The role of the firm in wage determination : an African case study

The conclusions and views expressed in this paper are those of the authors and do not necessarily reflect those of the World Bank Group. The paper is a-draft for discussion; please do not cite without the author's permission. DPH8110 Population and Human Resources Division Discuss on Paper No. 8110 THE ROLE OF THE FIR N WAGE DEERMINATION AN AFRICAN CASE STUDY, April 1981 Prepared by: J.B. Knight Institute of Economics and Statistics University of Oxford R.H. Sabot Uevelopment Economics Department Z The World Bank Washington DC 20433 USA Summary An establishment-based survey of employees in the Tanzanian manu- facturing sector, providing information not only on workers but also on their employers, is analyzed to test various hypotheses about wage determina- tion in developing countries. The relative importance of personal and firm characteristics is assessed: the firm characteristics are found to play a significant, albeit tho minor, role. The influence of certain fi.rm charac- teristics - such as ownership status, degree of unionization, size, and ability to pay - is examined. Some of the results run counter to the conven- tional wisdom, but a knowledge of the institutional context can prcvide explanations. 1. Introduction To what extent are wages in a developing country determined by the personal characteristics of employees and to what extent by the characte- ristics of their employers? Insofar as employers play a role in wage determi- nation, what are the firm characteristics which influence wages? These are the questions which we attempt to answer in this paper. We do so by means of an earnings function analysis of a sample drawn from the manufacturing sector of Tanzania. The establishment-based survey, conducted by one of the authors in 1971, covers about 1,000 employees in some 47 firms in Dar es Salaam. Information was gathered on the characteristics of the 24 larger firms (those with 50 or more employees), accounting for a total of 660 sampled employees. This permits the simultaneous analysis of the influence on wages of both personal and firm characteristics. Theories of wage determination focusing on individual productive characteristics and those focusing on group affiliation differ in their assumptions about the behavior of the labor market. In a perfectly competi- tive labor market group affiliation does not influence wages. Irrespective of differences among groups of workers in goods produced, in the technology or organization used to produce them, in the ownership or profitability of such production, its scale or its location, competition in the labor market will ensure that all workers with the same personal economic characteristics receive the same rate of pay. Differences in pay arise only if employers a discriminate among workers on the basis of personal non-economic characteris- tics or if the desirability of emptoyment varies among different jobs. - Implicit in theories of wage determination that emphasize group affiliation is -2- the assumption of labor market imperfections. Non-market forces which, directly or by restricting labor mobility, cause wages to depart from t'- competitive level, must be sufficiently powerful to prevent competition in the market from eroding wage differentials among homogeneous workers. In Section 2 wc sketch out various arguments, relating to the characteristics of firms, which could explain why group affiliation of this type matters. Most of the empirical work on the determination of wages has involved a testing of hypotheses within either the personal characteris- tics or the firm characteristics paradigm. In Section 3 we attempt a nesting of the two paradigms by including both firm and personal variables in che earnings function. Standardizing in this way for the personal variables, we examine the extent of segmentation among firms, and the relative importance of employer- and employee characteristics. We find that firms play a significant, albeit the minor, role. Sections 4-6 go on to examine the influence of particu- lar firm characteristics. Thus Section 4 assesses the role of firm ownership on wages and poses the question: do foreign firms pay more? Section 5 is concerned with the influence of trade union membership, and Section 6 with that of various proxies for ability to pay', such as firm size. Some of our results run counter to the conventional wisdom. However, recognition of the institutional context enables us to advance plausible explanations for them. Section 7 summarizes and concludes. 2. Why Should Firms Matter? Explanations of persistent inter-firm wage differentials for apparently homogeneous labor divide into two categories on the basis of the wage determination mechanism assumed to be at work. On one view, employers -3- are forced by employees to raise wages above the competitive level; on the other, employers may have reason voluntarily to raise wages. On the former view, wages are seen as the outcome of a process of -ollective bargaining. The outcome depends partly on the bargaining power of workers: the greater their degree of organization and their cohesive- ness, the greater their ability to impose Icosts' on employers through indus- trial action. The outcome also depends on the bargaining power of firms; the ability to pay high wages may reduce the will of employers to resist worker pressures. The other view of firm-related wage determination is that employers unilaterally raise wages above the floor provided by competitive market forces or minimum wage legislation. One explanation focuses on non- productivity benefits that may accrue to certain employers as a consequence of wage increases, the other on the effect of wages on labor productivity. There are thus four main sets of hypotheses; examples of each are introduced briefly below. It is a commonplace view that unionized firms pay higher wages than non-unionized firms, and that firms with strong unions or a high propor- tion of the labor force unionized will pay higher wages than firms with weak unions or a low proportion of workers unionized. Implicit here is the assump- tion that trade unions function much as they do in industrialized countries. However, the prediction of the effect of unionization on wages could be reversed if there was reason p believe that unions had become an instrument for implementing a government ;olicy of wage restraint. -4- Highly profitable firms will pay higher wages than firms operating at the margin. This cannot be a clear-cut prediction owing to two counter- arguments. The will to resist wage increases may be greater in highly profit- able firms if their exceptioinal p-irformance is due to superior management. Moreover, although it seems impidusible in the case of the many manufacturing firms ia which wage costs are a small proportion of total costs, high profit- ability could be the result of paying low wages. In addition to current profits, there are various other criteria for "ability to pay", reflecting long run profitability. One is high capital intensity: the lower the pro- portion of wage costs in total costs, the less sensitive are profits to a given wage increase. It is for this reason that technological dualism is sometimes said to give rise to labor market dualism. A positive relation between wages and capital intensity, however, supports this hypothesis only if it can be shown that the choice of technique was not in response to high wages. Secondly, ability to pay:may be positively associated with size of firm. Large firms may reap greater economies of scale, have more monopoly power and be more likely to be protected by government trade policies. In that case, they have a greater capacity to pay high wages both because of high profits and the ease with which they can sustain profits by passing on cost increases to consumers. It is commonly argued that foreign firms are willing to pay more than local firms. Multinational companies may see the paylent of high wages as a fo m of insurance in a hostile environment. It could-be a way uZ securing the loyalty of employees, of avoiding charges of 'exploitation', and of reducing political pressures for nritionalisation. Host governments may -5- choose not to discourage wage increases in multinationals if they regard higher wages as a means of wresting from foreign owners a larger national share of value added. This hypothesis is not independent of the ability to pay hypotheses: the multinationals must have the resources to pay high wages. However, it is often asserted that multinational operations in less developed countries tend to be highly profitable, as a result of their monopoly or collusive oligopoly positions, and that multinationals tend to use more capital intensive techniques of production, irrespective of the wage level, so making profits less sensitive to wages. There are various reasons why employers may perceive a relation- ship between wages and productivity. According to the 'efficiency wage' hypothesis, an increase in wages can improve the health, energy or morale of workers, and hence their productivity. According to the 'labor turnover' hypothesis, higher wages can diminish labor turnover in the firm, and hence reduce training costs. A third hypothesis is that, by raising wages relative to those in other firms, an employer can 'cream' the most productive members of the labor force. Wage differences among firms may in that case reflect differences in the productivity of their workers which are not captured by the personal tconomic variables in the multiple regression equations owing to the inevitable crueeness of the variables available as proxies for personal productivity. Because of differences in their characteristics, firms are liable to differ in the extent to which they wish to raise the wage for efficiency wage reasons or for labor turnovej reasons, and in the extent to which they find it profitable to cream workers. These differences are liable 6- to be revealed by firm variables in the regression analysis. In the first two cases, labor market segmentation exists among firms, in the sense that there are no differences in labor productivity at the time of hiring, and tL,2re is no tendency for wage differences to be reduced through competition. In the creaming case, the labor market is not segmented in so far as the wage differ- ences reflect differences in personal productivity. 3. The Magnitude of Segmentation Among Firms There were large differences in average wages among the firms of our sample. Table 1 indicates that average wages among the 24 larger firms ranged from 192 sh. to 722 sh. per month; the mean of the distribution was 379 sh. and its standard deviation 368 sh. But we need somehow to measure the extent to which this variation was due to differences among firms in labor force composition in order to isolate the influence of a worker-s firm on his wage. Even a simple comparison of average wages and average length of school- ing of employees in the various firms (Table 1) suggests that compositional differences were important: the correlation coefficient was 0.67. How best can the influence of firm affiliation be isolated? In the many earnings function studies for developing countries, the influence of firm affiliation on earnings is generally ignored because the-data used in most such studies are generated by urban or even national hougehold surveys. With such a widespread frame it is generally impracticable to gelect a sample sufficiently large to yield enough observations within individual enterprises for statistically significant results. Thus the few indirect assessments of the impact of firm affiliation are generally based on -7- Table 1: AVERAGE WAGES AND YEARS OF SCHOOLING BY FIRM, AND THE COEFFICIENTS ON THE FIRM DUMMY VARIABLES Average Average Years Coefficients on the Fin Wages of Schooling Firm Dummy Variable 1/ 1 721.9 6.1 0.26* 2 564.7 5.5 0.32* 3 545.4 4.6 0.34 4 544.8 4.3 0.47* 5 542.9 5.3 0.33* 6 532.0 6.2 0.29 7 488.2 6.4 0.19* 8 393.9 3.8 0.14 9 384.2 4.7 0.33* 10 367.3 4.4 0.14 11 358.8 3.8 0.14 12 341.8 4.0 0.01* 13 340.5 4.4 0.11 14 319.0 2.8 -0.07* 15 289.6 3.5 -0.13 16 275.2 4.8 -0.30* 17 271.5 2.5 -0.17* 18 267.3 2.2 0.10 19 259.2 1.5 0.12* 20 219.1 6.2 -0.15* 21 217.2 2.5 -0.25* 22 211.6 2.9 -0.04* 23 207.7 3.5 -0.21* 24 192.3 2.6 0.17* * Coefficient significant at least at the five percent level. 61/ The sub-category of 23 small firms is the bas& in the dummy variable analysis. -8- aggregative rather than individual data. A typical study might regress sectoral data of average educational attainment, occupational distribution, firm size, capital intensity and other characteristics of enterprises on average earnings by sector. Significance of a firm variable in such equations is interpreted as evidence in support of the general proposition that, for wages of individual workers, firm affiliation matters. 1/ Our data permit us to improve on these methods by using a simple, though rarely used, technique. We add to the earnings function dummy variables representing the firms in which the workers in the sample are employed. Whether the dummy coefficients are significantly large is then a direct test of the proposition that firm affiliation matters. This procedure is made possible by the character of the sample. Because the sample was chosen from a limited number of manufacturing establishments, randomly selected in the first stage, respondents are dispersed among a smaller number of firms. The proce- dure is preferable to the conventional method using sectoral averages for two reasons. First, the hypothesis that firm affiliation matters is best tested using data on individual workers and firms. Secondly, the firm dummy variables capture all the influence of firm affiliatioa on wages, and not just that associated with particular characteristics of firms such as profitability and degree of unionization. The lack of significance of variables measuring specific firm characteristics does not exclude the possibility of segmentation among firms owing to other unspecified characteristics. 1/ See, for example, Fields and Marulanda (1976). -9- An earnings function was estimated with the natural logarithm of earnings as the dependent variable. 1/ The coefficients on each of the firm dummy variables are listed in column 3 of Table 1; the base sub-category represents the remaining 23 firms with fewer than 50 employees. Moot of the coefficients (14 out of 24) are significant at least at the 5 percent level, suggesting that, even after controlling for differences among workers in personal characteristics, firm affiliation continues to matter. Indeed, judging by the coefficients, to some workers it matters a great deal. To take the extreme cases, an employee in firm 4 with the same personal characteristics as another in fit i 16 never--heless earns more than double. 2/ Of the 14 firms with significant coefficients, 7 pay at least 15 percent more than the group of small firms which constitute the base dummy, and 3 pay at least 15 percent less. It would be misle.ding on the basis of this evidence to conclude that labor market segmentation among firms has great influence on the structure of wages in Tanzania's manufacturing sector. Consideration of the proportion of total variance in log earnings explained by certain gro-ps of variables puts such segmentation in perspective (Table 2). A worker-s firm is a poor predictor of his wages: in'a ranking of independent variables according tt their predic- tive power, the firm dummies would be placed near the bottom. When added to 1/ In this, and in all other regression equations reported below, estimates were made with both the absolute wage and 'ts natural logarithm as the dependent variable. In no case were the results substantively affected by the choice of dependent variable; only the logarithmic results are presented in the tables. 2/ Because a dummy variable is dichotomous, its coefficient in a semi- logarithmic equation understates the percentage increase in the dependent variable (see Halvorsen and Palmquist, 1980). - 10 - the personal variables, the firm dummies raise the explanatory power of the equation by only 3 percentage points (raising R2 from .63 to .66) 1/; on their own they produce an R2 of .10. The firm dummy variables contribute only 13 percent of the explained variance; the personal variables explain 87 percent, of which those variables more likely to represent productive charac- teristics account for 59 percent and those personal variables more likely to represent discrimination and segmentation among workers for the remaining 28 percent (Table 2). Before we attempt to explain the reasons for labor market segmentation among manufacturing firms, we should recognize that the influence of such segmentation is relatively minor. There are two possible techniques for determining whether the payment of premium wages is associated with particular characteristics of firms such as high profitability, a high proportion of workers unicnized, and foreign ownership. We can simply replace the firm dummy variables in the earnings function with the variables that measure the firm characteristir. Alternatively, we can use the coefficients on the 24 firm dummies, which constitute a hierarchy of firms ordered on the basis of the wages they pay to workers with the same personal characteristics, as the dependent variable in another equation with 24 observations. The characteristics of firms would be the independent variables in such an equation. The latter technique appeals because of the clear separation between firm and individual variables. 2/ But the former technique proved more 1/' evertheless, a hierarchical F test indicates that the contribution of the the firm dummies is significant at the one percent level. 2/ It wras suggested by Rees and Schultz (1970) who considered using, but did not pursue, this approach in their study of the Chicago labor market. Table 2: 1%E RELATIVE CONTRIBUTION OF FILM AFFILIATION TO THE EXPLAINED VARIANCE IN LOG EARNINGS Percentage of total explained variance that is explained by selected groups of variables in: a. Regression with b. Regression with firm dummy firm characteristie variables variables Personal variables 87.3 95.4 of which: economic 58.7 61.2 (education, experience, occupation, formal training) non-economic 28.6 34.2 (sex, race, age, employment status) Firm variables 12.7 4.6 Note: The method cf Epportionment was as follows. We estimated an earnings function using a full set of independent variables; by means of the function we then estimated the predicted earnings (% ) of each employee. The coefficient on each independent variable (i) in turn was set equal to zero, and the predicted earnings (Q ) were estimated using the hypothetical function. The variances of 0 and were calculated, and the percentage contribution of 1. to the explained variance in earnings estimated as (var (0) - var (4 ) )/ (var (0) - -ar (7 ) See Knight and Sabot (1981b). - 12 - helpful in assessing the effect of particular firm characteristics, and its results are reported below. We do, however, put the firm dummies to one further use. When the 24 coefficients of the firm dummies are the dependent variaoles and the ..ha.acteristics of firms the independent variables, the 2- highest value obtained for the corrected R is only 0.38. Similarly, when the variables representing ftrm characteristics are substitut -- for the firm dummy variables, the contribution of firm affiliation is reduced from 13 to 5 percent of the explained variance of log earnings (Table 2). Thus there is a presumption that characteristics of firms other than those we have measured also have an impact on wages. I/ 4. Do Foreign Firms Pay More? The conventional wisdom that foreign firms pay more 2/ requires careful scrutiny. The interesting q, 3tion is not whether they do pay more but w!7ether, if that is true, they do so because they are foreign. A simple comparison of foreign and local firms might well reveal sig- nificant differences in mean earnings. However, it would fail to distinguish between the contribution of multinational status alone and that of other characteristics (such as high technology products, capital intensity, profita- bility, large scale production, or a unionized labor force) which might be associated with multinational status. This can be done, subject to the limitations imposed by multicollinearity, in a multiple regression analysis of the sort we attempt below. But th, problem is more complicated. It is correct to standardize completely for such characteristics as capital iptensity 1/ Indeed, both methods of assessment suggest that the other characteristics explaii just over 60 percent of the contribution of firms to the variance of wages. 2/ See Arrighi (1973), Reuber (1973), Knight (1975), Lim (1977). - 13 - only if capital intensity is causally independent of ownership status. If foreign firms are more capital-intensive wholly because they are foreign, no standardization is required: standardization should logically be performed only on that part of the difference in capital intensity which is independent of ownership status. The relevant but difficult question becomes: What long run changes in the characteristics of a firm would there be as a result of a change in ownership status? The difficulties of answering this question and of unravelling complex inter-relationships is illustrated by the following statement from a recent survey of evidence on choice of techniques in manufacturing in less developed countries: the MNCs are frequently pictured as the special villains of the appropriate-technology effort. They are, so the argument goes, tied to their capital-intensive technology in the developed countries.... And, even if they are considering adaptations, they frequently pay higher wages than do locally owned firms, and they can obtain their capital abroad at cheaper rates, so they would have less incentive to adapt. I/ The former statement implies that higher capital intensity flows from multi- national status; the latter that higher capital intensity flows from higher wages, which apparently flow from multinational status. As we shall see, the problems of collinearity between certain variables and ownership status make causation difficult to esta'lish. 1/ White (1978),-pp. 42-3. He cites as proponents of this view Stewart in Streeten (ed.: (1973), and Streeten (1972). The manufacturing sector of Tanzania does not divide neatly into two ownership categories. Whether the government owns all or part of a firm may have an influence on wage; in it. The government has an opportunity to implement its wage polic:-es in the parastatal enterprises. I/ Foreign participation in some firms falls short of complete ownership. Thus we distinguish five groups of firms: wholly government owned and thus wholly local (A); wholly private and wholly fcreign (J2); wholly private, but paZt local and part foreign (J3); partly government owned and partly foreign (J4); and wholly privately owned and wholly lccal (J5). Initially we ignore the issue of the influence of government participation on wages, and focus our attention on the differences in characteristics and wages between foreign and local private firms. Seventeen of the 24 larger firms fall into these cate- gories (8 foreign and 9 local); their evmployees comprise two thirds of the labor force of the firms for which we have data on firm ch2racteristics. Table 3 shows that employees ir. foreign private firms earn on average 15 percent more than those in local private firms. It also reveals differences in firm characteristics between foreign and local private firms which are consistent with the caricature of multinational enterprises in developing countries. Whether judged by the number of employees or by value added, foreign firms are larger. They are three times as capital intensive, and labor producttvity as measured by value added per employee is dubstantially higher in foreign firms. Were it not for the extraordinarily poor performance 1/ A parastatal organization is defined by thi government of Tanzania as "not an integral part of the gover-iment, but an institution, organization g or agency which is wholly or mainly financid or owned and controlled by the government. The criterion of such nublic Waterprises would be ownership by the government of )0 percent or more of the capital shares, or c'her forms of governmental participation and effective influence in all the main aspects of management of the enterprise." Quoted by Jackson (1978), p.39, from government statements. - 15 - Table 3: CHARACTERISTICS OF FOREIGN AND LOCAL PRIVATE FIRMS: MEAN VALUES OF PERSONAL AND FIRM VARIABLES Variable Foreign Local (J2) (J5) Wages (W) 384 334 Years of Education (E) 4.1 3.2 Years in Current Employment (Ll) 5.1 6.0 Years in Prev'ous Employment (L4) 3.4 3.3 Formal Training* (F4) 0.09 0.06 Supervisory* (01) 0.04 0.03 Clerical* (02) 0.20 0.07 Skilled/Headman* (03) 0.24 0.28 Semi-Skilled* (04) 0.35 0.39 Unskilled* (05) 0.19 0.22 Age 15-19* (Al) 0.05 0.06 Age 20-34* (A2) 0.75 0.69 Age 35-49* (A3) 0.16 0.18 Age 50- * (A4) 0.04 0.03 Regular Employment Status* (R2) 0.97 0.87 African Race* (Tl) 0.88 0.88 Number of Employees (X) 126 91 Value Added (V'000) 1,963 1,007 Value Added per Employee (VE) 15,580 11,060 Capital/Labor Ratio (K) 22,810 7,140 Profit as Percentage of Costs (P) 12 15 Proportion of Workers Unionized (U) 91 86 Proportion of Unskilled and Smiskilled in Total Employment (H2) 0.55 0.60 Number of Observations (N) 213 227 * The mean value of a dummy variable, marked by an asterisk, indicates the proportion of the sample in the particular sub-category. - 16 - of one, loss-making, enterprise, the multinationals would be more profitable than local firms; 1/ as it is, on average those in the 1971 survey are somewhat less so. True to their image of being particularly accommodating to labor organizations in host countries, foreign enterprises have a higher proportion of their labor force unionized. Turning to the characteristics of employees in the two ownership categories of firm, we have reason to believe that foreign firms are more skill-intensive as well as more capital-intensive. Employees in foreign firms have received more formal education, and a higher proportion have received training provided by the firm. White collar workers comprise a substantially higher proportion of the labor force in the foreign firms. On the other hand, workers in multinationals have had, on average, marginally less experience in the firm and less experience in wage employment generally. However, this is more likely to be explained by differences in the age of enterprises or It their rates of growth than by higher labor turnover in foreign enterprises. 2/ It must be stressed that these various differences in group characte- ristics are not necessarily the result of ownership status. The heterogeneity of the manufacturing sector means that such differences may reflect differences in the composition of production, with foreign firms disproport!onatly repre- sented in particular types of manufacturing activity, including motor assembly and chemical products. Do these differences between the two ownership groups in employee and firm characteristics account for the difference in mean earnings? Table 4 1/ Despite the danger that multinationals will use 'transfer prieing' to understate their profits in order to evade taxation or exchange controls in a developing country. 2/ On average the local firms in the sample had begun operations 11 years and the foreign firms 6 years before the survey. - 17 - Table 4: EARNINGS FUNCTION FOR THE LARGE FIRM SAMPLE COMBINING PERSONAL AND FIRM VARIABLES Independent Firm Variable l/ Coefficients a b Wholly government firm 2/ (JI) -0.487 ** Foreign private firms 2/ (J2) 0.058 Part-local, part-foreign private firms 2/ (J3) 0.095 Part-government, part-foreign firms 2/ (J4) 0.167 * Proportion of workers unionized (U) -0.765 * -0.107 Firm size: 80 or more employees 2/ (X2) 0.118 * 0.121** Proportion of workers unskilled or semi-skilled (H2) -0.995 ** -0.791** Profits as a percentage of costs (P) 3.020 0.091** Capital - labor ratio (K) 0.000 0.000 N 3/ 628 628 -2 R 0.675 0.654 S.E. 0.358 0.369 F 62.97 63.41 ** Indicates in this, and in subsequent tables, that the coefficient is significant at the one percent level. * Indicates in this, and in subsequent tables, that the coefficient is significant at the five percent level. I/ The regression was estimated using personal as well as firm variables; but only the coefficientsfof the latter are reported here. 2/ The base sub-category in The dummy variables are private local firms and firm size 50-79 emplc?es. 3/ Although the number of observations, corresponding to the number of employees in the 24 large firms, is 628, the number of values taken by the firta variables is 24. The appropriate tests of significance for the firm characteristics are unclear; the asterisks are based on the normal test. - 18 - - shows that they do so in large part. According to the coefficient on the dummy variable signifying foreign ownership (with local private firms as base), the wage premium paid by foreign firms, is reduced from 15 to only 5 percent; the coefficient is not significantly positive. This conclusion is qualified by consideration of a difference in the structure of wages between private foreign and local firms. The tet premium paid to Asian as opposed to African employees is over a third higher in local than in foreign firms. We attribute this to the high proportion of local firms owned and managed by Asians. 1/ Were it not for this feature of local firms, a wage premium in foreign firms might be discernible. Multinationals have a favorable impact on wage structure in the sense that their-s appears to be economically more rational than that of local employers. This is suggested by Table 5, which presents separate earnings functions for foreign and local private firms. The influence of personal variables which might be interpreted as representing human capital is greater for the former than for the latter group. Conversely, the influence of personal variables which might be interpreted as proxies for a worker-s social station is greater in local firms. The higher premium paid to Asians by local firms is an example of the latter case because in Tanzania Asians were traditionally accorded higher status than Africans. Employers accepting the traditional hierarchy may, through their wage structure, densciously or unconsciously help to sustain it. The returns to experience o;n the job and to previous employment experience are markedly higher in foreign Ehan in local firms, whereas the converse is true of the returns to aging. Given that 1/ Knight and Sabot (1981a). - 19 - Table 5: EARNINGS FUNCTIONS FOR EMPLOYEES OF FOREIGY AND LOCAL PRIVATE FIRMS Coefficients Independent Variables Foreign Local Years of Education (E) 0.043** 0.027** Years of Current Employment (LI) 0.040** 0.010* Years of Previous Employment (L4) 0.022** 0.011** Supervisory (01) 0.517** 0.491** Clerical (02) 0.484** 0.491** Headman/Skilled (03) 0.250** 0.261** Semi-Skilled (04) -0.014 0.138** Age 15-19 (Al) -0.261 -0.379** Age 20-34 (A2) -0.126 -0.224** Age 35-49 (A3) 0.010 -0.004 Casual Employment Status (R2) -0.358** -0.199** African Race (TI) -0.522** -0.814** Constant -20.8 -55.3 2 0.49 0.72 S.E. 0.49 0.30 F 11.01 31.1 Number of Observations (N) 213 227 Note: The base sub-categories of the dummy vIiables are Unskilled Occupation (05), Age 50 - (A4), Regular Employment Status (RI), and Non-African Race (T2). - 20 - years in current and in previous employment are included as independent variables in the regression analysis, it is implausible that the coeffi- cient on age can be interpreted as a return to human capital. In more tradi- tional societies age per se is accorded status. These results may thus reflect a greater tendency among local employers to have their wage structure conform to a traditional social hierarchy. Government is in a better position to influence wages in the para- statal organizations than in the private sector. It is important to discover whether parastatal status raises or lowers earnings, because this will help us to assess the influence of government wage policy. If parastatals do pay homogeneous labor more than other firms, then a government policy of curbing parastatal pay might be justified on both distributive justice and allocative efficiency grounds. However, if the difference in mean wages is due to differences in labor force composition, then the distributional benefits are more arbitrary, and the allocative effects harmful insofar as parastatals constrained from paying high wages may be unable to attract or retain the more productive workers they want to employ. Parastatal organizations are represented in our sample by 4 partly government owned and partly foreign firms (J4). Another form of partnership, 2 private firms with part local and part foreign ownership (J3), is also analy7ed. The premium paid by foreign firms appears to vary with the precise ownership arrangement. Table 4 indicates that foreign firms which enter into partnership with local firms pay 9 percent more than local private firms (the premium is not statistically significant), whereas a partnership between foreign firms and the government is associated with particularly high premiums. The coefficient on J4 is significant and indicates a premium of 19 percent in relation to the base sub-category, J5. - 21 - How is this remarkable result to be explained? Does it mean that the influence of government has been to raise wages rather than to lower them? The most plausible explanation is suggested by the pattern of nationalization in Tanzania. The parastatals are more capital-intensive, larger and more profitable than private firms. The values of K, Q and P for the J4 group (41,475, 7:0 --ind 85) are far higher than those for J2 and J5 shown in Table 3. These .haracteristics - capital-intensity, size and profitability - may well be the reason for the form of ownership rather than being the result or simply a coincidence, i.e. those firms were the first targets for state participation. Prior to their affiliation with government, such firms may have pa4d particularly large premiums because their economic characteris- tics and forzign status made them both able to pay and vulnerable. This interpretation .ild be favored by Jackson: ... rationalization measures were aimed largely at the larger capital-intensive multinational corporations who mostly tended to pay above-average wa,es. 1/ Consistent rith this argument is the fact that the one firm in the sample which was, rnm ir- inception, wholly government-owned (Jl) paid wages signi- ficantly below w:ges in the group of private local firms, as revealed by its dummy coefficient in the multiple regression analysis. We have provided an explanatin of the high wage premium paid by parastatals in 1971 which does not atetibute the premium to government policy. Indeed, there is reason to beiieve that government policy after 1967 was to reducp pay, i.e. that the parastatals were subject to a more stringent wage policy than was the rest of the economy. 2/ This was effected 1/ Jackson (1979). 2/ There is an alternative explanation for the large premium on the J4 variable. Large foreign firms may not have been persistently high payers. Rather, the managements of these firms may have raised pay in anticipation of the parastatal wage policy. - 22 - through the Standing Committee on Parastatal Organizations (SCOPO). SCOPO was established by the President in 1967 to examine pay in parastatal organiza- t1ons. A year later it issued a directive which prescribed common basic pay scales for all parastatal employees. I/ These were in many cases below the pay of persons then in post. The scales were not revised between 1968 and the year of our survey. 2/ However, in 1971 many parastatal employees were still receiving higher pay than the scales indicated because individuals were permitted to retain their pre-SCOPO pay levels, albeit without increments, on a personal basis. Our finding of pure ownership status effects suggests that the government policy of attempting to erode the premium paid by para- statals was well advised on grounds of allocative efficiency as well as equity. 5. Do Unions Drive Up Wages? In the model of wage determination based on countervailing power, trade unions are expected to be more successful in raising the wages of their members, the greater is their strength and aggressiveness. This in part will depend on certain economic variables. For instance, unions are likely to be more successful in raising wages, the lower is the employer's elasticity of demand for labor (and so the smaller are the adverse employment effects) and the smaller is the relatfve size of the unionized sector (so facilitating a : 3 9 relative gain). Factors such as these will determine the size of the coeffi- cient on an independent variable indicating unionization in an earnings function. 1/ SCOPO (1968). 2/ Nor indeed until 1974. - 23 - The view that trade unions are generally able to raise real wages in less developed countries is not held unanimously. Unions may be ill- organized and lack financial or political power, and be weakened by the abundance of unemployed and poor workers keen to step into wage jobs. As a result, unions need not have the impact that is apparent from their actions and statements. I/ In particular, where government clashes with the unions and subordinates them, they may cease to play an effective role in wage determination. If they become instruments of goiernment policy, and that policy is one of wage restraint, unions may even be effective in depressing wages. The role of trade unions in Tanzania appears to be better depicted by the 'government control' than by the 'countervailing poxwer' model. 2/ The government adopted a series of measures during the 1960s which effectively subordinated union aims to its own. It was able to obtain the acquiescence of workers and their organizations in a policy which severely restricted the rights to strike and to engage in free collective bargaining. In return it offered them new and substantial benefits in the form of greater security, and participation. In 1962 the Trades Disputes (Settlement) Act was passed which prohibited strikes and lockouts unless a prescribed set of conditions had been. fulfilled. In 1964 the National Union of Tanganyika Workers Act established NUTA as the sole trade union, dissolving the Tanganyika Federation of Labor and its member unions. NUTA was subsequently affiliated to the Tanganyika African National Union (TANU), the ruling political party, and enjoined -to do 1/ See, for instance, House and Rempel (1976). 2/ For detailed accounts see Friedland (1969) and Jackson (1979). -24 - everything in ts power to promote the policies of TANU.' In 1967 the govern- ment adopted a- incomes policy limiting increases in average wages within an establishment tc a maximum of 5 percent, except under extraordinary circum- stances. In the same year it established " Permanent Labour Tribunal to mediate in all labor disputes, and it effectively barred all strikes and lockouts. Compensating benefits for these restrictions and regulations include severance allowances, introduced in 1962; the establishment in 1964 of the National Provident Fund providing pensions for retired workers or their widows, and for disabled workers; passage of the Security of Employment Act, also in 1964, which provided for the establishment of Workers' Committees in all enterprises employing at least 10 NUTA members, and placed restrictions on the summary dismissal and fining of employees; and the establishment of Workers' Councils in 1970, and of Managemcnt Commit:ees in 1976, which increased workers' participation in m:.iagerial decisions. The success of the government attempt to transform the role of unions in Tanzania is re- flected in the trends in strike activity. Table 6 shows that the number of strikes, workers involved, and total working days lost rose throughout the 1950s; in 1960, the year before Independence, roughly one quarter of all wage employees were involved. Strike activity steadily declined in tne 1960s, and after 1968 very few strikes took place The .countervailing nower' model of the role of trade unions in the wage determination process is clear in its preliction of the sign of the coefficient on the unionization variable: it will be positive. However, what prediction is implied by the 'government control' model, which seems to be more appropriate in the Tanzanian casa? The coefficient could be - 25 - Table 6: STRIKE ACTIVITY IN TANZANIA 1955-1973 Year Number of Number of Workers Number of Working Strikes 1/ Involved Days Lost 1955 42 8,877 12,562 1956 54 17,695 58,066 1957 114 39,786 165,328 1958 153 67,430 296,746 1959 205 82,878 402,693 1960 203 89,495 1,494,773 1961 101 20,159 113,254 1962 152 48,434 417,474 1963 85 27,207 77,195 1964 24 3,582 5,855 1965 13 884 1,825 1966 16 2,062 8,845 1967 25 3,224 7,224 1968 13 1,906 5,757 1969 4 874 2,141 1970 3 357 726 1971 3 654 3,026 1972 0 0 0 1973 0 0 0 1/ Excluding strikes lasting less than one day. Source: Jackson (1979), Table 1. - 26 - either zero or negative, dependiig on whether the unions are passive or actively depress wages. If wages are determined competitively or by the statutory minimum in the non-unionized sector, and if the competitive or statutory wage is similarly effective in the unionized sector, unions should be irrelevant to wages. However, if NUTA attempts to hold down wages against worker demands or employer initiatives, and is better able to do so in more unionized firms, unionization may actually reduce pay in a firm. Various difficulties arise ir attempting, by means of the sign and size of the coefficient on the unionization variable, to test these competing arguments. Consider the hypothesis that unions raise wages. Insofar as the union movement has a general political impact, for instance by influencing government minimum wage legislation, this effect will not be reflected in the coefficient. If non-unionized firms choose to raise the pay of their employees in order to avoid the threat of unionization, the coefficient will again be an understatement. On the other hand, it is also possible that the coefficient will exaggerate the influence of unions. Employers in unionized firms, faced with having to pay relatively high wages, are in a position to 'skim the cream' by selecting the most productive from the pool of available workers. Insofar as they are able to select and retain the more able workers, the coefficient on unionization may be in part a proxy for productivity not-captured by the independent variables represeantiag= economic characteristics. 'This last argument applies also if unions depress - wages: the negative coefficient may partly represent lower productivity. - 27 - A further problem arises in the precise formulation of the unioniza- tion variable. The data permit two proxies for union strength: whether a worker is a member of NUTA (as a personal variable) and the proportion of employees who are members of NUTA (as a firm variable). The choice should depend on whether the union negotiates for its members alone or negotiates t) establish rates of pay in the enterprise as a whole. The latter is a more appropriate assumption for Tanzania. Both proxies for union strength were tried: they gave similar results, and only those for the firm variable are presented below. Unfortunately, even the firm variable -- union members as a pro- portion of all employees in a firm (U) - may be a poor proxy for union strength. There are no non-union firms among the 24 large firms in the sample. The proportion of the employees unionized ranges from the lowest value of 69 percent to the highest of 100 percent. It may be that, while the strength of the union varies among firms, it is not influenced by varia- tion of U over this narrow range. It is surprising, therefore, that the unionization variable produces a decisive result. The coefficient is negecive, as predicted by the 'govern- ment control' hypothesis, and it is significant (regression a in Table 4). The value of the coefficient implies that a rise in the value of U by 10 percentage points would reduce pay by roughly 7 percent. The union movement in Tanzania, it seems, is an instrument of government poJicy rather than the representative of wage-earner interests. The alternative explanation of the sign of the unionization coeffi- cient is that it represents a non-causal association: unionization may simply - 28 - be a proxy for some other firm characteristics wiich are responsible for reducing the wage. That this might be the case is suggested by regression b in Table 4, showing that the coefficient on U remains negative but is no longer significant when the firm ownership variables are dropped from the regression. As Table 8 below shows, unionization is negatively correlated with profits, with the share of unskilled and semi-skilled employees in the total, and with parastatal ownership. It is unlikely that the collinearity between U and J4 can explain the negative coefficient on U, however, because U is positively correlated with J2 and J3, the other ownership categories with positive coefficients. Nevertheless, some doubt must remain about the interpretation of the negative association between unionization and earnings. 6. Does Ability to Pay Raise Wages? For managers of firms on the margin of profitability the will to resist pressures for higher wages is reinforced by the knowledge that the existence of the firm is imperiled. Countervailing power models of wage determination suggest that indicators of ability to pay, such as firm size or capital intensity or, more generally, profitability, are likely to prove good indicatorg-of a firm's rate of pay. But if there is no upward pressure on wages from their worker-, there is no reason why firms able to pay more should do sPunless they would themselves benefit. We should not be surprised, therefore, if the proxies used to measure ability to pay in the earnings function are not significant. Indeed, this proved to be the case for capital intensity (K) and profits as a proportion of total cost (P) - 29 - (Table 4, regression a). However, when the firm ownership dummies are dropped from the equation (regression b), the coefficient on profits as a proportion of costs becomes significantly positive, reflecting the correlation (+0.489) between profits and the highest paying ownership category. This suggests that it is those firms which have both the means and the motive to pay more -- the most profitable foreign firms, which were included in our parastatal category - that do so. The conventional view is that large firms pay more than small firms. Simple supporting evidence, showing mean earnings by size of firm, is not hard to find for rich and poor countries. 1/ Few econometric attempts have been made to isolate the effect of firm size by standardizing for other variables. One such study for Kenya found that earnings tended to rise with firm size for the education levels 0-11 years of schooling, but not for those with more than 11 years of schooling. 2/ It is of interest, therefore, to discover whether firm size has an effect in our sample. However, we should not be surprised if the results are weak. The reasons for expecting a posi- tive association between firm size and earnings apply within a sub-sector and only incidentally to the manufacturing sector as a whole. The indicator of firm size chosen was the number of employees. 3/ When firm size was entered as a continuous variable (X),-the coefficient was negligible and insignificant. A search for a non-linear relationship using9dummy variables suggested a distinction between thi e firms witn 50-79 employees and those with 80 or more (X2): wages in the larger firms 1/ Phelps Brown (1977), pp. 275-6, Taira (1966), p. 285. 2/ Thias and Carnoy (1972, Tables 2.10, 3.11, C.1. 3/ Gross output and value added in continuous form were also tried as indicators, but the results were the same as for X. - 30 - are 13 percent higher (Table 4). Further dit-agregation within tie above- 80 category failed to reveal successive wage increments wi h incr'asing firm size. It is possible that the occupational structure of employment in a firm influences the level and structure of wages paid by the firm. The greater the proportion which a particular type of labor constitutes of total employment, the more costly to the firm is a wage increase for those workers. We therefore hypothesize that the greater the proportion of unskilled and semi-skilled employees in the total employment of a firm, the lower the wage it pays them. If wage pushfulness were equally strong at the higher occupational levels, it would also be the case that the greater the proportion of unskilled and semi-skilled employees -- and thus the smaller the proportion of skilled and white collar employees -- the higher the wage paid in skilled and white collar occupations, and hence the wider the occupational wage structure. However, these predictions flow from the countervailing power model of wage determination, and if it is suspect, so also are the predictions. To test them we estimated separate earnings functions for each occupational group, and included among the firm variables either unskilled employees as a proportion of total employees (HI) or unskilled and semi- skilled employees as a proportion of total employees (H2). Table 7 presents the results. The evidence is consistent with the prediction that a relatively large base to the occupational pyra*Ld depresses wages at the base. The coefficient on HI is negative, but not significant for unskilled earnings, and the coefficient on H2 is significantly negative for unskilled plus semi- skilled earnings. The evidence is not consistent, however, with the other prediction that a relatively large base to the occupational pyramid raises - 31 - wages at the top. The coefficients on Hi and H2 are negative and significant for all occupations combined, and negative for each occupation separately. Moreover, the numerical size of the negative coefficient increases with skill level. As the proportion of unskilled and semi-skilled employees increases, the occupational structure of wages is compressed. We can only speculate on the explanation. The differences among firms in I and H2 may reflect differences in products and production pro- cesses. For instance, a high proportion of skilled and supervisory workers (low Hi and H2) might be associated with high productivity, and hence high pay, of such workers. Examples would be engineering workshops, motor repair and other non-standard skill-intensive products. In these establishments, which employ a high proportion of craft artisans, the low values of HI and H2 may be a proxy for the skill level of the skilled. The sample of manufac- turing firms is too small and diverse for the satisfactory analysis of industrial sub-categories. However, the 7 firms in our sample of 24 which appear to fall into this category have lower values of HI and H2 (17 and 45 percent respectively) than the 24 firms as a whole (26 and 60 percent). If this interpretation is correct, then the ability to pay argument may still apply to relatively unskilled wages. Years ago Sumner Slichter found that the unskilled wage in the U.S. tended to be high in industries where the semifskilled and skilled wage was also high, and where wage costs were low and net Profits high in relation to sales. 1/ He attributed this result to the g infltfnce of managerial policy as opposed to market forces. Correspondingly, it is ,ossible that in Tanzania a high proportion of unskilled and semi- I/ Slichter (1950). - 32 - skilleI4 labor causes firms to keep these wages low, and that internal labor market considerations then account for the low wages found also at the higher occupational levels. The variables Hl and H2 bring out a general problem in trying to estimate the effect on earnings of any one fira characteristic, such as occupational structure. This is shown by the size of the correlation coeffi- cients between the independent variables in Table 8. For instance, 12 is collinear with X (positive), with the highest paying firm category, J4 (positive), and with U (negative). Size of firm (X) is correlated positively with the high paying firm categories (J3 and J4) and negatively with U. The collinearity is not only due to the small number of firms in the sample: some clustering of firm characteristics is likely to be endogenous. Although our results on the influence of particular firm characteristics must be qualified for these reasons, they reveal a coherent story consistent with the institutional setting in Tanzania. - 33 - Table 7: COEFFICIENTS ON OCCUPATIONAL STRUCTURE VARIABLES IN EARNINGS FUNCTIONS STRATIFIED BY OCCUPATION I/ Occupation Numbers Coefficient on HI Coefficient on H2 White Collar 102 - 1.51 * - 1.58 ** Skilled 155 - 0.64 * - 1.02 ** Semi-skilled 231 - 0.47 - Unskilled 172 - 0.10 - Unskilled and Semi-skilled 403 - 0.79 ** All Occunations 660 - 0.49 ** - 0.96 ** 1/ The data relate to all employees of the 24 larger firms. - 34 - Table 8: FIRM VARIABLE CORRELATION COEFFICIENTS J2 J3 J4 U X H2 P J2 1.000 -0.193 -0.382 0.391 -0.343 -0.207 -0.205 J3 1.000 -0.160 0.266 0.619 0.171 -0.082 J4 1.000 -0.499 0.485 0.206 0.489 U 1.000 -0.288 -0.391 -0.227 x 1.000 0.501 -0.069 H2 1.000 0.048 P 1.000 - 35 - 7. Conclusions The methodological contribution of this paper is that, by combining personal and firm characteristics, we have been able to make an advance both on earnings function studies which include only personal characteristics and on studies of firm wage determination which use only firm characteristics and the average characteristics of employees in the firm. We have been able to test hypotheses concerning the role of the firm in wage determination while standardizing for personal characteristics. Our examination of the relative importance of the personal and the firm characteristics indicates that the personal variables are the predo- minant influence on earnings. When firm variables were introduced into the earnings function, they improved the explanatory power of the equation but only to a minor extent. The dummy variables representing individual firms could explain about 13 percent of the variance in explained earnings, i.e., their contribution to the overall inequality of pay was very limited. Never- theless, the coefficients on most of the firm dummy variables were statis- ftically significant, and, when these were replaced by firm characteristics, some of the characteristics had significant coefficieftts. The wage of a worker is indeed affected by the firm in which he is employed. Although the evidence badically supports the competitive labor market paradigm, the market is by no means perfectly compe_-Ltive. IMwas possible to test the hypothesis that foreign firms pay more by comparing the coefficients on the dummy variables representing owner- ship status: foreign firms pay only 6 percent more than locally owned private firms. However, we found evidence that foreign firms wage structure is - 36 - economically more rational and conforms less to social hierarchy than that of local employers. Given that the government policy after 1967 was to restrain wages, we expected the coefficient on the dummy variable repre- senting parastatal enterprises to be negative: on the contrary, it was positive and larger even than that for foreign companies. The most plausible explanation is suggested by the pattern of nationalization in Tanzania. The parastatals were previously the most profitable, most capital-intensive and largest foreign firms, which had paid the highest wages. Nationalization had been fairly recent, and the implementation of the government policy of curbing pay would take time. The finding on parastatals is therefore support for, and not refutation of, the view that foreign firms pay more if they are sufficiently profitable. The role of trade unions in Tanzania appears to be better depicted by the 'government control than by the 'counterprevailing power model. The government adopted a series of measures during the 1960s which effectively subordinated union aims to its own. In the multivariate analysis, unioniza- tion turned out to depress wages. The sign on the firm unionization variable was negative, and the coefficient significant. If there is no upward pressure on wage from workers, there is no reason why firms able to -ay more should do so unless they would themselves benefit. evertheless, various indicators of ability to pay were included as explanatq*y vari,.6les in the earnings function. Although profitability and capital-intensity proved to have insig- nificant coefficients, there was some evidence in favor of the ability to pay hypothesis. First, there was a limited tendency for earnings to increase with - 37 - siz:± of firm; secondly, the higher the proportion which the unskilled and semi-skilled constituted of total employment in a firm, the lower their earnings; thirdly, the correlation between profitability and parastatal status suggests that employers with both the means and the motive pay more. Some of the results on particular firm characteristics run counter to the conventional wisdom. Nevertheless, due consideration of the institu- tional background in Tanzania has enabled us to provide explanations for them. - 38 - REFERENCES Arrighi, Giovanni (1973). International corporations, labor aristocracies and economic development in Tropical Africa,' in G. Arrighi and J.S. Saul, Essays on the Political Econowy of Africa, Monthly Review Press, New York. Fields, Gary S. and Nohra de Marulanda (1976). Intersectoral wage structure in Colombia, Yale University Economic Growth Center, Discussion Paper No. 251. Friedland, William H. (1969). Vuta Kamba: The Development of Trade Unions in Tanganyika, Hoover Institution Press, Stanford. Halvorsen, Robert and Raymond Palmquist (1980). 'The interpretation of dummy variables in semilogarithmic equations, American Economic Review, Vol. 70, No. 3, June. House, W.J. and H. Rempel (1976). 'The impact of unionizaticn on negotiated wages in the manufacturing sector in Kenya, Oxford Bull-tin of Economics and Statistics, Vol. 38, No. 2, May. Jackson, Dudley (1979). 'The disappearance of strikes in Tanzania: incomes policy and industrial democracy,' Journal of Modern African Studies, Vol. 17, No. 2, June.. Knight, J.B. (1975). Wages in Africa: what should a foreign firm do?', Oxford Bulletin of Economics and Statistics, Vol. 37, No. 2, May. Knight, J.B. and R.H. Sabot (1981a). 'Labor market discrimination in a poor urban economy', World Bank, Washington D.C., mimeo. Knight, J.B. and R.H. Sabot (1981b) 'Why wages differ: earnings iunction analysis in a poor urban economy', World Bank, Washing.on D.C., mimeo. Lim, David (1977). 'Do foreign companies pay higher wages than their local counterparts in Malaysian manufacturing?', Journal o' Development Economics, Vol. 4, No. 1, March. Phelps Brown, Henry (1977). The Inequality of Pay, Oxford University Press, Oxford. Rees, A. and G.P. Schultz (1970). Workers and Wages in an Urban Labor Market, University of Chicago Press, Chicago. Reuber, Grant L. (1973). Private Foreign Investment in Development, Oxford University Press, Oxford. - 39 - Slichter, Sumner (1950). 'Notes on the structure of wages-, Review of Economics and Statistics, Vol. 32, No. 1, February. Standing Committee on Parastatal Organizations (1968). Directive No. 3: Salary Scales in the Parastatal Sector, Dar es Salaam, August. Stewart, Frances (1973). 'Trade and Technology' in P. Streeten (ed.), Trade Strategies for Development, John Wiley, New York. Streeten, Paul (1q72). 'The Multinational corporation and the nation-state, in P. Streettr, The Frontiers of Development Studies, John Wiley, New York. Taira, Koji (1966). -Wage differentials in developing countries: a survey of findings,' International Labor Review, Vol. 93, No. 3, March. Thias, H.H. and M. Carnoy (1972). Cost-Benefit Analysis in Education: A Case Study of.Kenya, World Bank Staff Occasicnal Papers, No. 14, Baltimore. White, Lawrence J. (1978). -The evidence on appropriate factor proportions for manufacturing in less developed countries: a survey,- Economic Development and Cultural Change, Vol. 27, No. 1, October. POPULATION & HUMAN RESOURCES DIVISION DISCUSSION PAPERS [A * by the number indicates the paper is Bank-confidential or otherwise restricted.] 81-1 Oey A. Meesook, "Demographic Characteristics of Individuals and the Measurement of Lifetime Income". February 1981. 81-2 Susan H. Cochrane, "The Economics of Fertility with Examples from Asia". February 1981. 81-3 Rashid Faruqee, "Social Infrastructure and Services in Zimbabwe" February 1981. 81-4* Richard Sabot, et al. , "Cognitive Skills: Their Determinants and Influence on Earnings in Two Poor Urban Economies". February 1981. 81-5 Dean T. Jamison, et al., "Improving Elementary Mathematics Education in Nicaragua: An Experimental Study of the Impact of Textbooks and Radio on Achievement". March 1981. 81-6 Rashid Faruqee, "Analyzing the Impact of Health Services: Narangwal and Other Experiences". March 1981. 81-7 J. B. Knight and R. H. Sabot, "Labor Market Discrimination in a Poor Urban Economy". March 1981. 81-8 Oey A. Meesook, "Interrelationships between Demographic Factors and Income Distribution: Problems of Measurement, Description and Interpretation." April 1981. 81-9 Susan H. Cochrane, "Fertility Attitudes and Behavior in the Nepal Terai." April 1981. 81-10 J.B. Knight and Richard Sabot, "The Role of the Firm in Wage Determination: An African Case Study." April 1981.