t-1) for each of
Intercept 0.90 the eight sectors
Mineral Metals 0.43 indicated. The first
Intercept 0.00 two lines show that
Chemical and Dyes 0.43 di(t)=-0.05 + 0.53
Intercept 0.04 di(t-1)
Paper and Printing 0.11 for the textile industry. The
Intercept 0.57 coefficient o.53 of
Wood Stone and Glass 0.07 lagged di=ln(Wi/W)
Intercept 0.88 is significant only for
Skin and Hides 0.73 the textile industry
Intercept 0.17 and not significant in
Misc. O.13 all other industries.