Wage misalignment in CFA countries: were labour market policies to ...

JOURNAL OF AFRICAN ECONOMIES, VOLUME 9, NUMBER 4, PP. 475–511

Wage Misalignment in CFA Countries: Were Labour Market Policies to Blame? Martín Rama1 The World Bank Development Research Group

Wage rigidity, stemming from highly distortive labour market policies, is a natural candidate to explain the overvaluation of the CFA franc after the adverse external shocks of the 1980s. This paper uses a variety of data sources to assess wage rigidity in CFA countries until the 1994 devaluation, and to analyse whether it was due to labour market policies. The paper shows that wages were high in CFA countries, compared with both wages in similar countries and the labour earnings of similar individuals within the same countries. It also shows that wages were rigid in real terms, in the sense of following closely the fluctuations of government wages and consumer prices, but it finds no evidence of nominal wage rigidity. Labour market policies may not be at the roots of wage misalignment and real rigidity, though. From an international perspective, minimum wages were not high enough to account for the observed wage misalignment. Moreover, their adjustment over time was highly responsive to real shocks. Private sector unions, in turn, seemed more instrumental in achieving wage moderation than wage drift. Their members usually had lower wages than similar, non-unionised workers, which probably reflects the ‘subordinate’ nature of the labour movement. The most likely candidates to explain wage misalignment and real rigidity in CFA countries in the 1980s and early 1990s are therefore government pay policies and (possibly) limited competition in product markets.

1

I am grateful to Ingrid Ivins, John McIntire, Peter Niessen and Abdoulaye Seck for support in gathering the data and to Raquel Artecona for assistance in processing them. Geoffrey Bergen, Alberto Chong, Gaurav Datt, Alvaro Forteza, Larry Hinkle, Jyotsna Jalan, seminar participants in Abidjan and Lomé and two anonymous referees provided very helpful insights and suggestions. The findings, interpretations and conclusions are entirely my own, and should not be attributed to the World Bank.

© Centre for the Study of African Economies, 2000

476 Martín Rama

1. Introduction In January 1994 the CFA franc was devalued from 50 to 100 per French franc. This devaluation marked the end of almost half a century of pegged exchange rates between the 13 countries of the Communauté Financière Africaine and their former colonial power.2 The 50-to-1 parity had been set in October 1948, at a time when CFA stood for Colonies Françaises d’Afrique. The convertibility of the CFA franc was guaranteed by the French Treasury. In return, CFA countries were supposed to follow strict rules of monetary discipline. Not surprisingly, their inflation was significantly lower than that of other Sub-Saharan African countries (Devarajan and Rodrik, 1991; Conway and Greene, 1993), and for some time their growth performance was better too (Devarajan and de Melo, 1987; Guillaumont et al., 1988). Although more recent studies put this apparent superiority in doubt (Assane and Pourgerami, 1994; Savvides, 1995), their results have been questioned on methodological grounds (Guillaumont and Guillaumont Jeanneney, 1995). At the roots of the 1994 devaluation lie two developments of the 1980s: the dramatic fall of export prices for cocoa, coffee and oil, and the substantial appreciation of the French franc against the US dollar. At first, an effort was made to adjust to these two shocks by cutting government expenditures and raising tax revenue, but this strategy failed to achieve the deflation that was required to restore competitiveness. A consensus gradually emerged: the problem was the downward rigidity of prices, resulting in turn from the downward rigidity of formal sector wages. In the presence of a nominal rigidity, devaluing the CFA franc appeared as the only way out of the recession. Labour market policies and institutions are a natural candidate to explain the stubborn currency overvaluation observed in CFA countries, despite apparent adjustment efforts. These countries replicated the kind of labour legislation adopted by France immediately after World War II, with a vengeance. There were national minimum wages, but also detailed salary grids setting the minima for workers in each sector at every level of employment, based on their formal training and seniority. Lay-offs required a government authorisation 2

The Communauté Financière Africaine is actually made of two zones, based on similar but separate agreements with France. The west zone includes Burkina Faso, Benin, Côte d’Ivoire, Mali, Senegal and Togo. The central zone includes Cameroon, Central African Republic, Chad, Congo and Gabon.

Wage Misalignment in CFA Countries 477

that was highly political in nature and could take months or even years to be granted, if at all. In some countries, hiring was under strict government control too, through large and inefficient placement offices centralising information on vacancies and job seekers. Finally, trade unions were well implanted in the formal sector of the economy, partly due to their role in the process that led to independence from France, but also because of significant barriers to competition in product markets. The goal of this paper is to evaluate whether and how the labour market policies characterising CFA countries contributed to the currency overvaluation observed before the 1994 devaluation. Note that the ‘how’ issue is as relevant as the ‘whether ’ issue. Labour market policies and institutions may be a source of wage rigidity in two substantially different ways. Formal sector wages may be high due to legal minima and mandated benefits. In this case there is nominal wage rigidity, implying that wages cannot decline in monetary terms. The effects of nominal rigidity on competitiveness can be offset through devaluation. But private sector wages can also be high due to trade union activities, to government pay policies, to incumbent workers taking advantage of large hiring and firing costs, or to employers and employees sharing the rents created by imperfectly competitive product markets. In these latter cases there is real wage rigidity, usually involving relatively stable ratios between formal sector wages on the one hand and government wages and consumer prices on the other. In the medium run, devaluation can do little to affect these ratios, and hence to restore competitiveness. Unfortunately, the literature is of not much help regarding the link between labour market policies and currency overvaluation in developing countries. Even in studies done by or for the World Bank, different views can be found. For instance, López and Riveros (1990) examined data from four Latin American countries and concluded that distortions in the formal labour market were a source of real wage rigidity, implying a low responsiveness of the real exchange rate to devaluation. On the other hand, Horton et al. (1994) considered a broader set of countries and concluded that inflation was effective in reducing real wages, thus raising the real exchange rate. It is probably difficult to find more opposite views on any economic issue. Given this ambiguity of the literature, a careful analysis of labour markets in CFA countries is warranted.

478 Martín Rama

2. Policies Affecting Wages In the early 1990s, the CFA zone was characterised by some of the heaviest labour market regulations in the world applied to some of its tiniest formal sectors. As Table 1 shows, employment in firms and agencies complying with regulations represented at most 5% of the labour force. Furthermore, roughly half of this employment was in the public sector. Only a handful of private sector firms was therefore confronted by minimum wages, hiring and firing constraints and other labour market regulations. At first glance, the small size of formal sector employment suggests that labour market policies and institutions were essentially irrelevant in CFA countries. However, most of the exports from these countries, as well as a key portion of the inputs used in the rest of their economies, were handled by formal sector firms. It is therefore of paramount importance to understand how these firms were affected.

Table 1: The Potential Coverage of Regulations

Country

Labour force (’000 persons)

Burkina Faso (1990, 1993, 1991) 4167.1 [100.0] Central African R. (1990) 1384.5 [100.0] Côte d’Ivoire (1990) 4598.6 [100.0] Mali (1990, 1991) 2958.7 [100.0] Niger (1990, 1991, 1991) 3619.0 [100.0] Senegal (1988) 2347.6 [100.0] Togo (1990, 1991, 1991) 1396.3 [100.0]

Formal sector employment (’000 persons) Private Public

35.6 [0.9] 13.0 [0.9] 131.5 [2.8] 15.8 [0.4] 79.6 [3.4] 22.4 [1.6]

34.2 [0.8] 116.0 [2.5] 40.2 [1.4] 39.7 [1.1] 66.3 [2.8] 33.7 [2.4]

Figures in parentheses indicate the years the data refer to, while figures in brackets are percentages of the labour force. The formal private sector may include state-owned enterprises in some countries. Data on the labour force for all countries, except Senegal, are from the International Labour Office. Data on private sector employment in Côte d’Ivoire are from the Banque de Données Financières. In Senegal, they are from Banque de Données Economiques et Financières. Other data for Senegal are from the Ministry of Finance. For other countries, data on formal sector employment are from the Statistiques Economiques of BCEAO.


The purpose of this section is to describe the institutional setting of CFA countries in a way that is general enough to fit most of them. Some specific examples are given, particularly for Côte d’Ivoire and Senegal, which are the two biggest economies in the west CFA zone, and the first and third largest in the CFA region as a whole. Among the countries for which data on the potential coverage of regulations could be assembled, Côte d’Ivoire and Senegal are also those with the largest formal sectors as a proportion of their labour force (see Table 1). Since the relevance of labour market policies and institutions should increase with the size of the formal sector, these two countries arguably provide an upper bound when assessing the distortions created by the institutional setting. The colonial administration took a dominant role in the establishment of labour policies for what are nowadays the CFA member countries. The first measures were introduced as early as 1926. A code du travail was passed for all French overseas territories in 1952, partly as a result of a strike organised and led by Guinea’s Sekou Touré with the support of France’s communist labour unions. For most of the period after independence from France, this code served as the basis for labour market regulation at the country level. Government pay policies were partly inherited from the colonial period too. In 1950, a statute known as the second Lamine Gueye law equalised pay between African and French employees of the colonial administration. As a result, at independence the average wage in the Ivorian civil service was more than twice that of an industrial worker (Niessen, 1995). For decades, labour market policies in CFA countries were characterised by four main features: extended reliance on minimum wages; detailed salary grids by occupation and experience for all firms in sectors with collective bargaining; centralised hiring mechanisms; and stringent constraints on firing. Other labour market interventions, such as payroll taxes and mandated benefits, do exist in CFA countries, but they are not excessively onerous compared with the rest of the developing world. For instance, in Côte d’Ivoire workers enrolled with the social security system have to contribute 4% of their salary to the old-age pension programme and an additional 5.5% to the family allowances programme. While contributions to the social security system represent a tax rather than a delayed payment, at 9.5% the tax rate is not particularly out of line compared with other countries. Minimum wages were introduced in most CFA countries in the

480 Martín Rama

1960s, and have been sporadically adjusted upwards ever since. Two different national minimum wages exist: one for urban workers (the SMIG), the other for agricultural labourers in large plantations (the SMAG). But urban workers in sectors with collective bargaining have their minimum wages set by highly detailed salary grids (the salaires minimaux conventionnels). The lowest level in these grids is usually equal to the SMIG, sometimes augmented by a small premium. Other levels vary across occupations and increase with seniority, with the overall structure being relatively stable over time. These grids are negotiated between trade unions and employers’ associations, and have legal force for all workers and firms in the corresponding sector. Collective bargaining outcomes obviously depend on the strength of the labour movement. In CFA countries, trade unions are often numerically small, with the possible exception of Senegal (see Bergen, 1997). Out of the public sector, they tend to be relatively weak. But the ability of the labour movement to affect labour market outcomes should not be judged based on its membership only: the nature of its links with the government and political parties matters too. For a long time, private sector unions were under tight government control in countries like Burkina Faso, Côte d’Ivoire and Gabon, thanks to the co-optation of union leaders by the ruling party. Public sector unions, on the other hand, had a significant impact on government pay policies and were in some cases independent from the ruling party. A third important feature of labour market policies in CFA countries was the legal monopoly the government had over hiring decisions. On paper at least, all vacancies had to be reported to central placement offices which were also supposed to register all job searchers and to make all placements. Private firms were not allowed to bypass these offices and advertise their vacancies, say through the press. Only a few private sector placement agencies existed, usually operating under cover as training providers. Moreover, the government offices tended to be quite inefficient. For instance, the Office de la Main-d’Oeuvre de la Côte d’Ivoire (OMOCI), with 175 employees in its payroll, only managed to make an average of 3,440 placements per year in 1988–91. In Togo, there were a meager 152 placements per year during the same period, which represents roughly 0.3% of total employment in the formal sector of the economy (Runner, 1992). Countries in the CFA zone were also characterised by significant firing costs. At first glance, these costs may not look excessively high by international standards. In Côte d’Ivoire, for example, severance


pay was set at 40% of the monthly wage for each of the first 5 years of service, and the percentage declined for subsequent years. Even counting the advance-notice compensation, which could amount to 3 months of salary, total severance pay remained within the standard range for countries at that level of development. The problem is that a worker ‘abusively’ fired had the right to an indemnity (dommages et intérêts) paid by the employer — and any lay-off for economic reasons used to be considered ‘abusive’ by the tribunals, which set this indemnity at the equivalent of several annual salaries. The labour market policies of CFA countries have experienced significant changes in recent years. The depth of these changes varies across countries, and is still negligible in some, but there is a clear trend towards greater flexibility in all of the four areas discussed above. In Cameroon, the enforcement of labour laws was deliberately weakened after 1982 (Barba Navaretti et al., 1996). In Côte d’Ivoire, the emergence of a new trade union confederation in 1990, reflecting a more general drive towards democracy, introduced competition in the collective bargaining process. Also, the suppression of the government monopoly over hiring decisions, in 1991, allowed private sector firms in the formal sector to recruit by themselves or, more accurately, to do it openly. In Senegal, the possibility of renewing temporary contracts over many years, introduced in 1987 and extended in 1989, implicitly put a cap on firing costs for newly hired workers. In all three countries, a gradual recognition that economic lay-offs should not be treated in the same way as ‘abusive’ ones has significantly reduced the cost of restructuring. It is also important to distinguish between regulation and practice. Labour market policies which look distortive on paper may not be very harmful if the administrative capability to enforce them is weak (see Squire and Suthiwart-Narueput, 1997). The small number of placements made by the central government offices in charge of hiring indicates that most of the recruitment was actually handled by the private sector. In the same vein, it is worth noting that most of the time and effort of labour inspectors was usually devoted to solving individual labour conflicts. Consequently, very few plants were visited to check that they did comply with labour standards. In the unlikely event of one such visit, some employers add, inspectors could be easily bribed to avoid paying a fine. Not surprisingly, employers did not feel excessively constrained by labour regulations. In a survey of some 200 manufacturing firms con-

482 Martín Rama

ducted in Cameroon, in 1994, only 2% of the interviewees considered that labour regulations were a large or severe problem, compared with 85% who answered that the problem was slight or non-existent. The responses were 3 and 75% respectively for wage costs, 4 and 85% for rules regarding lay-offs, and 5 and 63% for the cost of lay-offs. Differences in these responses across plant sizes and sectors of activity were relatively minor, except for the cost of lay-offs. More generally, labour issues were ranked as the least acute problem faced by firms, after taxes, corruption, price controls, government rules and difficulties in obtaining licenses (Gauthier, 1995). This sanguine assessment could be due to the fact that labour market reforms were deeper in Cameroon than elsewhere in the CFA region. However, a similar picture emerges from studies done for other countries. In Senegal, before any deregulation of the labour market had actually taken place, most managers viewed labour market regulations more as a nuisance than a constraint (Terrell and Svejnar, 1989). In Côte d’Ivoire, the most stringent labour regulations were reportedly bypassed by firms using alternatives such as subcontracting and apprenticeship (World Bank, 1993). 3. The Extent of Wage Misalignment 3.1 Compared with Other Countries One way to assess the extent of wage misalignment in CFA countries is to compare their average wages with those of other countries at a similar development level. Controlling for the development level is key in this respect. The salaried relationship is characteristic of the formal sector of the economy. Labour productivity and earnings tend to be much lower in agriculture, and for the urban self-employed. As a result, the ratio of average wages (mainly formal) to average productivity (for the economy as a whole) should be larger in poorer countries, even if wages were not especially high. Several cross-country comparisons are presented in Table 2. The wage figures used to construct this table are from a cross-country database of labour market indicators currently in preparation at the Development Research Group of the World Bank (Rama and Artecona, 2000). This database includes information on average labour costs in manufacturing for more than 100 countries, including several from the CFA zone. Most of the figures are from plant-level surveys covering

Wage Misalignment in CFA Countries 483 Table 2: Labour Costs Across Countries

Dependent variable: labour costs in manufacturing, in % of per capita GDP (B) (C) (D)

Explanatory variables

(A)

CFA countries (dummy variable) Per capita GDP (’000 PPP dollars) Per capita GDP squared (million PPP dollars) Urban population (% of the total) Mean years of schooling for ages 25 and above Labour force in manufacturing (% of the total) Countries in Latin America and the Caribbean (dummy variable) Sub–Saharan African countries (dummy variable) Independent term

582.3** (5.097) –72.27** (–4.843) 2.820** (3.938)

No. of observations R2 F test

510.6** (8.512) 118 0.559 42.6

533.6** (4.038) –62.59** (–5.152) 2.460** (4.172)

522.1** (4.395) –71.88** (–4.663) 3.350** (4.547) 1.260 (1.183) –28.47** (–3.052) –0.041** (–4.251)

29.26 (0.649)

484.7** (3.541) –68.23** (–4.288) 3.120** (4.598) 1.578 (1.271) –25.10** (–3.133) –0.038** (–4.328) 0.594 (0.015)

98.80 (1.192) 450.0** (8.159) 118 0.568 31.1

77.38 (0.921) 500.43** (8.394) 109 0.629 42.5

557.79** (7.718) 109 0.624 43.7

Estimated by ordinary least squares (OLS), using White heteroskedasticitycorrected estimators. The data are for period 1985–93. Values in parentheses are t-statistics. Significant coefficients at the 1% level are indicated by two asterisks.

large establishments. Labour costs are calculated as the ratio of the total payroll to paid employment. The payroll usually includes bonuses, allowances and social security contributions by both employers and employees. Data are for the entire decade preceding the CFA franc devaluation, but in many countries there are only a few annual observations available. For the regression analyses, the average labour cost in manufacturing was divided by the contemporary output per capita, with the latter variable being an indicator of the average

484 Martín Rama

productivity of labour. Because both the numerator (labour costs) and the denominator (output per capita) are measured in the same currency, the 1994 devaluation does not directly affect their ratio. Two key explanatory variables in the analysis are output per capita measured in dollars of comparable purchasing power (PPP) and its square. The quadratic term allows for a non-linear relationship between the dependent variable and economic development. Other explanatory variables are the share of urban population and the mean years of schooling of the adult population. Although these two variables are correlated with the level of development, they have proven significant when explaining wage variation across countries (see Freeman, 1994). The share of the labour force in manufacturing is also included among the explanatory variables. This share can be seen as a proxy for the relative size of the modern sector of the economy. Based on the ‘new’ growth literature, the possibility that regional factors play a role is considered as well — hence the dummy variables for Latin American and Sub-Saharan African countries (see Barro, 1991). The econometric results reported in Table 2 show a convex relationship between the ratio of labour costs to labour productivity and the level of development.3 This relationship is downward sloping at low levels of development. It reaches a minimum for industrial countries, where the average labour cost is comparable to per capita output. Also, the ratio of labour costs to labour productivity is larger, ceteris paribus, in countries with low levels of education and small manufacturing sectors. It appears to be larger in countries with a large urban population as well, but the corresponding coefficient is not statistically significant. Similarly, labour costs seem higher in Latin American and Sub-Saharan African countries than in other areas, but the associated coefficients are not significant either. From the point of view of this study, the most striking result in Table 2 concerns the dummy variable for countries in the CFA region. Whatever the specification chosen for the regression, the coefficient multiplying this dummy variable is positive and statistically significant. Its order of magnitude remains unaffected, at around 500%, regardless of the set of control variables included in the regression. At 3 Following a standard practice in cross-country growth regressions, it is assumed that the variance of the error terms in the regressions can vary from country to country. Hence the use of White-heteroskedasticity-corrected estimators.


the level of development of CFA countries, the predicted level of average labour cost in manufacturing is about six times per capita output. It follows that average labour costs in CFA countries were almost twice as high as in other countries at a similar level of development. This finding can be considered as evidence of wage misalignment. 3.2 Compared with Non-wage Labour Earnings A different way to assess whether wages were particularly high in CFA countries is to compare them with non-wage labour earnings within the same countries. This kind of comparison requires information on individual earnings, as well as on individual characteristics such as education, work experience and other measures of human capital. These control variables are needed to make sure that comparisons are made across similar workers. The standard practice in this respect is to estimate Mincerian equations, where the explained variable is the log of labour earnings and the explanatory variables are individual characteristics like those listed above. The analysis is completed by splitting the sample between workers in the formal sector, who are covered by labour market policies and institutions, and workers in the informal sector, who are not.4 The criteria used to identify the formal sector vary across studies. In some cases, the sample is split according to the size of the establishments individuals work in, the hypothesis being that large establishments are more likely to be subject to inspection. But more elaborate criteria are considered too. For instance, the study on Côte d’Ivoire by Vijverberg and van der Gaag (1990) applies factor analysis, involving 15 job attributes, to data from the 1985 Living Standards Survey (CILSS). The first component of these data is then used as an indicator of the formality of jobs. This component turns out to be highly correlated with six of the job attributes. It is higher when the wage is subject to minimum wage legislation, when there is a signed contract, when the worker receives paid holidays, when he or she is eligible for paid sick leave, when there is a retirement plan and when the worker receives social security benefits. 4

Some of the studies also correct for the self-selection of workers into the different sectors. They take advantage of the fact that workers in the formal and informal sectors have different observable characteristics to make inferences about other, unobservable characteristics that may affect their earnings. Ignoring these other, unobservable characteristics could bias the coefficients associated to the observable ones.

486 Martín Rama

A different approach is adopted by Miller and Vallée (1995) in the case of Cameroon. Using a plant-level-based survey of workers produced by the Regional Program on Enterprise Development (RPED), these authors consider three sectors: formal regulated, formal unregulated and informal. Firms in the formal regulated sector are characterised by multiple ownership, by some form of public ownership or by having complained about being constrained by labour regulations. At the other end, firms in the informal sector are characterised by individual ownership, and by either having apprentices or carrying no formal accounts. All other firms are classified as being formal but unregulated. This criterion is thus different from the one used by Vijverberg and van der Gaag. However, the divide between the formal and informal sectors would probably be similar if both were applied to the same sample. The results of a series of studies done along these lines for selected CFA countries are reported in Table 3. The figures in this table indicate the percentage change in earnings that an individual with ‘average’ characteristics would experience if he or she was transferred from the informal to the formal sector. The figures are based on the preferred estimate in each of the studies. The list of studies considered in Table 3 is certainly incomplete. For instance, a careful analysis of the labour market in Côte d’Ivoire, by Appleton et al. (1990), was not included, because it does not provide an estimate of the wage premium for formal sector jobs. It could also be argued that the samples of some of the studies that were included in Table 3 are too small to make any reliable inference. However, the paucity of empirical analyses in this particular region of the world makes the results reported in this table interesting on their own. These results provide further evidence of wage misalignment in CFA countries. With the possible exception of Mali, the gap between formal and informal sector earnings is high by international standards. Of course, the gap can be expected to be positive even in countries with relatively flexible labour markets. Formal sector firms may voluntarily pay higher wages as a way to attract better workers, to reduce the turnover rate, to boost morale or to elicit higher effort levels. Informal sector firms, by contrast, are less likely to use ‘efficiency’ wages of this sort. However, the gap between formal and informal sector earnings seldom exceeds 30%, even in countries where the distortions created by labour market policies are agreed to be large

Wage Misalignment in CFA Countries 487 Table 3: Wage Premium for Formal Sector Jobs

Study

Estimation technique

Burkina Faso

Cameroon

Berthélémy and OLS with dummy Bourguignon variable for formal (1992, p. 98, sector jobs table 4.4)

Côte d’Ivoire

Mali

203.7% [CILSS, 1985; n = 311]

Lachaud (1993b, p. 54, table D)

OLS with dummies for covered and uncovered wage jobs

Miller and Vallée (1995, p. 176, table 6.11)

OLS with sample split between formal and informal firms

70.2% [RPED, 1994; n = 517]

Vallée and Thomas (1994, p. 170, table 5.12)

OLS with sample split between regulated and non-regulated firms

34.3% [RPED, 1993; n = 740]

Vijverberg and van der Gaag (1993b, p. 34, table 6.2)

OLS with factor analysis for job coverage

57.1% 60.3% 40.9% 9.6% [IIES, 1990–1; [IIES, 1990–1; [IIES, 1986–7; [IIES, 1991; n = 219] n = 278] n = 244] n = 202]

65.2% [CILSS, 1985; n = 295]

All studies control for individual characteristics, but the set of controls varies across studies. The wage premium is based on the coefficient (c) multiplying the formal sector dummy variable. In the case of Lachaud (1993b), it is based on the difference between coefficients for covered and uncovered wage jobs. In the case of Vijverberg and van der Gaag (1990), the formal sector dummy is the first of the principal components of a set of job characteristics. For large values of coefficient c, in absolute terms, the per cent premium is approximated as 100[exp(c) – 1]. Data sources and sample sizes are reported in brackets. CILSS is the Côte d’Ivoire Living Standards Survey, a household survey with national coverage. IIES is a series of labour force surveys carried out by the Institut International d’Etudes Sociales in the capital cities. RPED is the Regional Programme for Enterprise Development, a plant-level survey which includes data on up to ten workers in each firm.

488 Martín Rama

[see, e.g., MacIsaac and Rama (1997) on Ecuador]. In CFA countries, by contrast, earnings gaps of 60% or more are not uncommon. 4. The Nature of Wage Rigidity 4.1 Real Rigidity The effect of a devaluation on economic activity depends crucially on the degree of wage indexation. Numerical simulations for the CFA zone suggest that the long-run effect could even be negative if real wages were rigid in the sense of following consumer prices closely (Bourguignon et al., 1995). The link between wages and other nominal variables, including consumer prices, is evaluated in this section using industry-level data for Côte d’Ivoire and Senegal over a period of two decades.5 Data are from the records of formal sector firms kept by the governments of these two countries, known as the Banques de Données Financières (BDFs). BDFs include information on the total wage bill and employment, organised according to various breakdowns. Although all firms in the two countries should in principle report to the BDFs, in practice only formal sector firms do so, and not in all years. The result is an unbalanced panel of roughly 2,000 firms in Côte d’Ivoire and 1,000 firms in Senegal. This section focuses on the determinants of changes in average labour costs in the formal sector firms covered by the BDFs. Labour costs include wages and salaries, contributions to social security by both workers and employers, and training expenses. The data do not allow the disaggregation of labour costs into these three components. On the other hand, the BDFs do report separate information on labour costs for permanent and seasonal workers, excluding training expenses, as well as separate information on permanent and seasonal employment. Unfortunately, the allocation of labour costs and workers to these separate categories does not appear to be reliable, as the resulting fluctuations in labour costs for permanent and seasonal workers are extremely erratic — hence the focus on average labour costs for all workers. The data are for period 1976–94, although the lag 5 There is an assessment of how labour costs reacted to the 1994 devaluation of the CFA Franc in Cameroon. Using plant-level data, Barba Navaretti et al. (1996) showed that the median wage fell by 11% compared with output prices between 1992–3 and 1994–5, whereas the average wage increased by about 13%. The assessment is thus inconclusive, which is not surprising given that it was carried out almost immediately after the devaluation.


structure used in the regression analysis substantially reduces the number of observations available. A simple specification is considered to assess how labour costs react to changes in other key nominal variables of the economy. The chosen dependent variable is the annual change in the average labour cost per worker, at the industry level. There are 53 such industries in Côte d’Ivoire and 35 in Senegal. The key explanatory variables are the annual inflation rate, the annual increase of the minimum wage (SMIG) and the annual variation of average government wages. The annual output growth rate (in real terms) and the annual change in the level of the terms of trade are used as controls; in the case of Senegal they are both lagged one year. All the explanatory variables are defined at the economy-wide level. They can be considered exogenous, as each of the industries is too small to affect macroeconomic aggregates. Because labour costs may not fully adjust within a year, the specification also includes a lagged value of the dependent variable.6 The results, reported in Tables 4 and 5, suggest that average labour costs followed closely the fluctuations of other nominal aggregates. Admittedly, each of the two sets of regressions has its own weaknesses. In the case of Côte d’Ivoire, the impact of changes in the minimum wage could not be assessed, as the SMIG remained constant over most of the period. Due to the lag structure of the regressions, the level of the SMIG was in fact the same for all the available observations. In the case of Senegal, regression tests indicate potential problems with the chosen instruments. Attempts to use different lags of the endogenous variable as instruments did not improve the results. Despite these unrelated shortcomings, the results in Tables 4 and 5 are quite similar. In both countries, the key explanatory variables have a significant impact on labour costs when they are considered separately. The individual significance of some of them drops when they are entered 6

If the error terms of the regression are serially correlated, the introduction of a lagged value of the endogenous variable among the right-hand-side variables may bias the estimates. Serial correlation implies that the error term in a specific sector and year is not independent from the corresponding error term in the previous year. The latter, in turn, was one of the determinants of the change in labour costs in the previous year. It follows that one of the right-hand-side variables (the lagged dependent variable) is by construction correlated with the error term, thus violating one of the assumptions of the linear regression model. The preferred approach to deal with this problem is the Generalised Method of Moments estimator for dynamic models of panel data introduced by Arellano and Bover (1995) and Blundell and Bond (1998). The results reported in this section are based on this estimator.

490 Martín Rama Table 4: Determinants of Increases in Labour Costs in Côte d’Ivoire


Change in the log of consumer prices Change in the log of average government wages Lagged change in the log of average labour costs Change in the log of output in real terms Change in the log of the terms-of-trade index Independent term Sargan test (p value) 2nd-order correl. (p value) 3rd-order correl. (p value) No. of observations

Dependent variable: change in the log of average labour costs per worker in formal sector firms (A) (B) (C)

0.4080* (1.874)

–0.2064*** (–3.349) –0.2248 (–0.588) 0.2616*** (3.058) 0.0251** (2.434) 0.393 0.110 0.576 505

0.3947*** (2.823) –0.1768*** (–2.880) –0.5409 (–1.420) 0.2947*** (4.082) 0.0490*** (7.685) 0.891 0.498 0.737 505

0.0936 (0.322) 0.2530 (1.423) –0.2224*** (–3.939) –0.2471 (–0.600) 0.2244*** (2.868) 0.0415*** (2.730) 0.400 0.167 0.678 505

Estimated using the Generalised Method of Moments, with the second and third lags of the endogenous variable as instruments. Data on average labour costs per worker are from the Banque de Données Financières, which is composed of about 2000 medium- and large-size firms. There is one observation per industry per year, over the period 1976–94. Changes in the log of output and in the log of the terms-of-trade index are for the same year. Values in parentheses are t-statistics. Significant coefficients at the 10, 5 and 1% levels are indicated by one, two and three asterisks respectively.

jointly in the regression, but this is probably due to the high correlation between them. Moreover, the magnitude of the coefficients multiplying the key explanatory variables is quite large. Finally, in both countries, the coefficient multiplying the lagged dependent variable is significantly negative, which could reflect a strongly convergent dynamic process. However, this result could also be due to substantial measurement error in the dependent variable. If this is correct, the negative coefficient would just reflect the fact that abnormally high

Wage Misalignment in CFA Countries 491 Table 5: Determinants of Increases in Labour Costs in Senegal


Dependent variable: change in the log of average labour costs per worker in formal sector firms (A) (B) (C) (D)

Change in the log of 1.1595*** 1.2362*** consumer prices (5.524) (6.668) Change in the log of 0.5155** 0.2265 minimum wages (2.164) (1.181) Change in the log of average 0.3193** 0.2788*** government wages (2.487) (2.671) Lagged change in the log of –0.3316*** –0.3229*** –0.2758*** –0.3303*** average labour costs (–14.576) (–13.766) (–12.077) (–14.444) Change in the log of output in 3.2730*** 1.8936*** 1.4473** 3.7511*** real terms (4.183) (3.157) (2.458) (5.866) Change in the log of the 1.9589*** 1.1460*** 1.5918*** 2.185*** terms-of-trade index (5.394) (3.7550) (5.837) (6.673) Independent term –0.0897*** –0.0167 –0.0089 –0.1184*** (–3.106) (–0.923) (–0.510) (–5.020) Sargan test (p value) 0.002 0.009 0.000 0.005 2nd-order correl. (p value) 0.218 0.469 0.053 0.144 3rd-order correl. (p value) 0.100 0.334 0.336 0.235 No. of observations 378 378 378 378

Estimated using the Generalised Method of Moments, with the second and third lags of the endogenous variable as instruments. Data on average labour costs per worker are from the Banque de Données Economiques et Financières, which is composed of about 1000 medium- and large-size firms. There is one observation per industry per year, over the period 1976–94. Changes in the log of output and in the log of the terms-of-trade index are for the previous year. Values in parentheses are t-statistics. Significant coefficients at the 10, 5 and 1% level are indicated by one, two and three asterisks respectively.

or low values of the dependent variable in one year are in general followed by more plausible values the year after. In spite of their indexation to other nominal variables, labour costs seem to be responsive to real shocks, as captured by the economywide fluctuations in output and the terms of trade. The coefficients multiplying the change in the terms-of-trade index are positive and

492 Martín Rama

significant in the case of Côte d’Ivoire. Moreover, their absolute size is plausible, given the degree of openness of the economy. In the case of Senegal, the coefficients multiplying both the change in output and the change in the terms-of-trade index are positive and significant. However, their absolute size appears to be too large to be credible. This result might be due to the lack of appropriate instruments to estimate the regressions in Table 5. There is independent evidence to suggest that wages are responsive to real shocks. If data from a representative household survey are used, instead of data from a sample of formal sector firms, a negative association between urban wages and local unemployment rates can be found. This is shown by Hoddinott (1996) using data from the 1985, 1986 and 1987 rounds of the CILSS. His results indicate that the negative influence of unemployment on labour earnings is stronger in the case of young workers and weaker in the case of public sector employees. Overall, and after controlling for a number of potentially serious econometric problems, most notably unobserved locationspecific effects, Hoddinott estimates the wage–unemployment elasticity of Côte d’Ivoire to be –0.12. This is similar to the –0.1 elasticity reported by Blanchflower and Oswald (1994) for the USA, the UK and other industrial countries. 4.2 Nominal Rigidity The case to devalue the CFA franc rested in part on the premise that formal sector wages were nominally rigid. The conventional wisdom is that the adverse shocks of the 1980s led to a drop in aggregate demand, hence creating a substantial deflationary pressure. These pressure was reinforced by an adjustment strategy that rested, allegedly, on lower public expenditures and higher taxes. While the earnings of those who were not protected by labour market regulations declined, the story goes, workers in the formal sector succeeded in preventing their wages from falling. Labour costs in formal sector firms thus failed to match the decline in aggregate demand. While this interpretation of the economic developments of the 1980s and early 1990s is plausible, there is little empirical evidence to support it. If the conventional wisdom was right, over the 1980s and early 1990s there should have been a decline of labour earnings in the informal sector and a relative stability of wages in the formal sector. Moreover, if the downward rigidity resulted, as it is claimed, from labour market


policies and institutions, trends in formal sector wages should have followed closely those of the SMIG. This is because the SMIG not only represents the wage of those who make the bare legal minimum: it also provides the ‘floor’ for the salary grids negotiated between trade unions and employers. However, none of these two predictions holds true in the cases of Côte d’Ivoire and Senegal, which are two of the countries with the largest formal sectors in the region. The conventional wisdom is rejected here based on data from the consumer price indexes of Côte d’Ivoire and Senegal. While the BDFs of these two countries provide information on average labour costs in their formal sectors, there is no similar, readily available data source for labour earnings in the informal sector. However, the consumption bundles used to calculate consumer price indexes include a few personal services whose production involves little more than labour. To the extent that these services are provided by individuals or microenterprises, they are not likely to be directly affected by labour market regulations. Fluctuations in their prices can therefore be expected to follow closely those of labour earnings in the informal sector. To minimise the possibility of a measurement bias, the focus of the empirical analysis is on services whose prices are defined based on units of output, rather than on the corresponding labour inputs. When labour inputs are considered, there is a risk that the statistical offices register some specific salaire minimum conventionnel, rather than the actual labour cost. This is what happens in practice with the prices of domestic service and cooking in the consumer price index of Côte d’Ivoire. Moreover, even if the statistical offices did gather the data from the actual providers of these services, the latter could declare fake wages to hide non-compliance with labour regulations. However, there should be no such biases when the unit of measurement is the price of a haircut, or the cost of a laundry service for a double-bed sheet. To the extent that no legal minimum price exists for these services, there are no incentives to report fake prices either. These prices are taken from the European consumer price index, which goes farther back in time than the African one. The resulting indexes for labour earnings in the informal sectors of Côte d’Ivoire and Senegal show that there was no deflationary trend over the 1980s and early 1990s. These indexes are represented in Figures 1 and 2 respectively.7 In Côte d’Ivoire, labour earnings in the 7

The informal sector series are non-weighted averages of the indexes for haircuts

494 Martín Rama Figure 1: Nominal Labour Earnings in Côte d’Ivoire

Indices with base 100.0 in 1980. Minimum wages and informal labour earnings are as of January, whereas formal labour earnings are annual averages.

informal sector increased roughly at the same pace as labour costs in the formal sector, at least until the end of the 1980s. In Senegal, they increased at a much faster pace. Figures 1 and 2 also show that both labour earnings in the informal sector and labour costs in the formal sector increased far faster than the SMIG. The plots in Figures 1 and 2 do not imply that formal sector wages are flexible downwards, but and laundries. Each of these two, in turn, is an non-weighted average of the indexes for all of the items available in the European consumption bundle of the two countries. In Côte d’Ivoire, the price index for haircuts is based on items 41011 and 41012, whereas the price index for laundries is based on items 31018, 31019 and 31020. In Senegal, items 118 and 119 were used for haircuts, and item 148 for laundries. Unfortunately, the disaggregated data were not available for Côte d’Ivoire in 1989–91.

Wage Misalignment in CFA Countries 495 Figure 2: Nominal Labour Earnings in Senegal

Indices with base 100.0 in 1980. Minimum wages and informal labour earnings are as of January, whereas formal labour earnings are annual averages.

they do suggest that downward rigidity, if it does exist, did not play a role in the series of events leading to the 1994 devaluation of the CFA franc. In particular, they reject the hypothesis that labour market policies and institutions were the obstacles preventing wages in the formal sector from adjusting to a more unfavorable international context. There is, in fact, some evidence that nominal wages are flexible downwards, at least in Côte d’Ivoire. The reason why this flexibility may not be apparent is because analyses based on averages tend to neglect composition effects. In a study using data collected at the plant level, Levy and Newman (1989) showed that average wages in the formal sector increased between 1979 and 1984, but that wages for well-specified classes of labour actually decreased. This discrepancy was due to the change in the composition of the labour force, which was characterised by greater education, training and experience in 1984 than in 1979.

496 Martín Rama

5. The Level of Minimum Wages 5.1 Compared with Other Countries Having established that wage misalignment was substantial in CFA countries over the late 1980s and early 1990s, the issue is whether labour market policies and institutions played a role in explaining this outcome. Put differently, it is necessary to understand why private sector firms paid their workers so much in excess of their alternative earnings. One obvious suspect in this respect is, of course, the minimum wage policy characterising CFA countries. This is because the SMIG affects not only the earnings of those who make the bare legal minimum, but also the ‘floor ’ of the salary grids negotiated between trade unions and employers, hence earnings in all sectors covered by collective bargaining. From an international perspective, however, the minimum wages of CFA countries are not high enough to account for the observed misalignment of their labour costs. Table 6 replicates the international comparison carried out in Section 3, with the explained variable being now the ratio of minimum wages (rather than average labour costs) to output per capita. The sample of countries is somewhat smaller than in section 3, because there are fewer data on minimum wages in the database of Rama and Artecona (2000). As before, the minimum wage figures are averages over the decade up to the devaluation of the CFA franc, but for some countries there are very few annual observations available.8 As in the case of average wages, minimum wages tend to represent a higher multiple of labour productivity in poorer countries. Not surprisingly, the coefficients multiplying income per capita have the same sign as those estimated in Table 2 (when dealing with average labour costs), although they tend to be less significant. On the other hand, the coefficients multiplying the urban share of the population, the median schooling of adults and the share of manufacturing in the labour force are all statistically insignificant, regardless of the specification. The most interesting feature of Table 6 is the positive and statistically significant coefficient multiplying the dummy variable for 8

A few countries (e.g. the UK) did not have a national minimum wage during this period. The minimum wage is set equal to zero in their case. Strictly speaking, therefore, the dependent variable in Table 6 is censored. However, the number of countries with no national minimum wage policy is small enough to justify the use of ordinary least squares in the estimation.

Wage Misalignment in CFA Countries 497 Table 6: Minimum Wages Across Countries


Dependent variable: minimum wage, in % of per capita GDP (A) (B) (C) (D)

CFA countries 127.5** (dummy variable) (4.523) Per capita GDP –8.869* (’000 PPP dollars) (–2.003) Per capita GDP squared 0.302 (million PPP dollars) (0.134) Urban population (% of the total) Mean years of schooling for ages 25 and above Labour force in manufacturing (% of the total) Countries in Latin America and the Caribbean (dummy variable) Sub-Saharan African countries (dummy variable) Independent term 90.0** (5.073) No. of observations 81 0.533 R2 F test 19.5

168.3** (6.121) –12.579** (–2.710) 0.461* (2.216)

4.326 (0.314) –53.96** (–3.031) 106.2** (5.109) 81 0.583 12.5

122.4** (4.224) –3.715 (–0.648) 0.149 (0.588) –0.435 (–1.108) –1.906 (–0.689) –0.001 (–0.297)

104.1** (4.308) 75 0.570 13.5

160.7** (6.239) –4.953 (–0.835) 0.257 (1.006) –0.593 (–1.370) –4.223 (–1.391) –0.002 (–0.654) 12.243 (0.748) –57.53** (–3.340) 130.4** (5.110) 75 0.588 11.1

Computed using White heteroskedasticity-corrected estimators. The data are for period 1985–93. Values in parentheses are t-statistics. Significant coefficients at the 5 and 1% level are indicated by one and two asterisks respectively.

CFA countries. As before, the estimated effect does not vary much with the set of explanatory variables included in the regression. Although the coefficients in columns (B) and (D) seem to be some 40 percentage points higher than those in columns (A) and (C), this is due to the inclusion of a Sub-Saharan African dummy in the specification. The coefficient on this variable is significantly negative, at around –50 percentage points. Since CFA countries also belong to the Sub-Saharan African region, the net gap with other developing countries is still

498 Martín Rama

around 120 percentage points, which is roughly the same as in columns (A) and (C). According to the regressions in Table 6, countries with the development level of the CFA region are expected to have minimum wages about twice as high as their per capita output. Observed minimum wages in these countries are therefore some 50% higher than predicted by the regressions. While such a discrepancy is far from being negligible, it is insufficient to account for the observed wage misalignment. Studies done for other countries reveal a relatively low elasticity of average wages to minimum wages. For instance, in Indonesia this elasticity was estimated at 0.1, in spite of substantial government efforts to enforce minimum wages (see Rama, 1996). The results in Table 5 suggest a higher elasticity in the case of the formal sector of Senegal, maybe in excess of 0.2.9 However, the results in this table should not be interpreted literally, as was already pointed out in the previous section. Using 0.1 as a benchmark, a 50% misalignment of minimum wages would translate into a 5% misalignment of average wages, which is a far cry from the almost 100% discrepancy observed in practice. 5.2 As a Function of Macroeconomic Conditions The minimum wages of most CFA countries declined, in real terms, during the 1980s and early 1990s. Over period 1977–92, the annual rates of variation were –0.65% in Burkina Faso, –0.77% in Benin, –0.84% in Senegal, –1.31% in Mali, –1.88% in Togo and –2.83% in Côte d’Ivoire. In most cases, the minimum wage declined at a faster pace than per capita output. Only in Niger did the minimum wage increase in real terms, at an annual rate of 3.21%. The observed decline of minimum wages in real terms suggests that policy makers in CFA countries were responsive to the overall deterioration in macroeconomic conditions. In general, governments can be expected to be more prone to raise minimum wages in ‘good’ times, and to freeze them in ‘bad’ times. As long as there is some inflation, freezing the minimum wage provides a mechanism to reduce the ‘floor ’ of formal sector wages. 9

This figure takes into account the dynamics of labour costs, as captured by the coefficient multiplying the lagged endogenous variable. If the estimates in column (B) of Table 5 are taken literally, a 1% increase in the minimum wage leads to a 0.52% increase in labour costs within the year, followed by a decline the year after. In the long run the net impact of the minimum wage increase is 0.52/1.32 ≈ 0.39. The estimates in column (D) yield a net impact of roughly 0.17.


This possible endogeneity of labour market policies in developing countries has been analysed at the conceptual level in several opportunities (see, e.g., Freeman, 1993; Rama and Tabellini, 1998). Countries in the CFA region provide an opportunity to test it at the empirical level. The simplest way to implement an empirical test of the endogeneity of minimum wages is to treat their annual change as a censored variable, partially reflecting the change of a latent variable. The implicit assumption is that the optimal change in the minimum wage, from the viewpoint of policy makers, is some stable function of macroeconomic variables such as the changes in aggregate prices, in the level of economic activity and in the terms of trade. These variables evolve over time, and so does the optimal change in the minimum wage. However, social and political constraints imply that the minimum wage cannot be cut in nominal terms. Given these constraints, the actual minimum wage would increase only when its optimal change is positive. When it is negative, in turn, the actual minimum wage would remain frozen in nominal terms, as was the case in CFA countries in most of the years preceding the 1994 devaluation. This simple model is applied to the annual changes of the SMIG in all of the CFA countries for which data are available, before the 1994 devaluation. In practice, this amounts to using a sample with slightly less than 200 observations, most of which are zeroes. The latent variable (i.e., the optimal change in the nominal SMIG, from the policy makers’ perspective) is assumed to be a function of annual changes in consumer prices, in economic activity and at the terms-of-trade level. All three macroeconomic variables are measured at the country level and included in the regression with a lag of at least one year, to avoid simultaneity problems. The hypothesis to be tested is that these macroeconomic variables have a positive impact on the latent variable, hence indirectly on the level of the SMIG. The results obtained when estimating this model, reported in Table 7, indicate that minimum wages were highly responsive to changes in macroeconomic conditions. Table 7 includes several columns with different specifications, to address the problems raised by the estimation of a censored-variable model with panel data.10 The results which 10

The main problem is how to deal with country-specific disturbances. In a linear regression, the standard solution is to use the fixed effects approach, which entails replacing the explained variable by its deviation with respect to the country mean. But in a censored-variable regression the latent variable is not always observable,

500 Martín Rama Table 7: Minimum Wage Endogeneity


Dependent variable: annual change in the log of the minimum wage Tobit estimates (A) (B)

Change in the log of terms of 0.3193** trade (lagged 2 years) (2.042) Change in the log of output at 1.0943** constant prices (lagged 1 year) (2.480) Change in the log of consumer 1.1033** prices (lagged 1 year) (3.312) no Country-specific dummy variables Independent term –0.2508** (–5.064) Number of observations 187 Pseudo R2 0.185 Chi-square test 24.49

Dependent variable: annual change in minimum wage (1 if raised, 0 otherwise) logit estimates (C) (D)

0.3177** (2.036) 1.0774** (2.420) 1.0924** (3.228) yes

2.7294** (2.162) 7.3127** (2.211) 9.0364** (3.520) no

2.7325** (2.116) 7.1094** (2.116) 9.0920** (3.440) yes

187 0.207 27.47

–1.8457** (–6.587) 187 0.116 24.91

187 0.134 28.76

All variables are defined at the country level. Values in parentheses are t-statistics. Significant coefficients at the 1% level are indicated by two asterisks. implying that the country mean cannot be calculated. One possible solution is to assume that disturbances follow the same distribution in all countries. This is the assumption underlying the specification in column (A). While it may be objected in the general case, it is plausible here, due to the influence of France on the economic policies of all CFA countries. In practice, this assumption holds true if there is a common policy rule across the CFA zone, with changes in the SMIG at the country level reflecting random deviations from that rule. A different way to implement the fixed effects approach is to include a dummy for each of the countries among the explanatory variables. Computing capabilities rule out this solution when the number of individuals in the panel is large, while the number of periods is small. But the panel considered here involves only a few countries, so that this is a feasible alternative. The results are reported in column (B). Although the implicit assumption in this case is that the policy rule may differ across countries, the similarity between the results in columns (A) and (B) is striking. In both columns, the Tobit estimation method is used. Columns (C) and (D), in turn, apply a methodology first introduced by Chamberlain (1980). The explained variable is now set equal to zero when the minimum wage does not change, and equal to one when it is raised. This transformation of the explained variable allows using the fixed effects approach in non-linear models, such as the censored-variable model. The coefficients in equations (C) and (D) are Logit estimates.


are easiest to interpret are those in columns (A) and (B). Because of the way the units were chosen, the regression coefficients in these columns are in fact elasticities. The units used for the dependent variable in columns (C) and (D) allow the verification of the sign and significance of the results obtained in the other two columns, but they yield coefficient values with no straightforward interpretation. While the levels of minimum wages in CFA countries could be criticised as being too high, the way these levels were adjusted to changes in macroeconomic conditions appears to be sensible. The estimated elasticity of the optimal minimum wage with respect to both aggregate prices and real output is almost 1. The unit elasticities imply that the real minimum wage fully incorporates the effects of real shocks, but is not affected by nominal shocks. Moreover, the elasticity of the optimal minimum wage with respect to the terms-of-trade level is about 0.3, which roughly corresponds to the average openness coefficient of CFA countries. Therefore, the way minimum wages were adjusted to real shocks took into account not only the domestic fluctuations in total productivity, but also the fluctuations stemming from external shocks. The results in Table 7 also indicate that, in the absence of any change in macroeconomic conditions, the real value of the optimal minimum wage would have decreased over time. Had the terms-of-trade level and the level of real output remained constant, the minimum wage would not have been adjusted, even in the presence of relatively high inflation. The independent term in column (A) implies that the minimum wage would have been raised only if prices increased by more than 25%. While this figure should not be taken at face value, it suggests that policy makers in CFA countries not only reacted in a sensible manner to real shocks, but also took advantage of inflation to gradually correct the initial misalignment of their minimum wages. 6. Other Potential Explanations 6.1 Trade Unions Trade union activity is another obvious candidate to explain wage misalignment. Although the labour movement is numerically weak in CFA countries, it is well implanted in the formal sector of the economy. Moreover, it has been traditionally well connected to the political party in power, which in principle could give it more leverage. Unfortunately, there are only a few studies analysing in detail what trade

502 Martín Rama Table 8: Wage Premium for Union Jobs

Study


Miller and Vallée (1995, p. 163, table 6.5)

OLS with dummy variable for union members

Terrell and Svejnar (1989, p. 102, table 7.1)

OLS with dummy variable for union members

Vallée and Thomas (1994, p. 168, table 5.9B)

OLS for formal sector jobs plus Logit for self-selection

Cameroon

Senegal

–8.1% [RPED, 1994; n = 517] –12.5% [TRSV, 1980–5; n = 513] –10.7% [RPED, 1993; n = 740]

All studies control for individual characteristics, but the set of controls varies across studies. The wage premium is based on the coefficient (c) multiplying the dummy variable for union affiliation. For large values of the c coefficients, in absolute terms, the per cent premium is approximated as 100[exp(c) – 1]. Data sources and sample sizes are reported in brackets. TRSV is a sample of seventeen large firms including data on individual workers.

unions do in CFA countries, and they generally adopt a sociological perspective (see, e.g., Ndiaye and Tidjani, 1995). The available economic studies, in turn, are based on quite small samples. Interestingly, they all find that union members earn less than similar, non-unionised wage earners. The available estimates of the union wage premium in CFA countries are reported in Table 8. The figures in this table indicate the percentage change in earnings an individual with ‘average’ characteristics would experience if he or she were transferred from a non-unionised to a unionised activity. These figures are based on Mincerian equations that include a dummy variable for unionisation among their explanatory variables. The fact that union wage premiums are negative suggests that union members may get other, non-wage benefits that compensate for the lower earnings. This is also the case in other countries, while the negative premium is highly atypical from an international perspective. In developing countries,


union members usually make 5–30% more than their non-unionised counterparts. Low or even negative union wage premiums would be consistent with the ‘subordinate’ nature of the labour movement in many CFA countries (see Nelson, 1991). Trade unions were instrumental in implementing the wage moderation policies set up by governments in this region in response to adverse shocks. Although this pattern might have changed as competition between unions developed in the workplace, it would be difficult to make them responsible for the wage misalignment observed in the late 1980s and early 1990s. 6.2 Government Pay Policies Fluctuations in formal sector wages have been shown in Section 4.1 to follow closely those in average government wages. This link is hardly surprising, given that public sector employment represents about one-half of formal sector employment in CFA countries, if not more. Government pay policies therefore emerge as a potential explanation for wage misalignment. The issue is whether government wages are high enough for this explanation to be relevant in practice. This issue can be addressed applying the same methodology that was used to measure the earnings gap between formal and informal sector jobs, or between union and non-union jobs. Note that some of the advantages of working for the government are not reflected in higher wage earnings. These advantages include more job security, better fringe benefits, lower effort levels, and the possibility of moonlighting or taking bribes. Because of these advantages, similar cash earnings for public and private sector employees would imply that the former are better off than the latter. The available empirical studies, summarised in Table 9, show that cash earnings were actually higher for public sector employees than for their private sector counterparts. The figures in this table indicate the percentage change in earnings an individual with ‘average’ characteristics would experience if he or she was transferred from a wage-earning job in the private sector to a government job. The precise definition of the government sector varies across studies, and may or may not include state-owned commercial enterprises. Although the number of studies is quite limited, some of them involve thousands of observations, which increases their credibility. With the exception of Côte d’Ivoire, where the wage premium is negative, public sector

504 Martín Rama Table 9: Wage Premium for Public Sector Jobs

Study


Burkina Faso

Cameroon

Berthélémy OLS with dummy and variable for Bourguignon government jobs (1992, p. 98, table 4.4) Lachaud OLS plus probit (1993a, p. 20, for self-selection table 6) of jobs

Côte d’Ivoire

20.5% [IIES, 1990–1; n = 234]

38.4% [IIES, 1990–1; n = 293]

12.6% [IIES, 1986–7; n = 277] –28.0% [Census, 1988; n = 181,757]

Levy and OLS with dummy Newman variable for public (1989, p. 106, sector jobs table 6)

–17.3% [ONFP, 1984; n = 10,835]

van der Gaag et al. (1989, p. 28, table 5)

FIML with switching regression model

Senegal

–14.3% [CILSS, 1985; n = 311]

Lachaud OLS with dummy (1993c, pp. variable for public 45–6, table B) sector jobs

Miller and OLS with dummy Vallée (1995, variable for 100% p. 163, table private firms 6.5)

Mali

34.8% 19.0% [IIES, 1991; [IIES, 1991; n = 379] n = 243]

11.1% [RPED, 1994; n = 517] –26.5% [CILSS, 1985; n = 513]

Depending on the studies, the public sector may or may not include state-owned enterprises. All studies control for individual characteristics, but the set of controls varies across studies. The wage premium is based on the coefficient (c) multiplying the public sector dummy variable, except in the case of Berthélémy and Bourguignon (1992), where it is based on the difference between coefficients for government and other formal sector jobs. For large values of the c coefficients, in absolute terms, the per cent premium is approximated as 100[exp(c) – 1]. Data sources and sample sizes are reported in brackets. ONFP is a plant-level survey covering large firms, but reporting data on individual workers.

employees thus appear to make around 10–30% more than their private sector counterparts.11 Taking non-wage benefits into account, these studies imply that 11 The Ivorian result was surprising enough to prompt van der Gaag et al. (1989) to use data on moonlighting to check that the estimated earnings differentials made sense. They found that moonlighting was more prevalent among public sector workers than among those in the private sector. They also found that the wage disadvantage of public sector employees was an important determinant of


working for the public sector entails significant rents. Moreover, the comparison in Table 9 is based mainly on wages. It therefore involves formal sector jobs, be these public or private. It was shown above that wages in the private, formal sector of the economy were much higher than other labour earnings. The results reported in Table 9 imply that public sector wages were even higher. This conclusion, together with the finding that wages in the private formal sector follow public sector wages closely, suggests that government pay policies were one of the main forces behind wage misalignment in CFA countries. 6.3 Product Market Distortions Another potential culprit for wage misalignment are the rents created by barriers to competition in product market. When firms have some monopoly power they can behave as price makers, rather than price takers. In this case, the prices they charge for their products involve some mark-up over unit costs, including labour costs, with the markup ratio being larger the higher their monopoly power. Moreover, as long as replacing incumbent workers is costly, the latter can capture some of the rents created by barriers to competition in product markets. The result is a relatively stable relationship between prices and wages on the one hand, and between wages and alternative earnings on the other hand. Countries in the CFA region are said to be characterised by significant monopoly power in product markets. A few firms operate in each industry, quite often under the umbrella of some convention spéciale with the government, involving tax exemptions and other preferential treatments. It would not be surprising if workers in these firms managed to secure a share of the resulting rents, under the form of wages well above other labour earnings in the economy. Note that rent sharing would be feasible even if trade unions did not push for high wages. Substantial firing costs, like those characterising CFA countries during most of the period under consideration, would be enough for barriers to competition in product markets to lead to both high domestic prices and high formal sector wages. Unfortunately, there is no empirical evidence to assess whether limited competition in product markets was one of the reasons for moonlighting. This latter result can be seen as an additional confirmation that the wage premium is negative in Côte d’Ivoire. A more disaggregated version of this analysis can be found in van der Gaag and Vijverberg (1988).

506 Martín Rama

wage misalignment in CFA countries. The standard way to test this hypothesis is to estimate Mincerian equations that include indicators of profitability at the firm or the sectoral level among the control variables. A positive coefficient multiplying these profitability indicators can be interpreted as evidence of rent sharing by workers. The only such estimate available for CFA countries uses the effective protection rate as one of the arguments in the Mincerian equation (Terrell and Svejnar, 1989). The coefficient multiplying this variable turns out to be statistically insignificant, but the sample is small and the effective protection rate is not the most appropriate measure for rents. This single estimate being inconclusive, the claimed link between wage misalignment and limited competition in product markets remains plausible, but it is by no means certain.

7. Conclusion This paper has shown that wage misalignment was substantial in CFA countries over the 1980s and early 1990s. Labour costs in manufacturing were much higher than in other countries with a similar development level, whereas formal sector wages were much higher than labour earnings in the informal sector of the economy. However, the paper has also shown that wage misalignment cannot be traced down to the labour market policies and institutions in force in any obvious manner. Particularly, it has produced evidence that minimum wages did not play a major role in obstructing the adjustment to the adverse external shocks of the 1980s. From a cross-country perspective, minimum wages were not high enough to explain a significant share of the observed misalignment of wages. From a time-series perspective, they decreased in real terms and were adjusted in a sensible manner to changes in the macroeconomic context. Trade unions do not appear to be at the root of wage misalignment either. None of these findings implies that the labour market of CFA countries is flexible. The results in the paper actually suggest that formal sector wages are rigid; but they appear to be rigid in real terms, rather than in monetary terms. Their level fluctuates in line with other key nominal variables, such as average government wages or consumer prices. One explanation for this finding is related to government pay policies. Given how large the public sector is compared with the private formal sector, and given also how high its salaries were in relative terms, government pay policies could have influenced labour


market outcomes in CFA countries. Another plausible, but more hypothetical explanation is the existence of barriers to competition in product markets. Monopoly power in product markets, combined with high costs of replacing incumbent workers, could indeed account for high domestic prices and high formal sector wages, and a stable relationship between the two. More generally, the results in this paper cast doubts on the conventional wisdom regarding adjustment in CFA countries. It seems widely accepted that the main obstacle to a deflationary adjustment was the downward rigidity of formal sector wages. But the analyses for Côte d’Ivoire and Senegal show that both wages in the formal sector and labour earnings in the informal sector increased substantially during the 1980s and early 1990s. If anything, labour earnings in the informal sector increased at an even faster pace than wages in the formal sector. This suggests that the alleged deflationary pressure failed to materialise. Although a detailed analysis of the reasons why this might have happened is clearly beyond the reach of this paper, a quick discussion is warranted to put any recommendations for labour market reform in the proper context. Two complementary explanations can be given for the failure of deflationary pressures to materialise. The first involves the budget institutions of CFA countries. Adverse external shocks may have led to an increase in the budget deficit due to a tax structure relying excessively on foreign trade (Nashashibi and Bazzoni, 1994). Marketing board mechanisms, aimed in principle at stabilising the domestic price of exports, may have played an important role in this respect. Since the ‘stabilisation’ tends to happen at low domestic prices, a US$1 decrease in commodity exports translates automatically into a US$1 decrease in government revenue. The same happens with oil exports. The monetary counterpart to this excess spending by the government can be found in the institutions underlying the CFA franc. In the 1970s, part of the central banking authority regarding the CFA franc was transferred from France to the region. Some relaxation of monetary discipline ensued, as reflected in the increase of the ceiling on outstanding government debt, from 10 to 20% of tax revenue in the previous year. Devarajan and Walton (1994) claim that this and other arrangements allowed significant fiscal laxity. Because all member countries share the same currency, the fiscal excesses of one country potentially spill over to the union as a whole. As shown by Aizenman (1992), the lack of coordination between decision makers, each of

508 Martín Rama

whom can effectively increase the money supply, can be expected to create an inflationary bias. Although this is an oversimplified interpretation of events, it sheds a new light on the relationship between wage misalignment and labour market policies. In this interpretation, both wages and prices in the formal sector of CFA countries were high to begin with, because of government pay policies and, possibly, because of significant barriers to competition in product markets. They became increasingly higher over time due to significant budget deficits, fuelled by the combined effect of ill-designed budgetary and monetary institutions. The policy implications of this interpretation are straightforward. A monetary mechanism imposing more discipline on government expenditures, a tax system relying less on exports, and more competition in product markets would be needed to reduce the risk of overvaluation when facing adverse external shocks. As regards labour market reform, it would certainly be key to improve microeconomic efficiency. Relaxing firing and hiring regulations, as several CFA countries already have already done, would facilitate labour reallocation across sectors and help firms survive when confronted with adverse shocks. Allowing a more decentralised wage bargaining, and not forcing firms and workers who are not represented in the negotiation to be subject to it, would also be welcome. More generally, reforms of the labour market policies and institutions of CFA countries in line with the recommendations of the World Bank (1995) should be beneficial. But it would be naive to see these reforms as a safeguard against wage misalignment if the budgetary and monetary institutions of CFA countries remain unchanged.

References Aizenman, J. (1992) ‘Competitive Externalities and the Optimal Seignorage’, Journal of Money, Credit and Banking, 24 (1): 61–71. Appleton, S., P. Collier and P. Horsnell (1990) ‘Gender, Education and Employment in Côte d’Ivoire’, SDA Working Paper No. 8, Washington DC: World Bank. Arellano, M. and O. Bover (1995) ‘Another Look at the Instrumental Variable Estimation of Error-components Models’, Journal of Econometrics, 68 (1): 29–51. Assane, D. and A. Pourgerami (1994) ‘Monetary Co-operation and


Economic Growth in Africa: Comparative Evidence from the CFAzone Countries’, Journal of Development Studies, 30 (2): 423–42. Barba Navaretti, G., B. Gauthier, J. de Melo and J. Tybout (1996) ‘High Noon in Cameroon?’, unpublished manuscript, Washington DC: World Bank.. Barro, R. (1991) ‘Economic Growth in a Cross-section of Countries’, Quarterly Journal of Economics, 106 (2): 407–43. Bergen, G. (1997) ‘Labor, Democracy and Development in Senegal’, unpublished manuscript, Washington DC: World Bank. Berthélémy, J.C. and F. Bourguignon (1992) ‘Growth and Crisis in Côte d’Ivoire’, unpublished manuscript, Paris: DELTA. Blanchflower, D. and A. Oswald (1994) The Wage Curve, Cambridge MA: The MIT Press. Blundell, R. and S. Bond (1998) ‘Initial Conditions and Moment Restrictions in Dynamic Panel Data Models’, Journal of Econometrics, 87 (1): 115–43. Bourguignon, F., J. de Melo and A. Suwa-Eisenmann (1995) ‘Dévaluation et compétitivité en Côte d’Ivoire’, Revue Economique, 46 (3): 739–49. Chamberlain, G. (1980) ‘Analysis of Covariance with Qualitative Data’, Review of Economic Studies, XLVII: 225–38. Conway, P. and J. Greene (1993) ‘Is Africa Different?’, World Development, 21 (12): 2017–28. Devarajan, S. and D. Rodrik (1991) ‘Do the Benefits of Fixed Exchange Rates Outweigh their Costs? The Franc Zone in Africa’, NBER Working Paper No. 3727, Cambridge MA. Devarajan, S. and M. Walton (1994) ‘Préserver la zone CFA: la coordination macro-économique après la dévaluation’, Revue d’Economie du Développement, 0 (3): 5–30. Freeman, R. (1993) ‘Labor Market Institutions and Policies: Help or Hindrance to Economic Development’, World Bank Economic Review, Proceedings of the 1992 Annual Conference on Development Economics, Washington DC: World Bank. – (1994) ‘A Global Labor Market: Differences in Wages Among Countries in the 1980s’, unpublished manuscript, Washington DC: World Bank. Gauthier, B. (1995) ‘The Business Environment: Regulation and Infrastructure’, in B. Gauthier (ed.), Economic Reform and the Development of the Manufacturing Sector in Cameroon, RPED Country Study Series, Washington DC: World Bank.

510 Martín Rama

Hoddinott, J. (1996) ‘Wages and Unemployment in an Urban African Labour Market’, Economic Journal, 106 (439): 1610–26. Horton, S., R. Kanbur and D. Mazumdar (eds) (1994) Labor Markets in an Era of Adjustment, Economic Development Institute, Washington DC: World Bank. Lachaud, J.-P. (1993a) ‘Les écarts de salaires entre les secteurs public et privé en Afrique francophone: analyse comparative’, Discussion Paper No. 53, Geneva: Institut International d’Etudes Sociales. – (1993b) ‘Pauvreté et marché du travail urbain en Afrique au sud du Sahara: analyse comparative’, Discussion Paper No. 55, Geneva: Institut International d’Etudes Sociales. – (1993c) ‘L’ajustement structurel et le marché du travail en Afrique francophone’, Discussion Paper No. 56, Geneva: Institut International d’Etudes Sociales. Levy, V. and J. Newman (1989) ‘Wage Rigidity: Micro and Macro Evidence on Labor Market Adjustment in the Modern Sector ’, World Bank Economic Review, 3 (1): 97–117. López, R. and L. Riveros (1990) ‘Do Labor Market Distortions Cause Overvaluation and Rigidity of the Real Exchange Rate?’, Policy Research Working Paper No. 485, Washington DC: World Bank. MacIsaac, D. and M. Rama (1997) ‘Determinants of Hourly Earnings in Ecuador: the Role of Labor Market Regulations’, Journal of Labor Economics, 15 (3): S136–S165. Miller, V. and L. Vallée (1995) ‘Labour Regulation and Segmentation in Cameroon’, in B. Gauthier (ed.), Economic Reform and the Development of the Manufacturing Sector in Cameroon, RPED Country Study Series, Washington DC: World Bank. Nashashibi, K. and S. Bazzoni (1994) ‘Exchange Rate Strategies and Fiscal Performance in Sub-Saharan Africa’, IMF Staff Papers, 41 (1): 76–122. Ndiaye, A.I. and B. Tidjani (1995) ‘Mouvements ouvriers et crise économique: les syndicats sénégalais face à l’ajustement structurel’, Série de monographies, 3, Dakar: CODESRIA. Nelson, J. (1991) ‘Organized Labor, Politics and Labor Market Flexbility in Developing Countries’, World Bank Research Observer, 6(1): 37–56. Niessen, P. (1995) ‘Labor Productivity in Senegal: a Discussion of Empirical Trends in Inter-sectoral Labor Productivity and Their Theoretical Causes’, unpublished manuscript, Washington DC: World Bank.


Rama, M. (1996) ‘The Consequences of Doubling the Minimum Wage: the Case of Indonesia’, Policy Research Working Paper No. 1643, Washington DC: World Bank. Rama, M. and R. Artecona (2000) ‘A Database of Labor Market Indicators across Countries’, unpublished manuscript, Washington DC: World Bank. Rama, M. and G. Tabellini (1998) ‘Lobbying by Capital and Labor over Trade and Labor Market Distortions’, European Economic Review, 42: 1295–1316. Runner, P. (1992) ‘Analyse du système d’observatoire de l’emploi et de la formation au Togo’, unpublished manuscript, Washington DC: World Bank. Savvides, A. (1995) ‘Economic Growth in Africa’, World Development, 23 (3): 449–58. Squire, L. and S. Suthiwart-Narueput (1997) ‘The Impact of Labor Market Regulations’, World Bank Economic Review, 11 (1): 119–43. Terrell, K. and J. Svejnar (1989) The Industrial Labor Market and Economic Performance in Senegal, Boulder CO: Westview Press. Vallée, L. and M. Thomas (1994) ‘Labor Market Segmentation and Labor Absorption in Cameroon’, in B. Gauthier (ed.), Manufacturing Enterprises under Adjustment in Cameroon: a Survey Perspective, RPED Country Study Series, Washington DC: World Bank. van der Gaag, J. and W. Vijverberg (1988) ‘Switching Regression Model for Wage Determinants in the Public and Private Sectors of a Developing Country’, Review of Economics and Statistics, 70: 244–52. van der Gaag, J., M. Stelcner and W. Vijverberg (1989) ‘Public–Private Sector Wage Comparisons and Moonlighting in Developing Countries’, LSMS Working Paper No. 52, Washington DC: World Bank. Vijverberg, W. and J. van der Gaag (1990) ‘Testing for Labor Market Duality: the Private Sector Wage in Côte d’Ivoire’, LSMS Working Paper No. 66, Washington DC: World Bank. World Bank (1993) Côte d’Ivoire: Private Sector Assessment, Washington DC: World Bank. World Bank (1995) Workers in an Integrating World, World Development Report 1995, New York: Oxford University Press for the World Bank.