WPS4688 Policy ReseaRch WoRking PaPeR 4688 Are Skills Rewarded in Sub-Saharan Africa? Determinants of Wages and Productivity in the Manufacturing Sector Louise Fox Ana Maria Oviedo The World Bank Africa Region Poverty Reduction and Economic Management Department August 2008 Policy ReseaRch WoRking PaPeR 4688 Abstract Using recent matched employer-employee data from Third, workers exercise some power in the wage setting the manufacturing sector in 20 Sub-Saharan African process but their influence varies by linguistic group. countries, the authors analyze how the supply of skills In common-law countries, union presence benefits all and legal origin of the country affect the wage setting workers equally, not just members, whereas in civil- process. The wage analysis yields three main findings. law countries, only older members enjoy higher wages. First, increasing returns to education, especially for older The authors also contrast wage premia with relative workers, suggest that the expansion of education in marginal productivities for different age, occupation, and Africa has reduced returns to education for entrants in education categories. The findings show that in general, the labor market. Second, age effects matter not just for older, highly educated, and highly ranked workers receive returns to education, but also for the wage setting process wage premia that do not reflect a higher relative marginal more generally. In particular, in civil-law countries, productivity. returns to seniority are rewarded only after a certain age. This paper--a product of the Poverty Reduction and Economic Management Department, Africa Region--is part of a larger effort in the department to analyze the relationships between labor market policy, economic growth, and poverty reduction in Sub-Saharan Africa. Policy Research Working Papers are also posted on the Web at http://econ.worldbank. org. The authors may be contacted at lfox@worldbank.org, and aoviedo@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Are Skills Rewarded in Sub-Saharan Africa? Determinants of Wages and Productivity in the Manufacturing Sector* Louise Fox and Ana Maria Oviedo The World Bank JEL classification codes: J21, J24, J31, O15, O49. Keywords: Sub-Saharan Africa, employer-employee data, wage determination, unions, productivity, human capital. *We are grateful to Vijaya Ramachandran, Manju Shah, John Haltiwanger, Alan Gelb, Benn Eifert, Rita Almeida, John Giles, Fabiano Bastos, Melissa Gaal, Mary Hallward-Driemeier, and participants at the Labor Market Group seminar for their helpful comments, suggestions, and data. The views expressed herein are those of the authors and do not necessarily represent those of the World Bank. E-mail: lfox@worldbank.org, aoviedo@worldbank.org. 1 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? I. Introduction The goal of this paper is to provide an update on the state of skills, wages, and productivity in the manufacturing sector in Sub-Saharan Africa (SSA). Although employment in the manufacturing sector has been growing at a steady pace in the last years, manufacturing still accounts for less than 10 percent of overall value added in SSA.1 As Table 1 shows, the share of manufacturing in GDP between 2002 and 2006 in 17 countries does not exceed 14 percent. (An exception is Swaziland, where it reaches 37 percent.) The contribution of manufacturing to total employment is also very small, accounting for less than 10 percent on average.2 INSERT TABLE 1 One reason cited for the slow development of the manufacturing sector is the low education level of the African workforce. A number of studies have highlighted the importance of education to economic growth, and the dampening effect on growth of the failure to expand education in the colonial period.3 African countries attained independence with a huge education deficit relative to the rest of the world. This massive skill shortage hurt African economies in every sector, public and private (see Figure 1). Figure 1 Literates as % of Adult Population 90 80 70 60 50 40 30 20 10 0 Sub-Saharan Africa Eastern Asia/Oceania Latin America/Caribbean Southern Asia 1980 1990 Source: SPESSA (1999) Since the 1960s, African countries, jointly with donors and other partners, have made a huge effort to expand access to education, with astonishing levels of success. While the workforce in SSA 1We exclude South Africa from our analysis. 2For example, the share of industry employment in total employment is about 7 percent in Madagascar and Uganda (World Bank, 2007a). 3See Bosworth and Collins (2003), and Ndulu et al. (2007). 2 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? countries still lags behind the rest of the world in average levels of education, we find that in some countries the average education level of manufacturing workers is as good as it was in Asian countries 20 years ago. Not all countries have followed the same patterns in expanding access. In a number of countries - primarily former British colonies- the "universal primary education" (UPE) model was adopted beginning in the mid-1990s. In this model, a large push was made towards access for all, but often with a quality decline in the initial period, at least in terms of inputs per student.4 Other countries have focused on expanding both primary and secondary school while trying to maintain quality. In this paper we focus attention on how the expansion of education has affected the relation between skills and labor market outcomes, specifically wages in the manufacturing sector. We are also interested in how skills with positive returns for wages relate to productivity at the firm level. There is an ample literature on returns to education in Africa, especially using household survey data from the 1980s and early 1990s. Appleton et al (1996) review some of them, citing 14 studies that analyze the returns to education for wage workers. Most used basic "Mincerian" specifications, although a few had additional information measuring other, usually unobservable, characteristics. All found positive and significant returns to education, but in contrast to what is observed in OECD countries, most researchers find that rates of return curves are still convex in Africa, especially for higher education, implying that skill shortages persist at this level. The cost of expanding higher education, combined with the brain drain (which has, in effect, made the market for skilled professions in some areas global) imply that this skill shortage is likely to persist for some time. Most authors have also found that returns to education are lower for women (see Table 2), and a few studies also find evidence of age effects (see Söderbom et al., 2006). INSERT TABLE 2 Recent analyses of the determinants of wages using repeated cross section data also suggest that returns to education are declining for new generations, especially for primary education (see Fox and Gaal, 2008). This is not surprising given how fast the labor force has been growing compared with the growth of physical capital. In addition, as education systems move from elite systems to broad ones, the selectivity factor associated with education (especially post-primary education) decreases.5 Finally, there is no doubt that in some countries, enrollment expanded faster than the supply of trained teachers and other inputs, lowering quality. 4See Avenstrup (2006) 5In other words, returns to education were over-estimated for older cohorts as education was correlated with other, unobservable, individual characteristics. 3 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? But wage setting might not solely depend on worker's ability. Indeed, the empirical literature has found ample evidence of a strong link between firm characteristics and worker wages (e.g., Krueger and Summers, 1988; Blanchflower et al., 1996, Abowd et al., 1999, Söderbom et al., 2004), suggesting that observable proxies of worker ability do not explain the entire wage setting process. Several recent studies use 1990s manufacturing data from the Regional Program for Enterprise Development (RPED) to analyze wages in African manufacturing, finding evidence of rent-sharing (e.g., Teal, 1996; Azam and Ris, 2001; Alby, 2003, 2007; Söderbom and Teal, 2001; Söderbom et al., 2004), efficiency wages (Azam and Lesueur, 1997; Söderbom et al., 2004; Mazumdar and Mazaheri, 2002), inefficient labor management (Fafchamps and Söderbom, 2006), job sorting (Benhassine et al, 2006), as well as a role of unions in wage setting (Manda, et al., 2001; Alby, 2007), which is another form of rent- sharing.6 Interestingly, the evidence also suggests that the unexplained gap between earnings and productivity tends to narrow in relatively more advanced countries (Van Biesebroeck, 2007). Figure 2 Doing Business Rankings General rank 103.7 158.8 Entry regulation rank 116.2 128.5 Licensing regulation rank 111.4 140.0 Labor regulation rank 83.4 131.3 Land registration rank 123.5 134.0 Access to credit rank 109.7 139.4 Investor Protection rank 87.2 122.1 Taxation regulation rank 71.0 133.2 Trade regulation rank 114.8 133.2 Contract enforcement rank 74.2 141.8 Bankruptcy regulation rank 73.5 140.3 Common Law (Anglophone) Civil Law (Francophone) Source: Doing Business. Higher ranks denote more stringent regulations. The worst score is 178. Labor regulations, in particular hiring and firing costs might also play a role in the process of upgrading the labor force to add more skilled labor. Increasing evidence in the literature suggests that the legal origin of a country has a significant impact on several measures of institutional quality, 6For a complete review of papers that use this data, see Bigsten and Söderbom (2006). 4 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? including regulation, which ultimately have an effect on development.7 In Africa, labor and other regulations vary substantially according to the legal origin of the country; particularly between common-law and civil-law systems (see Figure 2). Our paper adds to the literature on skills and labor market outcomes in SSA in several ways: first, we use recent matched employer-employee data covering 20 countries in Sub-Saharan Africa. To our knowledge, this is the first study to include such a large number of countries and to use the new Enterprise Surveys collected between 2003 and 2007.8 Our sample enables us to analyze common patterns in wage-setting mechanisms in manufacturing across much of the continent, with enough confidence that our findings are relevant at the aggregate level. Indeed, the combined manufacturing output in our sample countries equals 44 percent of the total manufacturing output in Sub-Saharan Africa.9 Second, we split countries according to their legal origin (and thus the source of their Western language use) to analyze wage determination, and in particular how the expansion of the educated labor force has affected returns to education. To this end, we analyze separately across age groups (and legal origins) how skills, captured in different ways, affect wages. In this sense, our paper complements a recent study by Alby (2007), who also looks at differences in wage setting according to legal origin, using Enterprise Survey data for 9 countries. Consistent with the literature, he finds evidence of rent-sharing, and he also finds that unionization has varying affects across both country groups (a result we find as well). However, Alby's study does not look at how returns to education vary with a worker's age, something for which we do find strong evidence. Finally, we investigate how differences in skills (associated with differences in wages) translate into differences in productivity. Following Hellerstein et al. (1999) and Van Biesebroeck (2007) we jointly estimate wage and productivity premia for the pooled sample and for each legal origin group. Our results suggest that productivity premia are much lower than their corresponding wage premia, which is consistent with previous findings by Bigsten et al. (2000). The rest of the paper is organized as follows. Section II describes the data. Section III describes institutional differences across countries, as well as differences in firm and worker characteristics in our sample. In particular, we look at differences in skills across common law (Anglophone) and civil law (Francophone) countries. In Section IV we present results of our wage regressions, in Section V we present results of our wage and productivity premia estimations, and we conclude in Section VI. 7See La Porta et al. (1998), Heckman and Pagés (2000), Beck and Levine (2004), and Loayza et al. (2005). 8The data collected under the RPED project covered the early to mid- 1990s. Thereafter, all firm surveys are done under the global Enterprise Surveys project. See www.enterprisesurveys.org. 9 Excluding South Africa. 5 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? II. Data We use surveys of manufacturing firms carried out by the world-wide Enterprise Survey project of the World Bank group between 2003 and 2007.10 The firms surveyed are for the most part registered for tax purposes; they employ at least 5 employees.11 The manager/owner of the firm is interviewed in detail about basic firm characteristics (age, legal status, etc.), as well as specific investment climate questions, for instance whether the firm has experienced power outages, what the delays have been when it requested a public service, how difficult it is to deal with public officers, etc. In addition, the manager is asked to rank the investment climate obstacles according to his/her opinion (infrastructure, access to finance, labor regulations, crime, corruption, etc., a total of 15 investment climate aspects). In short, the surveys contain a rich set of questions that can be used to relate firm performance to the investment climate of the country. Our sample covers 12 civil law countries (Benin, Burkina Faso, Burundi, Cameroon, Cape Verde, Democratic Rep. Congo, Guinea, Madagascar, Mali, Mauritania, Rwanda, and Senegal), all of them Francophone, except Cape Verde; and 8 common law countries (all Anglophone): Botswana, Gambia, Ghana, Kenya, Namibia, Swaziland, Tanzania, and Uganda. Within manufacturing, our sample has firms in the following sectors: food & beverages; chemicals, paints & pharmaceuticals; construction materials; metals; paper & printing; plastics; textiles, garments & leather; wood; and other manufacturing.12 A number of surveys in Africa include a worker module for a sub-sample of randomly selected firms in manufacturing. A maximum of 10 workers were interviewed in each firm.13 They were asked about their current and starting occupation, wages, tenure, experience, gender, and other relevant questions for Africa (health, transportation, etc.). In most cases, the sample is not random, since the goal was to interview at least one worker in each job category (manager, professional, skilled production, unskilled production, non-production).14 We use this information, plus the data in the firm surveys, to analyze employment growth and the determinants of wages in African manufacturing firms. Our total number of observations is roughly 10 Comprehensive information about these surveys is available at www.enterprisesurveys.org. 11 In some countries, separate surveys have been implemented for micro- and informal firms, but we are not using these data at present. 12 Hereafter, we refer to common law countries as "Anglophone" and civil law countries as "Francophone". 13 Most firms have at least 6 workers in their sample. 14 In the surveys conducted in the last 3 years, worker sampling is done as follows. In a random sample of manufacturing firms, only after the manager agrees, (a) if a list of workers exists, the interviewer(s) obtains the list of workers and randomly selects the predetermined number of employees; (b) if the list does not exists - that is, in most of the cases- the interviewer(s) walks in the factory and randomly selects workers in all categories (professional, skilled, unskilled, manager, etc.) ensuring that all categories are represented. (Personal communication with Giuseppe Iarossi, who oversees data collection in Africa.) 6 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? 17,000 workers in 3,800 firms.15 We exclude foreign workers (a very small fraction of the total sample) as their wage patterns are markedly different from national workers. III. Institutional differences and manufacturing characteristics across Sub-Saharan Africa Institutions The fast recent growth in manufacturing employment observed in our sample is remarkable achievement, especially given the constraints faced by firms.16 For instance, when firms are asked to express the degree of obstacle to their business posed by the lack of infrastructure (electricity, water, telecommunications), between 80 and 95 percent of firms qualify it as a major or severe obstacle.17 Comparing more "objective" data on business regulations, such as the Doing Business indicators, also places SSA at the bottom the ranking (see Table 3). These rankings alone have led many observers to predict a dim future for the sub-continent. INSERT TABLE 3 But as Figure 2 above shows, within SSA there are important differences, particularly between Anglophone (common law) and Francophone (civil law) countries. Notably, in virtually all categories of the Doing Business indicators (entry, licensing, registration, labor, contracts, trade, taxation, access to finance, investor protection, bankruptcy), the Anglophone group's ranking stands ahead of the Francophone group's. For instance, in the labor regulation category, the Anglophone group has better scores in hiring, rigidity of hours, and has lower non-wage labor costs. Furthermore, even though the actual firing costs (in terms of weeks of wages) are higher in Anglophone countries, the overall firing index is lower (that is, firing is less difficult legally). Firms Tables 4 and 5 provide information about the firms in our sample. Despite the similar industry composition of our sample across both legal groups, we do find a few differences in other firm characteristics: 21 percent of firms in the Francophone area are partly (or entirely) foreign owned, which is the case only for 13 percent in the Anglophone area. Thus, despite an apparently weaker investment climate, Francophone countries attract more foreign investment. INSERT TABLE 4 15The number of observations on the wage regressions fall to approximately 10,000 due to missing values. See Table A15. 16For a detailed discussion of employment growth in manufacturing and the role of regulations, see Fox and Oviedo (2008). 17See Fox and Oviedo (2008). 7 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 5: Selected firm characteristics Workers Exporter Privatized Foreign Part. foreign Provides reported for (percent) (percent) owned owned training (percent) (percent) (percent) taxes (percent) Anglophone 25 9 11 13 33 49.2 Francophone 25 3 19 24 29 66.9 Source: Enterprise Surveys But overall, the within group differences are much greater than those between groups: the share of exporting firms varies from 4 percent in Cape Verde to 59 percent in Kenya. Likewise, the share of firms that have at least some foreign ownership ranges from 7 percent in Ghana to 50 percent in Botswana. If we look at the size distribution of firms, we find similar patterns to other countries. On average over 50 percent of firms are small (less than 20 employees), but there is a large variance in size across countries.18 18This is a much smaller share than would be found in most business registers or censuses in African countries. Here the composition of the sample matters particularly: since it is restricted to formal firms, large firms tend to be overrepresented. 8 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Workers African countries have made major strides in increasing educational attainment. Since 1990, the annual change in primary enrollment has been well above the rate of population growth ­ nearly 5 percent between 1998 and 2002. Secondary enrollment has increased even faster but from a lower base (see Figure 3). Figure 3 Gross secondary enrollment ratio 100.0 90.0 80.0 70.0 tn 60.0 rceeP 50.0 40.0 30.0 20.0 10.0 0.0 Latin America & East Asia & Pacific South Asia Sub-Saharan Africa Caribbean Source: WDI (2007) 2000 2005 As a result, the supply of graduates with completed primary and some secondary education is increasing every year. Table 6 reports that countries such as Cameroon and Namibia have been able to shift their education distribution so that about 40 percent of their labor force has 10-12 years of education. Countries such as Uganda, who started later owning to a long civil war, were still able by 2002 to provide 70 percent of the labor force with at least 6-9 years of education, and another 8 percent with 10 and above. Unfortunately, rivals in South and East Asia have been able to increase education levels faster, in part because they have lower population growth. The educational distribution of the entire labor force (at the turn of the century) in Africa is quite similar to that of Asian countries 20 or 30 years earlier.19 INSERT TABLE 6 19See Psacharopoulos and Arriagada (1992). 9 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Nonetheless, manufacturing firms in Africa have been able to attract a well educated labor force, allowing upgrade in skill levels since the 1990s. Table 7 shows that on average, workers in our sample had 11 years of education; 12 years in firms with over 100 employees. This compares favorably with China, where workers in manufacturing have on average 9.8 years of education, and Morocco, where only 27 percent have more than 12 years of education.20 Even unskilled workers have on average over 8 years of education, which roughly corresponds to having completed a junior secondary education.21 Similarly, average education (years) for professional workers corresponds nearly to a university degree, and there is relatively little variation across countries within this category. Firms in both country groups have older workers with slightly less education. Larger firms and exporters tend to have slightly more educated workers. INSERT TABLE 7 We tried to compare the levels of education from the previous wave of enterprise surveys but the sample of countries differs, and there were also changes in the sampling methodology. For the countries where we do have data in both waves, the numbers are shown in Table 8. In Ghana and Kenya the share of workers with post-secondary education seems to have increased significantly.22 Likewise, the average number of years of education increased in both countries. INSERT TABLE 8 A word of caution is in order here: although the worker sampling has been made as randomly as possible, it is still likely that some worker categories are being overrepresented in this sample, particularly high-skilled workers, because of the stratification technique. While there is no precise way to establish this (nor to construct appropriate weights), we can still roughly compare our sample averages with a few responses from the firm's survey, which give us an indication of the "true" averages at the firm level. We can do this in particular for the distribution of education (albeit only for those firms interviewed before 2006) and for the distribution of occupations (only for the categories of skilled and unskilled production workers). INSERT TABLES 9 AND 10 20The source of education in China is the National Bureau of Statistics, One Percent Population Sample (2005). We're grateful to John Giles for providing us with this number. 21Readers should exercise caution in interpreting cross country comparisons of education years, because they do not necessarily reflect similar levels of skills due to the variation of education quality within and across countries. In this paper we do not correct for education quality, although in our regional regressions we argue that one potential reason for differences in returns to education is quality. 22However, numbers for Kenya must be interpreted with caution as differences might be largely due to sample composition (oversampling of large establishments). 10 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 9 shows the distribution of the workforce at the firm level according to years of education, and the distribution implied by the worker sample. We observe a systematic shift of the distribution in the workers' sample, with far higher shares of workers that have ore than 12 years of education than what the firm reports as the average. This is particularly the case for Kenya and Uganda. In Table 10 the bias towards skilled workers seems less evident, although more comprehensive distributions (only available in Kenya, Mali, and Senegal) suggest that managers and professionals are somewhat overrepresented in the workers' sample. However this bias is not as marked as the one for education. Tables 11 and 12 show union membership rates across countries and for different age brackets. The rates of unionization are not very high, which seems at odds with the picture of labor market rigidity from Doing Business discussed above. However, several studies of the African labor market point out that unions have by-and-large lost ground since the 1980s.23 INSERT TABLES 11 AND 12 Figure 4 Tenure years by union membership and worker age 16.0 14.0 ) rs 12.0 eay( 10.0 reu enteg 8.0 6.0 eravA 4.0 2.0 0.0 <20 [20-29] [30-39] [40-49] [50-59] 60+ Worker age Source: Enterprise Surveys Non-union Union There is substantial variation across countries, however, with very high rates in East and Southern Africa, and lower rates in Francophone countries (with the highest rate observed in Cape Verde). Unionization increases with worker age and tenure, which may be one of the reasons for the high severance pay costs in Anglophone countries (see Figure 4). Older firms have a more unionized labor 23See Mazumdar and Mazaheri (2002), Fox and Gaal (2008). However, this does not imply that union members receive equal treatment as compared to non-members. See Section VI. 11 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? force, and they are also more likely to have been privatized. We will return to this issue in the analysis below, on the role of unions in wage setting processes. IV. Determinants of wages We now turn to the analysis of the determinants of wages. In particular, we are interested in testing several hypotheses of labor market frictions that potentially affect the wage determination process. In a perfectly competitive labor market, workers' wages equal their marginal productivity, and provided that there are sufficient outside options and that labor supply is unlimited at any productivity level, wages should strictly reflect a worker's ability, that is, firm characteristics or worker characteristics unrelated to ability should not be significantly associated to wages. In practice, the labor market is highly imperfect, since the supply of qualified workers is limited, and there are relatively few job opportunities for unskilled workers. More able managers might also want to attract more able workers, hence offering them a premium above their productivity.24 In addition, legal restrictions on hiring and firing, minimum wages, unions, and other societal arrangements might interfere with the wage-setting process. In our estimations we concentrate on measuring how wages are affected by workers' age, education, experience, occupation in the firm, bargaining power, and whether the firm is likely to pay efficiency wages to attract more able workers. We use in turn a pooled sample of all workers across countries, and segmented samples for common-law (Anglophone) and civil-law (Franco- and Lusophone) countries. First, we construct a worker's monthly wages using the worker questionnaire, which contains two questions regarding current wages and allowances, namely the unit of payment (hourly, weekly, monthly, etc.) and the amount in current local currency units. We compute the monthly wage by multiplying the hourly rate obtained from the questionnaire, by the number of hours worked per week (where available) times four.25 In order to compare wages across workers and countries, we converted the nominal wages to PPP- US dollars by using the exchange rate corresponding to the year prior to the survey and the 2005 PPPs from the International Comparison Group (2008).26 Table 13 shows annual wages by country, 24See Abowd et al. (1999) a detailed study of this phenomenon in France. 25If the hours worked were not available we assumed the employee works 40 hours on average. In many cases there were inconsistencies in the answers of workers. In particular, many workers reported a payment unit different than monthly (say, hourly), but reported the amount corresponding to the payment frequency (usually monthly). This created an unusual number of outliers, so nominal wages were subsequently corrected by taking into account education, occupation, and payment frequency. This reduced the number of outliers considerably, however, it is still possible that some measurement error remains. If this error is randomly distributed, it should not affect our results. 26Using the official exchange rate only is misleading as it does not take into account difference in the cost of living across countries. In order to reduce measurement errors, we also tried normalizing the nominal monthly wage by taking the ratio of the monthly wage to the median wage of the unskilled production worker category 12 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? and Table 14 shows monthly wages by country and occupation. Table 13 displays the mean, median, and number of observations for both workers' annual wages as well as average annual wages calculated at the firm level.27 The systematic discrepancies between the workers' and the firms' averages confirm the fact that our worker sample is likely to over-represent highly-skilled workers and those that have highly-paid occupations. INSERT TABLES 13-14 Except for the wages of managers, median wages in Francophone countries are higher than in Anglophone countries. There are several possible explanations for this observation: (i) living costs may be higher in Francophone cities (even after the PPP correction). However, in this case, we would expect even the managers' median wages to be higher. (ii) It is possible that the market for professionals is an international one, while the other markets are local, although the variation of the median wage across countries goes against this hypothesis. Rather, because the category "managers" includes both employed managers as well as owners, it is likely that owners are actually paid less on average than some of their more qualified employees, which can drive the average manager wage down if the sample has relatively more small firms. (iii) There may be a better supply of educated workers in Anglophone countries. This explanation, (and to some extent, the cost of living), are both plausible, given that Rwanda and Madagascar have both much higher median wages than the others for all occupations.28 (iv) It is possible that unions push up the salaries of the lowest paid workers in Francophone countries (although the unionization rate is lower in Francophone countries). We test the hypotheses regarding workers' education, occupation, and the union wage effect in our regression model, including also a few firm attributes that we believe may capture efficiency wages. First, we estimate a basic Mincerian earnings model, which we then expand to try to account for unobserved ability and other individual characteristics which, if excluded, are believed to bias the OLS regression (see Appleton et al., 1996). Thus, we estimate the following model: lnWij = + 1(Ei )+ 2(Ii )+ 3(Oi )+ 4 Fj + 5(Ui )+ij ( ) (1) where lnWij is the log of the monthly wage or employee i in firm j, Ei is education in years of employee i, Ii is a vector of other characteristics of employee i used in the standard Mincerian (within country). This specification worked well and the regression results are qualitatively similar, but we lost some of variation at the lower end, so we decided to use the PPP-exchange rate. 27In the firm questionnaire the manager reports the annual cost of labor. This number is divided by the number of employees (permanent and temporary, correcting for the average duration of a temporary contract). It is then also corrected using PPP exchange rates. 28We tried removing Rwanda from the Francophone sample given that it seems to be an outlier, but the median wages were basically unaffected ­ they were still significantly higher in the Francophone area. 13 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? formulation, (age, tenure, experience, gender, whether the worker is full-time), Oi is occupation of employee i, where O is a dummy for the following categories: manager, professional, skilled production, unskilled production, and non-production. This variable potentially captures unobservable characteristics of the worker which allow them to perform a particular occupation.29 Fj is a vector of firm characteristics, namely size and exporter status. There is evidence in the literature that large firms pay higher wages, which is typically attributed to efficiency wages paid to more able and dedicated workers (Söderbom et al., 2004). The same is true for exporters, for whom the quality of the product is more important than for non-exporters. Ui is a dummy that indicates whether the worker belongs to a union, while U indicates whether there are any unionized workers f in the firm.30 As Alby (2007) finds in his analysis, rent-sharing can take different forms and benefit workers differently, hence, we might find union effects on wages related to a worker belonging to a union, or a general wage effect related to the presence of a union at the firm. Finally, ij is an error term that is assumed to be uncorrelated across firms, but not necessarily within firms. This means that in our estimations we cluster the errors at the firm level, which typically yields larger standard errors for the estimates. This model assumes that the effects of the independent variables on wages are consistent across cohorts. However, as suggested above, given the rapid expansion in both the size of the labor force and in access to education, this may not be true. In addition, there might be other reasons for wages to differ with age, for instance if unions lobby on behalf of older workers, or if workers' promotions are correlated with age.31 To test these hypotheses, we introduce a dummy variable for workers above the median age (32 years old). All our regressions also control for worker effort (proxied by hours worked), as well as other unobserved effects at the country and industry level. 29 Benhassine et al., (2006) argue that African firms are able to detect these characteristics and sort workers into occupations accordingly. 30 As Table 16 shows, around 43 percent of firms have at least one unionized worker in Anglophone countries, versus 25 percent in Francophone. 31 In fact, unionization is positively correlated with age (0.15). Having received at least one promotion at the firm is also positively correlated with age (0.08). 14 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Results Table 15 reports the results of the regressions with the pooled sample. Following Murphy and Welch (1990) we add quartic terms for tenure and previous experience in order to capture more accurately the age-earnings profiles.32 INSERT TABLE 15 Column (1) reports the results from the basic specification. In terms of individual variables, we have a few deviations from the standard Mincerian. We are able to subdivide experience into that acquired at the firm, and that acquired before the employee joined the firm (but in the same industry). The former is expected to be worth more than the latter, as during tenure at the firm, firm-specific as well as general human capital can be built up. In contrast to evidence from the US and other developed economies, returns to age, experience, and tenure seem to be linear in our case, since the quartic terms are insignificant (with the exception of age, which is negative but very close to zero). The linear returns to tenure and experience are equal to 0.014 and 0.016 respectively, well below the ones reported for the US, of around 0.03 (see, for instance Topel, 1991; and Altonji and Williams, 2005). The age dummy is positive and highly significant, indicating that older workers receive about 14 percent higher wages, for equal levels of education and experience. In the basic specification we include the number of years of education as a linear term and squared, and we find that both terms are positive and significant at the 1 percent level, suggesting increasing returns to education. In fact, everything else equal, for a worker with six years of education each extra year of education is associated with a wage increase of 4.4 percent; for a worker who is at the mean of education years (10.8) the increase is 5.4 percent, and for a worker with 16 years (the equivalent of a university degree) the increase of an additional year is 6.4 percent. In column (2), we add dummy variables indicating the level of educational achievement of the worker (i.e., the highest degree obtained).33 These dummies are expected to pick up the marginal effect of obtaining a certificate of completion for full secondary, vocational training, or university, and they also capture increasing returns to education.34 As expected, education still has a positive coefficient, however, it is much lower than in the previous regression, and not statistically significant. The squared term is still significantly different from zero, but it only represents a semi-elasticity of 0.1 percent. On the other hand, the coefficients for each degree obtained are large and significant, 32 As we see in the results, these turn out to have insignificant coefficients. The same is true for quadratic terms. 33 The categories are: Primary degree or no education, Junior secondary or secondary degree, Vocational training degree, and Tertiary or university degree. 34 Additionally, they might also capture signaling effects. 15 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? suggesting that employers place a much higher value on having completed an education.35 In fact, compared to those with only primary or no education, having completed a secondary education raises the wage by 16 percent, while workers with vocational training receive on average 35 percent higher wages. Workers with a tertiary or university degree have an 81 percent higher wage. We take this a strong evidence of increasing returns to education, which is consistent, for instance, with findings for Kenya and Tanzania by Söderbom et al. (2006). Experience appears to be worth more than years of education (not counting the degree obtained), with an additional year of previous experience or tenure adding a two percent increase in wages. Surprisingly, years of prior experience are worth the same as tenure years in this specification, but this shifts slightly after we add other variables (although statistically they are not different). With respect to other characteristics, female workers earn about 5 percent less, and this characteristic stays negative at this level through the analysis. Temporary workers earn between 25 and 24 percent less than permanent workers, which is as expected. In our first two specifications, we only augment the Mincerian model with firm size and whether the firm is an exporter. As expected, we find that exporter firms pay higher wages, about 14 percent higher. On the other hand, we only find weak evidence of a firm size effect, since the coefficient for medium firms is not significant. Large firms appear to pay between 9 and 12 percent more than small firms. It could be argued that our results are sensitive to sample composition, in the sense that some countries will carry a disproportionately large weight in the regression when in fact the size of their economy is small in comparison to the rest. We test this by performing the same regression analysis as previously, but we weigh each observation by the GDP of the corresponding country. Results are reported in column (3) of Table 15. The coefficients related to education and age are maintained, whereas those of gender, tenure, and experience fall somewhat and lose their significance, as does the firm size effect. The exporter effect is maintained. Although results are not radically different whether or not we weigh the observations, in all following exercises we choose to weigh the observations by GDP. In column (4) of Table 15, we add the variables for occupation, which appear to pick up a significant fraction of unobserved ability, as the net returns to educational achievement decline. Linear returns to education are about the same, but returns to all degrees fall, especially vocational and secondary, which fall by nearly 50 percent. Returns to prior experience also fall below the returns to years of tenure. The skills and characteristics found in managers and professional workers seem to be 35It is also likely that more able workers usually have completed their education. 16 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? especially valuable to the firm, which pays a premium of 85 percent to managers and 56 percent to professionals (relative to unskilled workers), even after controlling for education and experience. In this same regression we introduce the two variables aimed at capturing rent-sharing from the workers' side. We include a dummy for workers who belong to a trade union, and a dummy for workers who a relatives to the owner or manager of the firm. We find no significant wage premium for union members, and surprisingly, relatives of the owner or manager seem to earn about 12 percent less. Finally, in column (5) we add a firm-level variable aimed at capturing rent-sharing at the firm level.36 Specifically, we include a dummy for firms in which at least one worker belongs to a union, to indicate union presence at the firm. This could have an effect not just on members' wages, but on the entire wage bill. We find a positive but insignificant union-firm effect of about 7 percent. Institutional effects Broadly speaking, the regression models estimated so far with the pooled sample show a well behaved wage function ­ the signs of the coefficients were as expected, and consistent. Now we analyze the wage function separately for the two legal systems in our sample: common- and civil-law (i.e., corresponding to Anglophone and Franco/Lusophone countries). Recall that the overall regulatory environment is much less friendly in Francophone countries, but the hiring and firing costs are actually higher in some Anglophone countries than in the mean of Francophone countries. What do these institutional differences mean for returns to education, occupational premia, and rent- sharing? INSERT TABLE 16 Table 16 reports the results obtained for the specifications in columns (1), (3), and (4) of Table 15, estimated separately for Francophone and Anglophone countries. The differences are striking. Returns to years of education appear to be higher in Anglophone than in Francophone countries, however in both regressions the coefficient is insignificant. Returns to educational achievement on the other hand, are large and significant in both groups, and slightly larger in Anglophone countries, especially at the secondary and tertiary levels. (Vocational degree-holders earn roughly the same premium of 31-37 percent in both.) University or tertiary degree holders earn 93 percent more than uneducated workers in Anglophone countries, whereas they get about 72 percent more in Francophone countries. Secondary degree holders earn about 24 percent more than uneducated workers, but the premium for a secondary degree in Francophone countries is only 11 percent. 36 We also omit the quartic terms from previous specifications, as they were all insignificant. This does not alter our results. 17 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? In terms of experience, we find that after controlling for education only previous experience has an effect on wages in Anglophone countries, but not tenure; while an additional year of tenure increases wages by 2.1 percent, and an additional year of previous experience increases wages by 1.2 percent in Francophone countries. We also find that there is no significant wage penalty for being a woman in Anglophone Africa, whereas females in Francophone countries earn about 8 percent less. As for firm effects, the exporter premium is still high and significant in both groups. On the other hand, the firm size effect is only present in Francophone countries, and its significance is stronger for medium-sized firms, which suggests the presence of efficiency wages in these countries. In columns (2) and (3) we augment the specification by adding worker occupation, union membership (again at the worker and firm level), and relation to the owner or manager. As in the pooled results, the coefficients for educational achievement fall considerably after these additions, especially in the Anglophone group. They remain high and significant, however. Results on experience change somewhat, with tenure becoming positive and significant for Anglophone countries, and higher than previous experience. In the Francophone both tenure and previous experience coefficients fall slightly, but remain significant at the 1 percent level. Occupation categories have positive and significant coefficients in both groups, but the magnitudes of the coefficients are much larger in the Anglophone group. Indeed, a manager in an Anglophone country earns over 100 percent more than an unskilled worker, whereas in a Francophone country a manager earns about 53 percent more. Differences in other categories are not as large, but still significant, for instance, the semi-elasticity for professional workers is almost twice as large in Anglophone than in Francophone countries (0.73 vs. 0.41). If promotions at the firm were mostly determined by seniority, then presumably experience would not affect wages. However, the fact that tenure and experience still matter, especially in Francophone countries, suggests that there is job sorting, that is, workers are not randomly assigned to occupations and then rewarded only according to their observable skills (education, experience, etc.), but instead they are assigned to occupations that match their abilities independently of observable skills. Turning to rent-sharing evidence, we find that union membership brings about a 15 percent higher wage in Francophone countries, but has no effect in Anglophone. This is a somewhat unexpected result, because union density is lower in Francophone countries than in Anglophone. Does this mean unions are ineffective in Anglophone countries? On the contrary, we find that the presence of a union does benefit all workers regardless of membership in Anglophone countries, since these workers earn roughly 16 percent more than workers of firms where there is no union presence. It is important to keep in mind that there could be an endogeneity problem with this result, if unionized 18 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? workers are more productive, or if firms that have union presence have in general a more productive workforce. We will explore productivity issues in the next section. As for relatives of the owner/manager working at the firm, we find that they earn about 15 percent less than non-related workers, but only in Francophone countries. 37 Firm effects are maintained similar to those of the baseline specification, that is, firm size matters in Francophone countries, but not in Anglophone countries, while exporters pay higher wages in both groups. Age effects So far, results for all regressions, pooled and segmented, show a positive and significant effect of worker's age, even after controlling for education, experience, and occupation. We now explore whether returns to education, experience, occupation, and other characteristics differ according to age groups. Our motivation for this exercise is to find out whether the rapid expansion in education has changed returns to education (i.e., lowered them), so that firms would be essentially overpaying older workers. In addition, we want to know whether age plays a role in any other effects that we have already observed at the mean sample level. To investigate age effects, we divide workers into "young" and "old" according to the median age of 32 years.38 Then, we run the same specification as in column (5) of Table 15, but separately for each age group. That is, we allow all coefficients to vary with respect to the age group. Even if we might lose accuracy in the estimation du to the reduction in sample size, we still have enough observations in each case to have reasonably precise estimates. INSERT TABLE 17 Table 17 reports the results from this exercise. The first thing to notice is that even if the sample size is considerably reduced in each case, the R squared remains high, revealing that our specification explains between 30 and 60 percent of the overall variation in wages, which is similar to the values reported in Tables 15-16. Interestingly, our specification seems to be able to explain much more of the variation in the Francophone sample. This new set of results confirms by and large that age effects matter, and that they have different effects in each of our groups. We find that returns to educational attainment are significant only for 37Alby (2007) finds a similar difference in the union effect between civil law and common law countries. He hypothesizes that this result is a response to lower regulation in the Anglophone countries. 38This is the overall median, but the median age is about the same for both linguistic groups. In previously reported exercises, we also divided workers into several cohorts, however results suggested that there was a single cutoff, corresponding to "young" workers and the rest. In addition, we did a piece-wise linear spline on age to see if there were level and slope differences. Results are similar to those from Tables 15-16. 19 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? older workers in Anglophone countries, whereas in Francophone countries they are significant (except for secondary education) and have similar magnitudes for both age groups. As before, returns to years of education are only positive and significant at the 10 percent level for older workers in Francophone countries. Also similar to the last exercise, returns to experience and tenure do not seem to affect wages in Anglophone countries, while in Francophone countries they do increase wages, and these returns seem to be slightly higher for younger workers. Returns to occupation in Anglophone countries do not vary across age groups, but they do in Francophone countries. In the latter group, the magnitudes of the coefficients are much higher and significant for the older group.39 Finally, our results regarding bargaining power seem to confirm the initial results of Tables 15-16 that union membership matters in Francophone countries, and union presence matters in Anglophone countries. However, in this case we also find that union membership, and also being related to the owner, matter more for older workers in the Francophone group. This is more in line with our hypothesis stated earlier, that unions tend to protect older workers more. In contrast, union presence seems to benefit all workers equally in Anglophone countries (and being related to the owner makes no difference). In summary, our wage analysis reveals (i) increasing returns to education in all countries, but for older workers only in Anglophone countries, suggesting that the expansion of education in Anglophone Africa has reduced returns to education for entrants in the labor market; (ii) age effects matter not just for returns to education, but they affect the wage setting process more generally. In particular, in Francophone countries returns to occupation are rewarded only after a certain age; (iii) workers exercise some power in the wage setting process, but this influence varies by linguistic group: in Anglophone countries union presence benefit all workers equally, not just members, whereas in Francophone countries only older members enjoy higher wages. V. Are wage and productivity premia equal in African manufacturing? The evidence that certain categories of workers systematically get higher wages suggest underlying differences in worker ability, due to education, experience, as well as unobserved characteristics. While certain categories, such as education, are very likely to reflect productivity differences among workers, in other categories it is less certain that better paid workers are indeed more productive. This is a relevant question if, for example, we conclude that union members enjoy a greater 39There is some potential for small sample bias given that relatively fewer young workers hold high occupations. However, over 65 percent of the Francophone sample is less than 32 years old, and their levels of education are higher, so we believe the potential for this bias is small. 20 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? bargaining position in wage determination, so that they get paid more regardless of their productivity. But it could also be the case that union members are more productive, and therefore they are simply getting rewarded according to their productivity and not their union status. In this section we take advantage of our data to analyze productivity differences across worker categories, and to link these differences to differences in wages. Our methodology draws largely from Hellerstein et al. (1999), and Van Biesebroeck (2007), who use employer-employee data to analyze wage and productivity premia respectively in the United States, and in three African countries (Kenya, Tanzania, and Zimbabwe). In this respect, our analysis in closer to Van Biesebroeck's, with two major differences: first, we use more recent and comprehensive data. Second, we pool the countries' data together in order to increase our sample size and hence the precision of our estimations. We perform two exercises: one with the pooled data, and another where we split the sample according to the linguistic group (English or French), as we did in previous analyses. In essence, this exercise consists of estimating a production function at the firm level, but allowing the labor input to vary according to several "quality" characteristics of the worker. In parallel, a wage equation is estimated that includes the same quality characteristics. The coefficients from the production function and wage regressions are then compared in order to determine whether wage premia actually reflect higher productivity for a particular quality attribute. To illustrate this, consider the following example. Suppose that workers correspond to one of two categories, that we call "0" and "1", and let the production function take the form ln Y = ln A + ln K + ln L~ + (2) K L where L is quality-adjusted total labor, which equals L = L + 1 -1 L1; L1 is the total number of ~ ~ ( ) employees of category "1", 1 is the relative marginal productivity of workers of category "1", and L is the total number of employees. That is, workers who belong to category "1" are more productive than category "0" workers by a fraction 1 -1 .40 Then we can express (2) as: ( ) lnY = ln A+K ln K +L lnL1+ (1 -1) LL + 1 (3) = ln A+K lnK +L lnL+L ln1+ 1 -1)L~1 + ( ( ) 40Expressing marginal productivities in this form facilitates the convergence of the estimation algorithm. 21 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? where L=L0+L1 and L1 = ~ L1 . L Likewise, for the worker, wages are determined by the augmented "Mincerian" specification: K lnW = w0 + k X k + w1I + (4) k=1 where I=1 if the worker belongs to category "1", and zero otherwise, and Xk is a worker characteristic independent of the worker category. From equations (3) and (4) it is readily seen that the productivity premium of workers that belong to category "1" is equal to 1 -1 , while their wage premium corresponds to the semi-elasticity w1. ( ) The goal of the exercise is then to test whether the productivity premium is equal to the wage premium, that is, whether workers who get paid more are really more productive. This amounts to testing whether w1 = 1 -1 . ( ) This exercise is straightforward if we consider only two categories (e.g. male vs. female). However, considering more categories (for example, gender and race) implies that the new number of total worker categories needs to include combinations of both main categories (black females, black males, etc.). In our exercise we restrict the analysis to age, occupation (as before, divided into managers, professionals, skilled production, unskilled production, and non-production), educational attainment (none or primary, junior or senior secondary, vocational, and upper education), and union membership. Because of data limitations, we cannot obtain accurate estimates the proportion of workers of all characteristic combinations at the firm level. We can, however, use only the estimated proportions of each broad category, provided we make the following two assumptions.41 First, we restrict the relative marginal productivities of two worker occupations (managers and professionals, for instance) within one age group (e.g., 32 years old and younger) to be equal to the relative marginal productivities of the same two worker occupations within the other age group. The second assumption is that the ratio of two worker categories within a group (for instance, the ratio of managers and professionals within the young age group) is the same for the same categories in another group (managers and professionals in the old age group). These two assumptions allow us to simplify our expression of quality adjusted labor as follows: 41This follows Hellerstein et al. (1999) 22 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? L~ = [L +(O -1)O]×(M -1)M +(P -1)P +(S -1)SL L L (5) ×(E -1)E +(V -1)VL +(T -1)TL×(U -1)U L L where O is the number of workers older than 32 years old; M, P, and S are respectively the numbers of managers, professionals, and skilled-production workers; E, V, and T are respectively the numbers of workers with secondary, vocational, or upper education (tertiary and university) degrees; and U is the number of union members. In principle, we do have a sub-sample of workers for each of these categories at the firm level. Indeed, in both Hellerstein et al. and Van Biesebroeck's studies the shares of workers for each category are calculated at the firm level. However, because in our sample certain categories of workers are overrepresented, particularly those with high rank and education, our results could be sensitive to this sample bias. We therefore perform two sets of exercises, the first with proportions of each category calculated at the firm level, and the second with proportions at the country, industry, size, and firm age level.42 The dependent variable in the production function estimation is the log of value added (sales minus materials) converted to US PPP dollars. To measure the capital stock, we use the log of replacement value of machinery and equipment stocks, also converted to US PPP dollars. In the wage estimation, the dependent variable is the monthly wage converted to US PPP dollars (also used in the previous section). The estimation of equation (2) and the wage specification presented in equation (1) (and corresponding to column (4) of Table 15) is done jointly, with productivity estimated at the firm level, and wages at the worker level. All errors are clustered at the firm level. Because equation (2) is not linear in its parameters, the joint estimation is done by non-linear maximum likelihood. All estimations control for country, industry, and size effects, they also include a dummy for exporters, a dummy for union presence at the firm, and the average hours worked per week.43 In addition, to mimic our preferred specification in the wage determination analysis, in the wage equation we control for gender, permanent status, years of education, years of education squared, tenure, previous experience, worker age, and whether the worker is related to the owner or manager. 42Within each country, there are 10 industries (Food & beverages, Chemicals, paints & pharmaceuticals, Construction materials, Metals, Paper & printing, Plastics, Textiles, Garments & leather, Wood, Other manufacturing), three size classes (less than employees, 20 to 99 employees, and 100 more), and 5 firm age categories (less than 5 years old, 5 to 9 years old, 10 to 19 years old, 20 to 39 years old, and 40 years old and older). 43In the first exercise (Table 19) hours worked per week is the average of the workers' sample at the firm level. In the second exercise (Tables 29 and 30) we use the mean for the corresponding country, industry, size, and age class. 23 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 19 presents the results for the first exercise, for the pooled sample as well as the Anglophone and Francophone groups. To lighten the presentation of the table, we only report the coefficients pertaining to occupation, educational attainment, age, and union membership. Before we move on to the results' discussion, it is worth pointing out that in all our specifications returns to scale appear to be decreasing.44 This stands in contrast to Van Biesebroeck's results, who finds that in Kenya, Tanzania, and Zimbabwe returns are slightly increasing. However, it is not clear what measure of capital is used in his analysis. INSERT TABLE 19 The first three columns of the Table present the results of the pooled estimation. Recall that the productivity premium is equal to one minus the reported coefficient. The occupation coefficients are all greater than one in the pooled estimation, however, only the professionals' coefficient is significantly higher than one (but barely, as the p-value equals 0.08), as the tests at the bottom of the Table show. This suggests that for an average firm, having a larger share of professionals is associated with a significant increase in output, but this is not the case for having more managers or more skilled production workers. Moreover, the joint test of all occupation coefficients equal to one cannot be rejected. Because in the wage equation managers, professionals, and skilled production workers all receive higher wages, we could conclude that all three categories are overpaid with respect to their productivity. However, the hypothesis of equality of wage and productivity premia cannot be rejected for any occupation, which indicates that in fact workers are not paid above their relative marginal productivity. The same is true for educational attainment. Although all coefficients are higher than one, none of them is significantly so, as the Wald tests confirm, including the joint test, and the hypothesis of equal premia cannot be rejected at any education level (jointly as well as individually). Having older workers seems to hurt production, however the coefficient for the share of workers above 32 is not significantly less than one. For union membership, we obtain similar results, that is, the productivity coefficient is not significantly different from one, and the equality of premia cannot be ruled out. In summary, for the pooled sample our results suggest that the range of wage premia that workers in a given category receive is not statistically different from the range of marginal productivities of workers in the same category. 44 The coefficient for capital varies between 0.03 and 0.05, and the coefficient for labor is between 0.76 and 0.79. We also attempted at conducting this exercise separately for each industry, however the coefficients were too imprecisely estimated due to lack of a sufficiently large sample in most industries. However, if we only estimate equation (3) by non-linear least squares, we notice that most industries exhibit close to constant returns, with the exception of Food & beverages, Furniture, and Paper & printing, for which returns were slightly increasing. 24 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Results are slightly different when we split the sample into Anglophone and Francophone. In particular, vocational degree holders appear to be more productive relative to those with no education, whereas those with upper level education appear as less productive. Both coefficients are significantly different from one. In both cases, the hypothesis of equality of premia is rejected, which indicates that vocational degree holders are paid below their marginal productivity, while more educated workers are paid above. The same is true for managers, who get a sizeable wage premium, but whose productivity is not above the average. In Francophone countries there are no significant differences with respect to the pooled sample. To mitigate the potential problems caused by sample selection bias, we perform two additional exercises, but this time, instead of using the worker shares calculated at the firm level, we use shares calculated for the corresponding country, industry, size, and firm age level. First, we include all worker categories as was done in Table 19. In the second exercise, we take into account only two groups at a time, so that our first estimation includes age and occupation, the second has age and educational attainment, and the third has age and union membership. Results are presented in Tables 20 and 21. INSERT TABLES 20-21 Broadly, our results from the previous table are confirmed. In most categories productivity premia are barely higher than one, although not statistically different from one, with a few exceptions. Having older workers does seem to affect production in this specification, since the coefficient for the share of workers above 32 is significantly less than one. Another category that appears to have lower productivity is upper-education degree holders, in particular for the Anglophone group. On the other hand, vocational degree holders are more productive in Anglophone countries. These results were also apparent in Table 19. Further, in the pooled sample the hypothesis of equality of wage and productivity premia is rejected only for managers and older workers, which indicates that only in these two categories workers are paid above their relative marginal productivity. In the Anglophone group we find that managers, skilled production workers, and upper-education degree holders have wage premia that are significantly higher than their relative marginal productivities, whereas vocational degree holders' wage premium appears to be below their relative marginal productivity. Finally in the Francophone group we find that older workers, union members, and vocational degree holders appear to have wage premia above their relative marginal productivity. These results are largely confirmed when we estimate the equations separately for two groups at a time, as shown in Table 21. In summary, our results suggest that in general, highly educated and highly ranked workers receive wage premia that 25 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? do not reflect a higher relative marginal productivity.45 These results are consistent with previous findings by Bigsten et al. (2000), and by Van Biesebroeck (2007) for the least developed countries in his sample. However they contradict evidence for Ghana provided by Verner (1999), who estimates wage and production function equations and finds positive productivity premia for education. In contrast to our specification, Verner estimates linear equations for production function and wages. This amounts, if our specification from equation (3) is the correct one, to an identification problem for the productivity premium, which is indistinguishable from the marginal product of labor. VI. Concluding remarks Enhancing competitiveness in value-added sectors is a high priority for policy agendas in many African countries. Although our knowledge about the functioning of the non-agricultural private sector is expanding fast, we still know little about how institutions are affecting the efficiency of this sector. In this paper we investigate labor market outcomes, in particular wages, in the manufacturing sector in 20 SSA countries, specifically looking at (a) skill shortage and cohort effects, and (b) effects of regulatory environment and unions. We looked for them in countries which have less regulation but more unionization, and have expanded education faster (Anglophone countries with a common law system) and in countries which have more regulation but less unionization and have expanded education more slowly (Franco- and Lusophone countries with a civil law system). In both countries, we find evidence of skill shortages, especially with respect to higher education. We also find cohort effects on returns to educational attainment in Anglophone countries, with older cohorts getting higher returns. In Francophone countries we find that there are strong occupation premia, but only for older workers, whereas they are uniform across age groups in Anglophone countries (but still very significant in size). Interestingly, our productivity estimation results suggest that neither older workers, nor those with higher qualifications (by degree or occupation) are significantly more productive than younger ones, which points to wage premia above the relative marginal productivity of these groups. There are different reasons why older workers could be getting higher wages. One possible reason is high severance payments. The only way firms can afford to keep older workers is to depress the earnings of younger workers.46 In this case, the premium is of concern, as it could decrease the private returns to education such that parents would no longer be willing to pay the cost of secondary school. 45In a separate set of (unreported) exercises, we relaxed our assumptions of equal relative productivities and equal shares for two categories in two different groups. Unfortunately, our estimates turned out to be too imprecise, perhaps due to the sample bias issue. 46Depending on the company and the decade, both effects could be found in the airline industry in the U.S. once the product market became more competitive. 26 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? The social effects of the premium are unclear. Low returns to secondary education could create cobweb cycles, where periods of an expansion of enrollment are followed by shortage of graduates as parents respond with a lag to fluctuating returns. In order to realize the social returns and support increased investment in manufacturing, governments would then have to subsidize the expansion of secondary more than otherwise. At the same time, many workers in these firms are supporting a number of dependents ­ in their immediate family and outside of it. If firms were able to lay off older workers cheaply, in a situation of few if any mechanisms for income smoothing, the poverty effect could be substantial. Thus, the overall economic and social impact of the cohort premium is difficult to estimate, and will vary country by country. In OECD countries, firms have tried to shift the costs of older workers to the public purse through early retirement on public pensions. African governments cannot afford this solution. As for the union effect, we find important differences between Anglophone and Francophone countries: in Anglophone countries all workers benefit significantly from union presence at the firm, regardless of their membership. In Francophone countries only older union members get a higher wage than non-members. In summary, our results indicate that institutional aspects play an important role in wage determination, by giving certain groups of workers more bargaining power. But this effect differs according to the legal origin of the country. 27 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? References 1. Abowd, J., F. Kramarz and D. Margolis, 1999. "High Wage Workers and High Wage Firms," Econometrica, Vol. 67(2), pp. 251-333. 2. Alby, P., 2003. "Unequal Rent-Sharing and Wage Determination in the Formal Ivorian Economy," manuscript, Université de Toulouse. 3. Alby, P., 2007. "Wage Determination in Sub-Saharan Africa: Does the Legal Origin Matter?" manuscript, Université de Toulouse. 4. Altonji J., and N. Williams, 2005. "Do Wages Rise With Job Seniority? A Reassessment," Industrial and Labor Relations Review, Vol. 58(3), pp. 370-397. 5. Appleton, S., Hoddinott, J. and Mackinnon, J., 1996. "Education and Health in Sub-Saharan Africa," Journal of International Development, Vol. 8(3), pp. 307-339. 6. Association for the Development of Education in Africa, 1999. Statistical Profile of Education in sub- Saharan Africa, 1999 Edition, Paris. 7. Avenstrup, R., 2006. "Reducing Poverty Through Free Primary Education," in Attacking Africa's Poverty: Experience from the Ground, L. Fox and R. Liebenthal, eds., The World Bank, Washington DC. 8. Azam, J.P. and C. Ris, 2001. "Rent Sharing, Hold Up, and Manufacturing Sector Wages in Côte d'Ivoire," manuscript, Université de Toulouse. 9. Azam, Jean-Paul and Jean-Yves Lesueur, 1997. "Efficiency Wage and Supervision: Theory and Application to the Ivorian Manufacturing Sector," Journal of African Economies, Vol. 6(3), pp. 445-462. 10. Beck, T. and Levine, R., 2004. "Legal Institutions and Financial Development," NBER Working Papers 10417, National Bureau of Economic Research. 11. Benhassine, N., Fafchamps, M. and Söderbom, M., 2006. "Wage Gaps and Job Sorting in African Manufacturing," CEPR Discussion Papers 6003, C.E.P.R. Discussion Papers. 12. Bigsten, A. and Söderbom, M., 2006. "What Have We Learned from a Decade of Manufacturing Enterprise Surveys in Africa?," World Bank Research Observer, Oxford University Press, Vol. 21(2), pp. 241-265. 13. Bigsten, A., Isaksson, A., Söderbom, M., Collier, P., Zeufack, A., Dercon, S., Fafchamps, M., Gunning, J., Teal, F., Appleton, S., Gauthier, B., Oduro, A., Oostendorp, R. and Pattillo, C., 2000. "Rates of Return on Physical and Human Capital in Africa's Manufacturing Sector," Economic Development and Cultural Change, Vol. 48(4), pp. 801-827. 14. Blanchflower, D. G., A. J. Oswald and P. Sanfey, 1996. "Wages, Profits, and Rent-Sharing", Quarterly Journal of Economics, Vol. 111, pp. 227-252. 15. Bosworth, B. and S. Collins, 2003. "The Empirics of Growth: an Update," Brookings Institution, Washington DC. 16. Fafchamps, M. and Söderbom, M., 2006. "Wages and Labor Management in African Manufacturing." Journal of Human Resources Vol. 41(2), pp. 346­379. 17. Fox, L. and A.M. Oviedo, 2008. "Institutions and Labor Market Outcomes in Sub-Saharan Africa," manuscript, The World Bank. 18. Fox, L. and Gaal, M., 2008. Working Out of Poverty, The World Bank, Washington DC. 19. Heckman, J.J. and Pagés, C., 2000. "The Cost of Job Security Regulation: Evidence from Latin American Labor Markets," NBER Working Papers 7773. 20. Hellerstein, J., D. Neumark and K. Troske, 1999. "Wages, Productivity, and Worker Characteristics: Evidence from Plant-Level Production Functions and Wage Equations," Journal of Labor Economics, Vol. 17(3), pp. 409-446. 28 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? 21. ILO, 2004. Yearbook of Labour Statistics. 22. International Comparison Program, 2008. 2005 International Comparison Program. Tables of final results, the World Bank, Washington DC. 23. Krueger, A. and L. Summers, 1988. "Efficiency Wages and the Inter-industry Wage Structure," Econometrica, Vol. 56, pp. 259-293. 24. La Porta, R., López-de-Silanes, F., Shleifer A. and Vishny, R., 1998. "Law and Finance," Journal of Political Economy, Vol. 106(6), pp. 1113-1155. 25. Loayza, N., Oviedo, A.M. and Servén, L., 2005. "The Impact of Regulation on Growth and Informality Cross-Country Evidence," World Bank Policy Research Working Paper No. 3623. 26. Manda, D., Bigsten, A., and Mwabu, G., 2001. "Trade Union Membership and Earnings in Kenyan Manufacturing Firms," Working Paper No. 50, Department of Economics, Göteborg University. 27. Mazumdar, D. and Mazaheri, A., 2002. Wages and Employment in Africa, Aldershot, Ashgate. 28. Murphy, K., and F. Welch, 1990. "Empirical Age-Earnings Profiles," Journal of Labor Economics, Vol. 8(2), pp. 202-229. 29. Ndulu, B., L. Chakraborti, L. Lijane, V. Ramachandran and J. Wolgin, 2007. "Challenges of African Growth: Opportunities, Constraints, and Strategic Directions," Washington DC. World Bank. 30. Psacharopoulos, G. and Arriagada, A.M., 1992. "The Educational Composition of the Labor Force: An International Update," PHREE Background Paper Series No. 92/49, The World Bank. 31. Söderbom, M. and Teal, F., 2001. "Firm Size and Human Capital as Determinants of Productivity and Earnings," Working Papers Series 2001-9, Centre for the Study of African Economies, University of Oxford. 32. Söderbom Måns, Francis Teal and Anthony Wambugu, 2004. "Does firm size really affect earnings?" Development and Comp Systems 0409011, EconWPA. 33. Söderbom, M., Teal, F., Wambugu, A. and Kahyarara, G., 2006. "The Dynamics of Returns to Education in Kenyan and Tanzanian Manufacturing," Oxford Bulletin of Economics and Statistics, Vol. 68(3), pp. 261-288. 34. Stifel, D., Rakotomanana, F. and Celada, E., 2007. "Assessing Labor Market Conditions in Madagascar, 2001-2005," Working paper, The World Bank. 35. Teal, F., 1996. "The Size and Sources of Economic Rents in a Developing Country Manufacturing Labor Market," Economic Journal, Vol. 106, pp. 963-976. 36. Topel, R., 1991. "Specific Capital, Mobility, and Wages: Wages Rise With Job Seniority," Journal of Political Economy, Vol. 99(1), pp. 145-76. 37. Van Biesebroeck, J., 2007. "Wage and Productivity Premiums in Sub-Saharan Africa," WP 291, University of Toronto, Department of Economics. 38. Verner, Dorte, 1999. "Wage and Productivity Gaps: Evidence From Ghana," World Bank Policy Research Working Paper No. 2168. 39. World Bank, 2007(a). World Development Indicators 2007, The World Bank, Washington DC. 40. World Bank, 2007(b). Africa Development Indicators 2007, The World Bank, Washington DC. 41. World Bank, 2008(a). Doing Business Report. www.doingbusiness.org 42. World Bank, 2008(b). Enterprise Surveys. www.enterprisesurveys.org 29 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Appendix: Tables Table 1: Manufacturing, value added (% of GDP) Year 2002 2003 2004 2005 2006 Botswana 4.2 4.3 4.1 3.7 3.7 Gambia 5.7 5.3 5.3 5.2 Ghana 9.0 9.0 8.8 8.3 7.6 Kenya 11.3 11.2 11.2 11.5 12.1 Namibia 11.1 12.4 13.5 13.1 12.9 Anglophone Swaziland 37.9 38.4 37.9 37.2 36.8 Tanzania 7.3 7.2 7.0 6.8 6.9 Uganda 10.0 9.3 9.2 9.2 8.6 Benin 8.6 8.5 7.6 7.5 Burkina Faso 12.3 12.9 Burundi 8.4 8.5 8.5 8.8 Cameroon 20.6 20.2 19.2 18.7 18.5 Cape Verde 6.1 8.0 4.8 Congo, DR 5.4 5.4 6.3 6.6 6.5 Guinea 4.1 3.9 3.7 3.7 3.7 Francophone Madagascar 12.5 13.7 14.2 14.0 13.4 Mali 3.2 2.8 3.4 3.2 3.1 Mauritania 6.2 6.0 5.9 5.0 Rwanda 11.2 8.9 8.0 8.2 8.5 Senegal 12.5 11.7 11.5 10.9 10.2 Source: ADI (2007) Table 2: Estimates of returns to education using firm survey data Authors Countries Years Returns Increasing Comments Mazumdar and Cameroon, Cote d'Ivoire, early positive, yes Mazaheri (2002) Ghana, Kenya, Tanzania, 1990s significant Zambia, Zimbabwe Bigsten et al. Cameroon, Ghana, early positive, yes Returns to human capital are (2000) Kenya, Zambia, 1990s significant much lower if measured at Zimbabwe the firm level via the production function. Teal (in progress) Ghana 1993-95 positive, yes cited in Appleton et al. significant (1996) Söderbom et al. Kenya, Tanzania 1993- positive, yes Find also cohort effects (2006) 2001 significant (young vs. old). Returns are higher for older workers. The studies use manufacturing firm and worker data from the RPED dataset. Their dependent variable is worker wages. 30 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 3: Doing Business Rankings for Sub-Saharan Africa Mean* Minimum Maximum General rank 135.4 27 178 Trade regulation rank 129.9 16 178 Entry regulation rank 125.3 8 178 Land registration rank 123.2 26 173 Bankruptcy regulation rank 121.3 26 178 Labor regulation rank 116.4 10 176 Contract enforcement rank 115 33 176 Access to credit rank 112.8 13 176 Licensing regulation rank 112.5 9 178 Investor Protection rank 111 9 175 Taxation regulation rank 104.6 11 176 *Mean over 46 SSA countries. Source: Doing Business 31 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 4: Industry distribution by country (number of firms) st r and for uctsd pe uctsd tscud dustryin alsc oducrpdte rials pro pafo pro pros athereld er an Furniture Wood Oth Total Agro Chemi Mate construction perap rela Metallic rystudnI and Plastic Textiles Botswana (2006) 12 10 19 19 8 8 0 28 0 10 114 Gambia (2006) 5 0 2 8 11 1 1 4 0 1 33 Ghana2007 80 7 0 0 21 0 6 128 0 71 313 Kenya (2003) 69 25 17 8 42 18 23 68 12 0 282 Namibia (2006) 18 4 9 22 16 10 2 7 0 18 106 Swaziland (2006) 15 5 8 4 2 10 2 21 0 3 70 Tanzania (2003) 81 27 11 65 29 25 7 31 0 0 276 Anglophone Tanzania (2006) 70 15 9 60 27 16 6 56 0 14 273 Uganda (2003) 122 18 40 47 21 23 7 15 7 0 300 Uganda (2006) 92 8 10 93 53 22 0 17 0 12 307 Total 564 119 125 326 230 133 54 375 19 129 2,074 Benin (2004) 50 9 10 61 27 45 3 7 25 10 247 Burkina Faso (2006) 14 1 1 0 3 12 2 4 7 7 51 Burundi (2006) 23 14 2 22 6 8 1 26 0 0 102 Cameroon (2006) 31 11 6 0 8 19 6 7 18 13 119 Cape Verde (2006) 12 1 2 0 0 0 1 5 16 10 47 DR Congo (2006) 59 16 3 24 10 4 2 22 0 9 149 Guinea (2006) 27 3 0 29 18 0 0 41 0 17 135 Madagascar (2005) 45 17 0 64 20 6 13 89 0 38 292 Francophone Mali (2003) 51 16 19 13 22 11 7 10 5 0 154 Mauritania (2006) 29 5 4 13 12 8 0 5 0 4 80 Rwanda (2006) 21 7 2 7 4 9 1 7 0 1 59 Senegal (2003) 93 34 19 6 25 36 15 23 10 1 262 Total 455 134 68 239 155 158 51 246 81 110 1,697 Total 1,019 253 193 565 385 291 105 621 100 239 3,771 32 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 6: Education of the labor force and manufacturing workers (percent) Coverage Country Year < 6 yrs 6-9 yrs 10-12 yrs 12 + yrs Labor force Botswana 1996 22.1 63.4 13.8 Manufacturing(a) Botswana 2006 18.4 13.5 64.6 3.0 Labor force(§) Burkina Faso 2003 72.9 17.6 6.0 3.5 Manufacturing(a) Burkina Faso 2006 32.1 34.9 11.6 20.8 Labor force(§) Cameroon 2001 74.1 16.5 6.9 2.5 Manufacturing(a) Cameroon 2006 11.2 39.4 16.6 32.8 Labor force(§) Ghana 1998 65.0 25.1 8.6 1.2 Manufacturing(a) Ghana 2007 25.9 16.6 50.0 5.1 Labor force Guinea 1998 61.3 27.5 8.0 3.2 Manufacturing(a) Guinea 2006 52.2 31.5 8.4 7.6 acir Labor force() Kenya 2005 63.5 34.9 1.6 Manufacturing(a) Kenya 2003 20.6 47.7 3.0 28.7 Af Labor force(b) Madagascar 2005 54.0 39.5 3.9 2.6 Manufacturing(a) Madagascar 2005 19.4 14.7 48.0 17.9 Labor force Namibia 1997 15.6 37.2 42.1 4.7 Manufacturing(a) Namibia 2006 15.2 14.6 68.3 1.5 Labor force Rwanda 1996 41.7 53.8 3.4 0.3 Manufacturing(a) Rwanda 2006 27.7 31.6 16.9 23.7 Labor force(§) Senegal 2001 86.4 10.7 1.0 Manufacturing Senegal 2003 40.6 22.0 18.3 16.5 Labor force Tanzania 1991 53.5 43.2 Manufacturing Tanzania 2003 49.6 37.0 13.4 Labor force(c) Uganda 2005 17.4 47.8 31.1 3.7 Manufacturing Uganda 2003 24.2 40.5 34.0 iasAts Labor force Indonesia 1996 28.2 52.6 16.3 3.0 Manufacturing Indonesia 2003 12.2 35.9 Labor force South Korea 2003 26.1 43.6 30.3 Ea Manufacturing South Korea 2005 1.2 1.4 53.6 43.9 eldd Labor force Morocco 2003 36.7 40.2 10.2 7.6 &ts htr aci Manufacturing Morocco 2004 39.2 33.3 19.1 8.4 Labor force Oman 2000 26.3 34.9 25.0 13.8 Mi Ea No Afr Manufacturing Oman 2003 14.8 21.8 29.7 33.7 Labor force Pakistan 2002 47.4 18.4 11.2 22.2 htu ai Manufacturing Pakistan 2002 59.3 18.7 12.5 9.5 So As Labor force Sri Lanka 2003 68.7 16.8 14.6 Manufacturing SriLanka 2004 24.7 33.2 36.8 5.3 Sources: Labor force - ILO; Manufacturing - Enterprise Surveys (a): Data comes from workers' module of Enterprise Surveys (b): Individuals employed in primary, industry, and services sectors. Source: Stifel et al. (2007) (c): Employed population, 15-64 years of age. Source: Household Surveys. (§) Individuals older than 10. Less than 6 years includes completed primary. Source: Household Surveys (): Individuals older than 15 years old. Source: Household Surveys. 33 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 7: Average years of education of employees Anglophone Francophone By firm size Small (<20) 10.7 9.1 Medium (20-99) 10.8 11.3 Large (100+) 12.3 12.7 Total 11.1 10.6 By worker occupation Management 13.7 12.7 Professional 14.4 15.5 Skilled production 11.2 11.2 Unskilled production 10.0 8.3 Non-production 10.5 10.4 Total 11.1 10.6 By worker age <20 years old 9.2 7.2 20 to 29 years old 11.0 9.5 30 to 39 years old 11.5 11.2 40 to 49 years old 11.0 11.6 50 to 59 years old 10.8 10.7 60 years old and older 9.9 10.8 Total 11.1 10.6 Source: Enterprise Surveys Table 8: Average years of education of employees, 1990s and 2000s Burundi (1992-1995)* 4.9 Burundi (2006) 8.72 Ghana (1992-1995)* 9.5 Ghana (2007) 10.47 Kenya (1992-1999)* 8.6 Kenya (2003) 11.52 Tanzania (1992-1995)* 8.2 Tanzania (2006) 10.55 Source: Enterprise Surveys (workers' module). (*) Source: Benhassine et al. (2006) from RPED 34 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 9: Distribution of years of education firms and workers < 6 yrs 6-9 yrs 10-12 yrs 12 + yrs Kenya (2003) 3.3 23.4 71.5 0.9 Kenya (2003) 4.8 22.3 36.5 36.4 Uganda (2003) 24.7 40.8 31.4 3 Uganda (2003) 7.1 17.7 26.5 48.7 Benin (2004) 30.5 34.9 18.7 15 Benin (2004) 17.5 35.1 18.8 28.6 Madagascar (2005) 21.7 28.2 15.6 9.5 Madagascar (2005) 7.2 22.9 26.7 43.2 Mali (2003) 34.2 26.2 27.2 11 Mali (2003) 15.4 24.9 13.8 46 Senegal (2003) 40.6 22 18.3 16.5 Senegal (2003) 7.1 27.7 20.2 45 Note: The first row corresponds to the distribution of workers by education years at the firm level (comes from the firm survey). Shaded rows denote averages from the worker's surveys. Source: Authors' calculations from Enterprise Surveys 35 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 10: Occupation distribution (percent) WORKERS Skilled production Unskilled production Non-production Botswana (2006) 34.3 41.9 18.4 Gambia (2006) 42.6 19.7 18.9 Ghana (2007) 44.9 31.4 12.2 Kenya (2003) 36.2 21.9 24.7 ne Namibia (2006) 40.2 37.7 17.5 Swaziland (2006) 20.2 51.2 19.0 phool Tanzania (2003) 26.1 11.1 38.3 ngA Tanzania (2006) 36.1 38.8 16.0 Uganda (2003) 21.5 10.3 35.5 Uganda (2006) 23.8 56.8 12.8 Male 31.0 25.0 24.0 Female 31.0 26.0 31.0 Total 30.9 25.6 25.5 Benin (2004) 25.1 19.8 31.2 Burkina Faso (2006) 18.2 35.1 28.2 Burundi (2006) 36.3 41.1 12.0 Cameroon (2006) 28.7 25.9 26.2 Cape Verde (2006) 17.0 50.0 27.0 eno DR Congo (2006) 39.3 33.2 16.2 Guinea (2006) 25.0 57.4 10.2 phocna Madagascar (2005) 18.8 43.9 26.2 Mali (2003) 33.1 24.2 16.3 Fr Mauritania (2006) 21.2 55.2 10.9 Rwanda (2006) 18.6 35.6 24.3 Senegal (2003) 26.5 30.7 23.3 Male 29.0 35.0 19.0 Female 14.0 29.0 41.0 Total 26.0 33.0 24.0 FIRMS Skilled production Unskilled production Non-production Botswana (2006) 43.9 30.2 22.6 Gambia (2006) 59.8 15.5 17.0 Ghana (2007) 56.0 25.0 15.3 enoh Kenya (2003) 20.8 21.9 10.0 Namibia (2006) 41.6 33.4 20.1 oplg Swaziland (2006) 35.1 34.7 26.7 Tanzania (2003) 32.2 26.5 14.7 An Tanzania (2006) 54.1 21.5 15.3 Uganda (2003) 19.5 20.4 6.1 Uganda (2006) 40.4 33.0 19.9 Total 50.4 25.4 15.8 Benin (2004) 40.7 24.3 19.2 Burkina Faso (2006) 35.1 38.4 23.7 Burundi (2006) 24.6 50.4 22.5 Cameroon (2006) 38.1 39.2 20.7 eno Cape Verde (2006) 30.0 42.0 25.2 DR Congo (2006) 36.8 36.1 23.0 phocna Guinea (2006) 51.1 32.9 12.4 Madagascar (2005) 21.9 49.0 8.1 FrMali (2003) Mauritania (2006) 36.5 34.1 25.4 Rwanda (2006) 25.2 26.5 27.4 Senegal (2003) Total 34.2 36.4 19.4 * In most countries, the categories provided in the firm-level questionnaire were limited to: skilled production, unskilled production, and non-production. Source: Enterprise Surveys 36 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 11: Unionized workforce Workforce unionized at the Unionized firms firm level Unionized workers (percentage)* (percentage) (percentage)§ Botswana (2006) 18.1 11.7 5.0 Gambia (2006) 17.6 9.1 17.0 Ghana2007 50.4 24.4 42.0 Kenya (2003) 63.6 44.1 25.0 Namibia (2006) 50.5 36.2 29.0 Swaziland (2006) 37.2 21.8 33.0 Tanzania (2003) 51.6 40.1 38.0 Tanzania (2006) 46.7 34.8 34.0 Anglophone Uganda (2003) 10.1 6.5 5.0 Uganda (2006) 8.3 2.6 4.0 Male 22.0 Female 27.0 Total 42.8 24.1 23.0 Benin (2004) 6.5 3.6 4.0 Burkina Faso (2006) 35.3 27.2 16.0 Burundi (2006) 1.3 1.3 3.0 Cameroon (2006) 31.1 17.9 34.0 Cape Verde (2006) 46.8 31.2 30.0 DR Congo (2006) 23.9 15.1 30.0 Guinea (2006) 8.8 3.8 15.0 Madagascar (2005) 27.5 10.0 11.0 Mali (2003) 43.7 29.6 29.0 Francophone Mauritania (2006) 42.8 27.8 24.0 Rwanda (2006) 12.0 8.8 10.0 Senegal (2003) 46.5 30.3 27.0 Male 19.0 Female 15.0 Total 24.5 14.8 18.0 *: A firm is unionized if it has any number of unionized workers. Over half of the firms have no unionized workers at all. §: In the workers' sample. Source: Enterprise Surveys 37 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 12: Unionization rate by firm age (percentage) Anglophone Francophone Less than 5 years old 10.5 6.7 5 to 9 years old 22.3 8.5 10 to 19 years old 21.6 17.1 20 to 39 years old 31.6 25.2 40 years old and older 44.4 36.2 Total 24.0 14.8 Source: Enterprise Surveys 38 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 13: Annual wage statistics by country WORKERS* Mean Median Observations Botswana (2006) 10,059 5,950 352 Gambia (2006) 5,449 4,762 111 Ghana (2007) 4,497 3,062 658 Kenya (2003) 8,061 3,531 1,881 Namibia (2006) 8,835 5,134 321 Swaziland (2006) Tanzania (2003) 6,714 2,964 1,529 Anglophone Tanzania (2006) 5,825 3,636 579 Uganda (2003) 4,286 2,013 1,239 Uganda (2006) 5,424 4,355 660 Total 6,498 3,444 7,330 Benin (2004) 3,739 2,224 1,668 Burkina Faso (2006) 9,324 4,500 250 Burundi (2006) 3,182 2,309 365 Cameroon (2006) 8,043 5,259 617 Cape Verde (2006) 4,759 3,426 284 DR Congo (2006) 3,501 2,467 392 Guinea (2006) 3,912 2,791 316 Madagascar (2005) 12,013 8,348 1,723 Francophone Mali (2003) 4,341 2,270 686 Mauritania (2006) 8,503 6,364 212 Rwanda (2006) 11,965 5,419 171 Senegal (2003) 5,911 3,604 1,483 Total 6,728 3,987 8,167 39 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? (Table 13, continued) FIRMS Mean Median Observations Botswana (2006) 8,124 5,165 113 Gambia (2006) 2,817 2,405 33 Ghana (2007) 2,115 1,417 313 Kenya (2003) 10,161 2,583 231 Namibia (2006) 8,565 5,679 104 Swaziland (2006) 6,520 4,159 69 Tanzania (2003) 3,328 1,262 215 Anglophone Tanzania (2006) 4,373 2,727 272 Uganda (2003) 2,114 1,487 276 Uganda (2006) 4,017 2,540 307 Total 3,269 1,913 1,933 Benin (2004) 2,994 1,662 244 Burkina Faso (2006) 5,078 3,522 50 Burundi (2006) 3,131 1,924 102 Cameroon (2006) 8,691 5,390 118 Cape Verde (2006) 6,424 3,403 47 DR Congo (2006) 2,615 1,888 149 Guinea (2006) 3,790 1,406 135 Madagascar (2005) 10,991 7,845 217 Francophone Mali (2003) 2,899 1,616 135 Mauritania (2006) 5,469 4,606 80 Rwanda (2006) 4,793 2,867 59 Senegal (2003) 4,598 2,295 205 Total 4,853 2,472 1,541 *: Annual wage for a worker is computed as follows. The worker is asked the unit of payment (hourly, daily, etc.), the amount paid by unit (in current LCU), and the number of hours worked per week. First, we compute the monthly wage on the basis of the payment unit, the amount, and the number of hours worked per week (where relevant). Then, we multiply this number by the implied months worked per year, on the basis of the hours worked per week (assuming workers work on average 48 weeks per year). For firms, we take the total annual labor cost (in current LCU) and divide it by the number of employees (permanent and temporary). All wages are then transformed into USD-PPP (PPPs are from 2005). Source: Authors' calculations from Enterprise Surveys 40 L. Fox and A.M. Oviedo Are Skills Rewarded in Sub-Saharan Africa? Table 14: Monthly wages by country and occupation Median wage in USD-PPP Skilled Unskilled Non- Managers Professionals production production production Botswana (2006) 486 2,149 496 331 453 Gambia (2006) 761 198 331 228 397 Ghana (2007) 492 507 213 160 213 Kenya (2003) 1,119 832 256 166 236 Namibia (2006) 2,552 939 470 282 494 Swaziland (2006) 237 395 593 289 441 Tanzania (2003) 549 442 177 221 148 Anglophone Tanzania (2006) 530 606 253 189 202 Uganda (2003) 288 320 128 112 80 Uganda (2006) 726 492 258 194 194 Total 531 485 242 197 167 Benin (2004) 289 491 186 124 144 Burkina Faso (2006) 900 750 490 175 325 Burundi (2006) 564 146 175 117 146 Cameroon (2006) 1,215 996 438 319 319 Cape Verde (2006) 721 1,189 324 231 216 DR Congo (2006) 317 251 152 164 175 Guinea (2006) 205 410 164 148 246 Madagascar (2005) 1,649 1,539 742 536 660 Francophone Mali (2003) 394 315 189 110 132 Mauritania (2006) 455 909 505 404 313 Rwanda (2006) 2,151 1,340 645 229 376 Senegal (2003) 820 601 330 207 270 Total 505 681 300 239 258 Source: Authors' calculations from Enterprise Surveys (for calculation details, see Table 19) 41 Table 15: Wage determination regressions - pooled sample Pooled sample Dependent variable: Non-GDP-weighted GDP-weighted log of monthly wage (PPP) [1] [2] [3] [4] [5] Hrs. per week worked 0.004** 0.005** 0.007*** 0.001 0.001 [2.22] [2.39] [3.19] [0.76] [0.67] Female worker -0.045** -0.065*** -0.044 -0.059** -0.053* [2.00] [2.93] [1.61] [2.12] [1.85] Education (years) 0.032*** 0.012 0.011 0.002 0.005 [2.67] [0.99] [0.84] [0.19] [0.39] Education years squared 0.002*** 0.001** 0.001** 0.001** 0.001* [3.96] [2.42] [2.45] [2.14] [1.80] Junior/senior secondary degree 0.162*** 0.189*** 0.103** 0.112*** [4.76] [4.38] [2.42] [2.59] Vocational degree 0.346*** 0.328*** 0.146** 0.151** [7.25] [5.46] [2.45] [2.50] Tertiary/university degree 0.807*** 0.843*** 0.534*** 0.544*** [16.62] [14.18] [9.68] [9.91] Age > 32 (dummy) 0.143*** 0.133*** 0.157*** 0.114*** 0.118*** [5.48] [5.23] [4.83] [3.88] [4.05] Tenure (years) 0.014*** 0.016*** 0.012*** 0.014*** 0.011*** [4.54] [5.14] [3.47] [4.25] [4.02] Tenure^4 0.000 0.000 0.000 0.000 [1.02] [0.89] [0.65] [1.63] Previous experience (years) 0.016*** 0.016*** 0.013*** 0.009*** 0.009*** [6.13] [6.31] [3.85] [2.83] [3.89] Experience^4 0.000 0.000 0.000 0.000 [1.57] [1.61] [1.50] [0.36] Permanent, full-time worker 0.337*** 0.287*** 0.269*** 0.266*** 0.249*** [7.48] [6.53] [5.21] [5.24] [5.33] Manager 0.849*** 0.863*** [8.74] [8.70] Professional 0.559*** 0.583*** [10.09] [11.13] Skilled production worker 0.205*** 0.207*** [5.62] [5.66] Non-production worker 0.122*** 0.119*** [3.97] [3.70] Belongs to union 0.049 0.031 [1.34] [0.79] Related to owner/manager -0.124*** -0.126*** [2.69] [2.69] Medium firm (20-99 empl.) 0.06 0.043 0.035 0.035 0.021 [1.57] [1.17] [0.67] [0.58] [0.33] Large firm (100+ empl.) 0.122*** 0.085** 0.058 0.096 0.077 [2.72] [2.01] [1.09] [1.60] [1.25] Exporter 0.146*** 0.137*** 0.148*** 0.138*** 0.124*** [4.49] [4.39] [3.93] [3.61] [3.27] Unionized firm 0.067 [1.62] Constant 3.624*** 3.757*** 3.618*** 4.089*** 4.114*** [20.89] [21.80] [17.68] [25.81] [25.75] Observations 11580 11559 11559 9697 9435 R-squared 0.39 0.42 0.36 0.41 0.42 Notes: Column (2) reports the coefficients for the basic specification, weighted by the GDP of the country. Robust t- statistics in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%. Wages are converted to PPP US dollars (see Table 19). All regressions include country and industry dummies, and hours worked per week. Countries: Benin, Botswana, Burkina Faso, Burundi (excluded), Cameroon, Cape Verde, DR Congo, Gambia, Ghana, Guinea, Kenya, Madagascar, Mali, Mauritania, Namibia,Rwanda, Senegal, Swaziland, Tanzania, Uganda. Industries: Food & beverages (excluded), Chemicals, paints & pharmaceuticals, Construction materials, Metals, Paper & printing, Plastics, Textiles, garments & leather, Wood, Other manufacturing. Other excluded categories: <20 empl (firms), unskilled workers, no education/primary degree. 42 Table 16: Wage determination regressions - by legal origin Anglophone Francophone Dependent variable: Basic Augmented Basic Augmented log of monthly wage (PPP) (1) (2) (3) (1) (2) (3) Hrs. per week worked 0.010*** 0.003 0.003 -0.001 -0.002 -0.001 [3.43] [1.41] [1.13] [0.52] [0.70] [0.44] Female worker -0.034 -0.067* -0.054 -0.076*** -0.055* -0.052* [0.88] [1.74] [1.38] [2.78] [1.79] [1.66] Education (years) 0.013 -0.008 -0.006 0.002 0.008 0.01 [0.78] [0.52] [0.39] [0.13] [0.45] [0.54] Education years squared 0.001* 0.001* 0.001* 0.001** 0.001 0.001 [1.86] [1.79] [1.67] [2.40] [1.35] [1.08] Junior/senior secondary degree 0.239*** 0.141** 0.156** 0.107** 0.089* 0.098** [3.88] [2.10] [2.29] [2.27] [1.88] [2.06] Vocational degree 0.304*** 0.059 0.072 0.374*** 0.321*** 0.325*** [3.85] [0.74] [0.89] [5.50] [4.90] [4.88] Tertiary/university degree 0.928*** 0.497*** 0.496*** 0.717*** 0.597*** 0.613*** [11.30] [5.89] [6.10] [9.86] [8.95] [9.05] Age > 32 (dummy) 0.187*** 0.133*** 0.152*** 0.113*** 0.075* 0.062* [3.89] [2.94] [3.33] [3.17] [1.91] [1.65] Tenure (years) 0.005 0.011** 0.007* 0.021*** 0.017*** 0.017*** [0.92] [2.35] [1.79] [3.84] [2.98] [4.76] Tenure^4 0.000 0.000 0.000 0.000 [0.15] [1.59] [0.00] [0.20] Previous experience (years) 0.015*** 0.009* 0.009** 0.012*** 0.010*** 0.012*** [2.79] [1.71] [2.22] [3.61] [2.93] [4.40] Experience^4 -0.000* 0.000 0.000 0.000 [1.80] [0.86] [0.94] [1.11] Permanent, full-time worker 0.268*** 0.276*** 0.252*** 0.286*** 0.261*** 0.252*** [3.92] [4.01] [4.18] [4.42] [3.73] [3.62] Manager 1.048*** 1.072*** 0.525*** 0.540*** [7.58] [7.55] [6.12] [6.19] Professional 0.728*** 0.741*** 0.413*** 0.454*** [7.75] [7.63] [6.00] [7.33] Skilled production worker 0.239*** 0.246*** 0.170*** 0.170*** [4.88] [5.04] [3.82] [3.73] Non-production worker 0.139*** 0.134*** 0.100*** 0.104** [2.81] [2.62] [2.60] [2.58] Belongs to union -0.023 -0.073 0.141*** 0.153*** [0.43] [1.34] [3.31] [3.20] Related to owner/manager -0.092 -0.102 -0.152** -0.156*** [1.32] [1.43] [2.52] [2.62] Medium firm (20-99 empl.) 0.016 0.018 -0.021 0.073** 0.068* 0.085** [0.20] [0.17] [0.19] [2.10] [1.87] [2.31] Large firm (100+ empl.) 0.012 0.086 0.036 0.105* 0.087 0.117* [0.16] [0.88] [0.35] [1.68] [1.30] [1.96] Exporter 0.143*** 0.134** 0.108** 0.147*** 0.130*** 0.129*** [2.65] [2.41] [1.97] [3.64] [2.92] [2.88] Unionized firm 0.160*** -0.047 [2.78] [0.96] Constant 4.292*** 4.720*** 5.160*** 4.193*** 4.214*** 4.179*** [14.99] [17.90] [29.67] [21.26] [20.40] [20.19] Observations 5041 3785 3710 6518 5912 5725 R-squared 0.27 0.32 0.32 0.52 0.54 0.54 Notes: Robust t-statistics in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%. Wages are converted to PPP US dollars (see Table 19). All regressions include country and industry dummies, and hours worked per week. All observations are weighted by the country's GDP. Francophone: Benin (excluded), Burkina Faso, Cape Verde, Cameroon, DRCongo, Burundi, Guinea, Madagascar, Mauritania, Rwanda, Senegal. Anglophone: Botswana (excluded), Ghana, Gambia, Kenya, Namibia, Swaziland, Tanzania, Uganda. Industries: Food & beverages (excluded), Chemicals, paints & pharmaceuticals, Construction materials, Metals, Paper & printing, Plastics, Textiles, garments & leather, Wood, Other manufacturing. Other excluded categories: <20 empl (firms), unskilled workers, no education/primary degree. 43 Table 17: Wage determination by age groups All Anglophone Francophone 32 32+ 32 32+ 32 32+ Education (years) -0.011 0.007 -0.003 -0.026 -0.036* 0.043* [0.58] [0.50] [0.08] [1.59] [1.68] [1.70] Education years squared 0.002*** 0.001 0.002 0.002** 0.003*** -0.001 [2.61] [1.06] [1.36] [2.17] [3.18] [0.71] Vocational degree 0.004 0.320*** -0.111 0.309*** 0.349*** 0.303*** [0.06] [4.56] [1.20] [3.13] [4.46] [3.32] Junior/senior secondary degree 0.006 0.210*** -0.009 0.316*** 0.120** 0.079 [0.12] [4.02] [0.12] [3.86] [2.41] [1.24] Tertiary/university degree 0.407*** 0.660*** 0.246** 0.708*** 0.704*** 0.539*** [5.58] [9.49] [2.31] [6.73] [9.06] [5.81] Tenure 0.016*** 0.010*** 0.012 0.005 0.023*** 0.016*** [2.67] [3.24] [1.41] [1.11] [3.02] [4.35] Previous experience in industry 0.015*** 0.009*** 0.013 0.007* 0.020*** 0.011*** [2.59] [3.46] [1.40] [1.74] [3.62] [3.90] Permanent, full-time worker 0.249*** 0.218*** 0.227*** 0.276*** 0.320*** 0.116 [4.84] [3.06] [3.57] [2.84] [3.77] [1.39] Manager 0.731*** 0.928*** 1.022*** 1.090*** 0.177 0.744*** [5.57] [9.12] [5.71] [7.22] [1.36] [7.33] Professional 0.537*** 0.631*** 0.737*** 0.747*** 0.270*** 0.566*** [7.94] [9.20] [7.42] [4.64] [3.28] [7.35] Skilled production worker 0.150*** 0.264*** 0.191*** 0.306*** 0.089* 0.240*** [3.40] [5.42] [3.09] [4.78] [1.83] [3.62] Non-production worker 0.063 0.192*** 0.08 0.229*** 0.046 0.164*** [1.24] [4.17] [0.95] [3.52] [1.03] [2.64] Belongs to union 0.048 0.027 0.015 -0.136** 0.094* 0.192*** [1.00] [0.54] [0.24] [2.10] [1.85] [2.76] Related to owner/manager -0.096* -0.171*** -0.088 -0.129 -0.092 -0.227*** [1.80] [2.71] [1.07] [1.37] [1.48] [2.62] Unionized firm 0.081* 0.05 0.140** 0.172** -0.023 -0.084 [1.75] [0.91] [2.19] [2.15] [0.45] [1.23] Constant 4.189*** 4.232*** 4.804*** 5.401*** 4.340*** 4.185*** [21.46] [21.85] [16.37] [23.90] [18.37] [15.99] Observations 4760 4684 1983 1731 2777 2953 R-squared 0.4 0.4 0.28 0.34 0.59 0.51 Notes: Robust t-statistics in brackets. * significant at 10%; ** significant at 5%; *** significant at 1%. Wages are converted to PPP US dollars (see Table 19). All regressions include country, size, industry, and exporter dummies, hours worked per week, and dummies for female, and permanent full time workers. All observations are weighted by the country's GDP. Countries: Benin, Botswana, Burkina Faso, Burundi (excluded), Cameroon, Cape Verde, DR Congo, Gambia, Ghana, Guinea, Kenya, Madagascar, Mali, Mauritania, Namibia,Rwanda, Senegal, Swaziland, Tanzania, Uganda. Industries: Food & beverages (excluded), Chemicals, paints & pharmaceuticals, Construction materials, Metals, Paper & printing, Plastics, Textiles, garments & leather, Wood, Other manufacturing. Other excluded categories: <20 empl (firms), <32 years old (workers), unskilled workers, no education/primary degree. 44 Table 18: Average regressor values for basic regression sample All Anglophone Francophone Obs = 9435 Obs =3785 Obs = 5912 Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Female worker 0.25 0.43 0.26 0.44 0.23 0.42 Years of education completed 10.73 4.34 11.00 3.73 10.56 4.68 No education/primary degree 0.24 0.43 0.22 0.41 0.26 0.44 Junior/senior secondary degree 0.43 0.50 0.43 0.50 0.43 0.50 Vocational training 0.15 0.35 0.21 0.40 0.11 0.31 Tertiary/university degree 0.17 0.38 0.13 0.34 0.20 0.40 Age 34.09 8.97 33.54 8.78 34.45 9.07 Tenure 6.58 5.83 7.04 5.96 6.29 5.72 Years of prior experience 3.60 5.10 3.21 4.87 3.85 5.23 Permanent, full-time worker 0.91 0.29 0.90 0.30 0.92 0.28 Manager 0.05 0.23 0.06 0.24 0.05 0.22 Professional 0.08 0.27 0.05 0.21 0.09 0.29 Skilled production worker 0.30 0.46 0.36 0.48 0.26 0.44 Unskilled production worker 0.35 0.48 0.34 0.47 0.36 0.48 Non-production worker 0.22 0.42 0.19 0.39 0.24 0.43 Belongs to union 0.20 0.40 0.25 0.43 0.17 0.38 Related to owner 0.10 0.30 0.07 0.25 0.12 0.33 Small firm (< 20 empl.) 0.42 0.49 0.38 0.49 0.45 0.50 Medium firm (20-99 empl.) 0.36 0.48 0.38 0.49 0.35 0.48 Large firm (100+ empl.) 0.22 0.41 0.24 0.43 0.20 0.40 Exporter 0.34 0.47 0.36 0.48 0.32 0.47 Unionized firm Benin 0.12 0.33 0.20 0.40 Botswana 0.03 0.17 0.08 0.26 Burkina Faso 0.01 0.11 0.02 0.14 Burundi 0.04 0.19 0.06 0.24 Cameroon 0.06 0.23 0.10 0.29 Cape Verde 0.03 0.16 0.04 0.21 DR Congo 0.04 0.20 0.07 0.25 Gambia 0.01 0.10 0.02 0.15 Ghana 0.06 0.23 0.14 0.35 Guinea 0.02 0.15 0.04 0.19 Kenya 0.16 0.36 0.40 0.49 Madagascar 0.14 0.35 0.24 0.42 Mali 0.00 0.00 Mauritania 0.02 0.14 0.03 0.18 Namibia 0.03 0.17 0.08 0.27 Rwanda 0.02 0.13 0.03 0.16 Senegal 0.11 0.31 0.18 0.39 Swaziland 0.00 0.00 Tanzania 0.05 0.21 0.12 0.33 Uganda 0.06 0.24 0.16 0.36 Food & beverages 0.23 0.42 0.22 0.41 0.24 0.42 Chemicals & paints 0.07 0.26 0.06 0.24 0.08 0.27 Construction materials 0.04 0.19 0.04 0.20 0.04 0.18 Furniture 0.15 0.35 0.16 0.37 0.14 0.34 Metals 0.09 0.29 0.11 0.32 0.08 0.27 Paper & printing 0.09 0.28 0.06 0.24 0.10 0.30 Plastics 0.04 0.19 0.04 0.20 0.04 0.18 Textiles & garments 0.18 0.38 0.22 0.42 0.15 0.36 Wood 0.04 0.21 0.02 0.12 0.06 0.24 Other manufacturing 0.08 0.26 0.06 0.24 0.08 0.28 45 Table 19: Wage and productivity premiums combining age, occupation, education, and union status Pooled sample Anglophone Francophone Production Wage Equal premia Production Wage Equal premia Production Wage Equal premia function equation P-value function equation P-value function equation P-value N=1438 N=7726 N=583 N=3210 N=855 N=4516 Unionized firm -0.013 0.01 0.048 0.11 0.158 -0.049 (0.107) (0.038) (0.158) (0.062) (0.139) (0.046) Union member 1.023 -0.008 0.839 1.375 -0.085 0.137 0.771 0.057 0.074 (0.156) (0.038) (0.316) (0.059) (0.164) (0.045) Age (32+) 0.83 0.02 0.155 0.729 0.049 0.076 0.892 -0.011 0.615 (0.133) (0.026) (0.176) (0.044) (0.192) (0.031) Manager 1.229 0.713 0.169 1.02 1.052 0.050 1.476 0.518 0.938 (0.327) (0.077) (0.453) (0.143) (0.526) (0.078) Professional 2.502 0.48 0.229 2.344 0.643 0.533 2.102 0.394 0.488 (0.851) (0.059) (1.120) (0.130) (1.018) (0.063) Skilled production 1.052 0.198 0.385 1.002 0.268 0.217 1.481 0.16 0.310 (0.166) (0.036) (0.217) (0.055) (0.311) (0.044) Vocational degree 1.496 0.173 0.353 3.000 0.18 0.058 0.540 0.219 0.037 (0.347) (0.058) (0.966) (0.086) (0.319) (0.068) Junior/senior sec. 1.25 0.134 0.627 1.762 0.233 0.294 1.234 0.068 0.572 (0.240) (0.040) (0.511) (0.067) (0.295) (0.051) University 1.286 0.531 0.534 0.436 0.474 0.000 1.426 0.515 0.857 (0.401) (0.057) (0.278) (0.094) (0.498) (0.069) H0: Age +32 coefficient =1 0.199 0.125 0.575 H0: Vocational coefficient =1 0.153 0.038 0.149 H0: Secondary coefficient =1 0.298 0.136 0.427 H0: University coefficient =1 0.475 0.042 0.392 H0: Manager coefficient =1 0.484 0.965 0.365 H0: Professional coefficient =1 0.078 0.230 0.279 H0: Skilled production coefficient =1 0.755 0.993 0.123 H0: Unionized worker coefficient =1 0.884 0.235 0.161 Joint test: occupation coefficients =1 0.306 0.688 0.377 Joint test: occupation premia are equal 0.197 0.125 0.756 Joint test: education coefficients =1 0.554 0.011 0.235 Joint test: education premia are equal 0.569 0.000 0.114 Notes: Shares of uinon members, older workers, managers, professionals, skilled production workers, holders of vocational, secondary, and upper-level degrees are calculated as the mean values at the firm level. The estimation of the production function and the wage equation is done jointly, clustering the errors at the firm level. Standard errors in parentheses below coefficients. Bold coefficients indicate significance at the 10% level or better. The production function coefficients represent one plus the productivity premium; the wage regression coefficients are semi-elasticities. The equality of premia test is a Wald test of the null hypothesis that the wage regression coefficient is equal to the production function coefficient minus one for the corresponding category. 46 Table 20: Wage and productivity premiums combining age, occupation, education, and union status All countries Anglophone Francophone Production Wage Equal premia Production Wage Equal premia Production Wage Equal premia function equation P-value function equation P-value function equation P-value N=1438 N=7726 N=583 N=3210 N=855 N=4516 Unionized firm 0.011 0.011 0.057 0.109 0.197 -0.047 (0.105) (0.038) (0.154) (0.062) 0.137 0.046 Union member 0.892 -0.011 0.504 1.423 -0.088 0.091 0.595 0.054 0.006 (0.144) (0.038) (0.304) (0.059) 0.167 0.045 Age (32+) 0.641 0.02 0.007 0.802 0.051 0.278 0.552 -0.013 0.017 (0.137) (0.026) (0.229) (0.044) 0.18 0.031 Manager 0.944 0.708 0.031 0.545 1.047 0.000 1.409 0.514 0.893 (0.330) (0.078) (0.301) (0.146) 0.775 0.079 Professional 2.072 0.468 0.439 1.541 0.626 0.920 1.555 0.383 0.881 (0.777) (0.059) (0.824) (0.129) 1.137 0.064 Skilled production 0.902 0.196 0.117 0.855 0.266 0.051 1.768 0.158 0.237 (0.185) (0.036) (0.207) (0.054) 0.513 0.044 Vocational degree 1.332 0.167 0.620 3.37 0.171 0.045 0.311 0.216 0.066 (0.330) (0.058) (1.106) (0.085) 0.487 0.067 Junior/senior sec. 0.841 0.126 0.133 1.613 0.221 0.426 0.872 0.062 0.479 (0.190) (0.040) (0.498) (0.066) 0.269 0.05 University 0.631 0.519 0.000 -0.382 0.468 0.000 0.994 0.505 0.155 (0.224) (0.056) (0.249) (0.094) 0.359 0.068 H0: Vocational coeff=1 0.314 0.032 0.157 H0: Secondary coeff =1 0.402 0.219 0.635 H0: University coeff =1 0.099 0.000 0.986 H0: Manager coeff=1 0.866 0.131 0.598 H0: Professional coeff =1 0.168 0.512 0.625 H0: Skilled production coeff =1 0.594 0.482 0.135 H0: Unionized worker coeff=1 0.452 0.164 0.015 Joint test: occupation coefficients =1 0.402 0.387 0.480 Joint test: occupation premia are equal 0.065 0.000 0.648 Joint test: education coefficients =1 0.118 0.000 0.524 Joint test: education premia are equal 0.000 0.000 0.159 Notes: Shares of uinon members, older workers, managers, professionals, skilled production workers, holders of vocational, secondary, and upper-level degrees are calculated as the mean values at the country, industry, size class, and firm age class levels. The estimation of the production function and the wage equation is done jointly, clustering the errors at the firm level. Standard errors in parentheses below coefficients. Bold coefficients indicate significance at the 10% level or better. The production function coefficients represent one plus the productivity premium; the wage regression coefficients are semi-elasticities. The equality of premia test is a Wald test of the null hypothesis that the wage regression coefficient is equal to the production function coefficient minus one for the corresponding category. 47 Table 21: Wage and productivity premiums by occupation, education, union status, and age All countries Anglophone Francophone Production Wage Equal premia Production Wage Equal premia Production Wage Equal premia function equation P-values function equation P-values function equation P-values N=1438 N=7726 N=583 N=3210 N=855 N=4516 Unionized firm -0.003 0.01 0.041 0.112 0.149 -0.05 (0.104) (0.038) (0.162) (0.063) (0.136) (0.046) Unionized worker 0.978 -0.012 0.947 1.755 -0.089 0.020 0.56 0.054 0.003 (0.153) (0.038) (0.364) (0.059) (0.163) (0.045) Age (32+) 0.575 0.019 0.00 0.474 0.050 0.000 0.499 -0.014 0.002 (0.121) (0.026) (0.147) (0.044) (0.152) (0.031) Manager 1.018 0.708 0.06 0.864 1.046 0.010 1.758 0.515 0.762 (0.348) (0.078) (0.402) (0.146) (0.799) (0.079) Professional 1.909 0.467 0.54 1.336 0.625 0.695 1.211 0.381 0.840 (0.725) (0.059) (0.694) (0.129) (0.831) (0.064) Skilled production 1.006 0.197 0.34 1.19 0.271 0.786 1.368 0.155 0.623 (0.198) (0.036) (0.293) (0.054) (0.430) (0.044) Vocational degree 1.299 0.167 0.678 3.303 0.171 0.048 0.419 0.218 0.118 (0.316) (0.058) (1.085) (0.085) (0.507) (0.068) Junior/senior sec. 0.863 0.126 0.180 1.532 0.221 0.519 0.837 0.062 0.399 (0.196) (0.040) (0.488) (0.066) (0.267) (0.050) University 0.807 0.521 0.004 -0.36 0.466 0.000 1.041 0.507 0.157 (0.251) (0.056) (0.240) (0.094) (0.329) (0.068) H0: Manager coeff=1 0.96 0.735 0.343 H0: Professional coeff =1 0.21 0.628 0.800 H0: Skilled production coeff =1 0.98 0.516 0.393 Joint test: occupation coefficients =1 0.64 0.85 0.72 Joint test: occupation premia are equal 0.22 0.07 0.94 H0: Vocational coeff=1 0.344 0.034 0.252 H0: Secondary coeff =1 0.486 0.276 0.542 H0: University coeff =1 0.442 0.000 0.902 Joint test: education coefficients =1 0.344 0.000 0.627 Joint test: education premia are equal 0.016 0.000 0.190 H0: Unionized worker coeff=1 0.886 0.038 0.007 Notes: Shares of uinon members, older workers, managers, professionals, skilled production workers, holders of vocational, secondary, and upper-level degrees are calculated as the mean values at the country, industry, size class, and firm age class levels. The estimation of the production function and the wage equation is done jointly, clustering the errors at the firm level. Standard errors in parentheses below coefficients. Bold coefficients indicate significance at the 10% level or better. The production function coefficients represent one plus the productivity premium; the wage regression coefficients are semi-elasticities. The equality of premia test is a Wald test of the null hypothesis that the wage regression coefficient is equal to the production function coefficient minus one for the corresponding category. 48