SURVEY RATIONALE AND METHODOLOGY

The formation of a rural class depending on wage employment was already deeply rooted at independence (Castel-Branco Reference Castel-Branco1983; O'Laughlin Reference O'Laughlin2002: 517). After independence, the ruling Frelimo party, downplaying the heterogeneity of rural society, implemented policies on the basis of a crude dualist assumption, pitting a homogeneous subsistence-oriented peasantry against a ‘commercial’ sector. During the war, the class stratification of rural society continued, but differentiation processes were largely ignored by policy-makers and academics alike (O'Laughlin Reference O'Laughlin1996: 1; Wuyts Reference Wuyts and Addison2003).

Since the end of the war, new processes of economic and social change have begun to have substantial effects on rural inequality and the demand for wage labour. These changes have included: the rehabilitation of transport infrastructure; new foreign investment in agriculture; the immigration of farmers from Zimbabwe and South Africa (concentrated especially in Manica Province); the revival of tea plantations in Zambézia Province; the privatisation of other state-owned plantations; and the rapid integration of small- and middle-scale farmers into international commodity markets (especially the wage labour-intensive markets for tobacco and cotton). Similar dynamic influences on rural labour markets, in some cases even more pronounced, are a feature of many African societies. So the findings of this survey in Mozambique may well be relevant to other sub-Saharan African economies.Footnote ⁶

Unfortunately, the data collected in recent household surveys are not useful for analysing the impact of these uneven developments on the market for wage labour. The usefulness of these household surveys is limited because they adopt international statistical conventions for measuring labour market participation that are ill-suited to the complex reality of labour transactions in poor rural areas.Footnote ⁷ Where new instruments have been developed to overcome the problems evident in conventional household survey data, they clearly demonstrate the large amount of labour market activity missed in the types of national surveys conducted in Mozambique (Lund Reference Lund2004; Adato et al. Reference Adato, Lund and Mhlongo2004: 17).

The results from two recent household surveys in Mozambique, the Inquérito Nacional aos Agregados Familiares (IAF 2004) and the Questionário de Indicadores Básicos de Bem-Estar (QUIBB 2001), yield surprisingly different measures of the relative significance of wage labour, depending on the specific questions asked in each of these surveys and their interpretation (Table 1). In particular, conventional questions in both surveys about rural respondents' ‘main’ job suggest that wage labour is rare – only about 7·3% of household members in the QUIBB survey, or 4·7% in IAF, claimed to have been paid a wage or salary in their ‘main’ job.Footnote ⁸ In contrast, a much larger percentage of rural households (almost 21% of rural households in IAF, 17·4% in QUIBB) claimed that at least one household member had been employed as an agricultural labourer in the most recent agricultural season.Footnote ⁹

Table 1

Responses to questions about wage labour at household and individual level

Source: Authors' calculations from IAF 2004 and QUIBB 2001 database.

If the IAF demographic extrapolations are accepted, then it may be concluded that at least 561,000 rural households (representing over 2·6 million rural Mozambicans) have one or more members employed as seasonal or casual agricultural wage labourers. The results from the IAF survey also suggest that households in the poorest quintiles are more likely to contain members who are temporary agricultural workers.

During 2002–3, fieldwork for the Mozambique Rural Labour Market Survey (MRLS) was completed in three provinces in the centre and north of the country: Manica, Nampula and Zambézia, where 2,638 wage-employed respondents (slightly less than half of them women) answered a lengthy questionnaire and provided information not only about themselves but also about other household members. As a result, the survey collected data on some 16,000 individuals in these provinces. The respondents were employed by a wide range of different types of establishment (around 900 separate employers), varying from very small farms, bars and market stalls to large plantations employing thousands of temporary workers. The wide range of types of employment/establishment covered in the MRLS allows an analysis that is much more disaggregated than is usual in the literature discussing the main sources of income for poor and rich rural households. For example, Tschirley and Benfica (Reference Tschirley and Benfica2001) focus on the broad category of ‘non-agricultural’ wage income, ignoring the important differences between, for example, households dependent on salaried office jobs in the rural bureaucracy (teachers, health workers, administrators, etc.) and households dependent on the wages of a waitress or cleaner in a roadside barraca.

The sampling was purposive rather than random. There was, of course, no reliable sampling frame on which to base a random sample of rural wage workers. The available official statistics not only fail to collect information on employees in enterprises employing fewer than ten workers, but also exclude many enterprises employing more than ten workers if, as is often the case, these enterprises are not legally registered. Ministry of Labour officials lack the resources, training or incentives to investigate employment, especially seasonal and temporary employment in many local firms.

The research team therefore used a wide range of sources to construct, as completely as possible, its own provincial sampling frames on the basis of existing agricultural censuses (TIA 2002 and the Agricultural Census 1999/2000 (INE 2000) on which TIA was also based), recent household survey lists (from the QUIBB 2001 database), and visits to all the relevant provinces and districts to assemble up-to-date and reliable lists of large and middle-scale farms;Footnote ¹⁰ lists of those enumeration areas that had recently experienced dynamic economic growth and structural change, but had not been covered by previous household surveys (IAF, QUIBB);Footnote ¹¹ and the names and location of other large-scale employers in different districts of Nampula, Zambézia and Manica. The three central and northern provinces were selected because they account for the bulk of labour-intensive cash crop production (cotton, tobacco, sisal and tea), and Nampula and Zambézia contain a very large proportion of the Mozambican rural population.

This preliminary work to establish these sampling lists provided some assurance that the MRLS would not miss either the most significant rural employers, or those enumeration areas where wage employment had recently become particularly important in each of the provinces.Footnote ¹² The sample purposively included a relatively large number of small- and middle-scale farmers.Footnote ¹³ The most successful of these farmers, who account for much of the demand for agricultural wage labour, are non-randomly distributed in rural Mozambique and there is, therefore, no guarantee that their wage workers would be included in conventional, randomised sample surveys of households. The range of occupations and of types of employer captured was large enough to ensure that sufficient diversity was achieved to make the results statistically relevant, and to obtain a well-targeted coverage of the most important rural labour markets within the selected provinces.Footnote ¹⁴

The principle used in the purposive sampling in this research was that of ensuring ‘maximum variation’, so that the final sample size was a function of the extent to which ‘saturation points’ had been reached for several key strata of jobs (agricultural, non-agricultural, small-scale vs large-scale employers, local and foreign investors and so on). The combination of care taken to build appropriate sampling frames, the large sample size, and the principle of maximum variation provide grounds to believe that the MRLS is a fair representation of the wide range of different types of rural wage workers in the provinces selected, and of the various categories of employers (see Tables 2 and 3 for a breakdown).

Table 2

Sample distribution by province and employer category

Source: MRLS 2002/3.

Table 3

Sample distribution by province and category of non-agricultural labour

Source: MRLS 2002/3.

When the results of this research and the results of the nationally representative IAF survey (2002/03) are compared, it is clear that the purposive sampling of people working for wages in the MRLS succeeded in capturing many respondents who would certainly be classified by IAF as among the poorest Mozambicans. Thus, the bottom third of the MRLS sample, ranked according to a simple but robust household asset index,Footnote ¹⁵ is at least as poor as or poorer than the bottom quintile of households surveyed by IAF in the rural areas of the same provinces. As Table 4 shows, fewer of the poorest tercile of MRLS respondents in Manica own a bicycle, a watch or a radio than the poorest quintile of IAF respondents in that province. And in all three provinces, fewer MRLS respondents had eaten meat within the last week or owned a watch. Educational status is known to be closely associated with other measures of poverty in Mozambique (Simler et al. Reference Simler, Mukherjee, Dava and Datt2004); 80% or more of the poorest households in both MRLS and IAF (with the surprising exception of IAF respondents in Manica) failed to complete primary school. When other education statistics (and most of the demographic statistics) for the whole sample of rural wage workers in the MRLS were examined, the results were not statistically different from the IAF rural samples in the same provinces.Footnote ¹⁶ In sum, the purposively sampled survey is a policy-relevant complement to standard survey techniques; both the questionnaire design and the MRLS sampling procedures were planned specifically to fill in gaps that are common to household surveys in Africa (Sender et al. Reference Sender, Cramer and Oya2005).

Table 4

Assets and education compared: MRLS and INAF surveys

Sources: INAF 2004; MRLS 2002/03.

This paper also reports findings from 120 respondents to a different questionnaire administered to a sample of small- and medium-scale agricultural employers. In addition, the researchers conducted semi-structured interviews with thirty-three large-scale employers, all employing more than fifty workers at the peak of the agricultural year, and collected the life histories of fifteen female wage workers.Footnote ¹⁷ Thus, this paper aims to combine the quantitative findings of large- and smaller-scale purposive surveys with primary qualitative material to question both the policy conclusions reached in much of the current literature on rural Mozambique, and some of the methods commonly used.

VARIATIONS IN METHODS OF PAY: MONTHLY, DAILY AND PIECEWORK PAY

Payment arrangements in rural labour markets are extremely complex. They are difficult to investigate and summarise (Hatlebakk Reference Hatlebakk2004; Rogaly Reference Rogaly2005). The literature on piece-rate systems and farm wage differentials attempts to explain marked differences between how workers are paid, even when they are doing similar things and in comparable locations. The focus in this literature is often on the costs of labour recruitment and supervision faced by employers, and on the implications of different payment methods for labour productivity (often in the context of debates about ‘efficiency wages’). There are also many other explanations, less firmly rooted in the neoclassical economic literature, for observed variations in payment methods (Bardhan & Rudra Reference Bardhan and Rudra1986; Newman & Jarvis Reference Newman and Jarvis2000; Ortiz Reference Ortiz1999; Rogaly Reference Rogaly2005; Rubin & Perloff Reference Rubin and Perloff1993; Wells Reference Wells1996). In Mozambique, forms and levels of payment vary within provinces from one rural area to another; they appear to be influenced by the strategies adopted by individual employers, making any general statement about or explanation of prevailing methods on Mozambican farms questionable.

Table 5 shows the distribution of the main methods of payment in the MRLS sample, based on workers' responses to both coded and open-ended questions.Footnote ¹⁸ However, in many of these responses the distinction between (unwritten) contracts to purchase labour time and contracts to complete a specific task was unclear, so the classifications in Tables 5 and 6 should be regarded as ‘best estimates’, rather than definitive.

Table 5

Wage payment methods for agricultural and non-agricultural workers, by type of employer

Note: ^a Number in brackets shows the total number of observations for each category. Note that the total 2620 differs from the 2638 reported in Table 2 as a result of some missing values for pay modalities. The total of agricultural labourers may slightly differ from the same total in other tabulations due to missing values in some of the categories used.

Source: MRLS 2002/3.

More than two thirds of the respondents employed as agricultural labourers were not paid on a monthly basis. An even larger proportion (almost 80%) of the agricultural workers employed by local farmers (small Mozambican farmers or medium-scale private farmers known as privados) were paid either by task or on a daily basis. In contrast, over 80% of agricultural workers employed on foreign-owned firms were paid monthly wages.Footnote ¹⁹

Payment methods on the larger farms, particularly on ‘foreign’ or ‘national’ corporations, were more clearly defined than on other types of farm, where the variation in payment methods and rates was particularly large. Table 6 shows that larger farming enterprises are more likely to employ monthly paid workers than smaller farmers. However, many of the large employers recorded as paying a monthly wage to temporary workers were, in fact, applying a daily rate (often derived from the legal monthly minimum rate), although the number of days of work required per month was at the discretion of individual employers.Footnote ²⁰

Table 6

Wage payment methods by size of establishment (agricultural workers)

Notes: ^a 1–10 workers; ^b 11–50 workers; ^c 50+ workers.

Source: MRLS 2002/3.

Larger farms also need to employ a great many temporary workers for one to three months to meet seasonal labour peaks. These workers are usually required to complete the task set by their employers before they receive a ‘daily’ payment. The tasks involved varied, as did the judgement of what could be done in a day and how many hours a casual worker should be expected to work. On some farms there was a stable daily wage paid for seasonal tasks, but these tasks would vary in intensity. Yet on other farms the daily wage varied according to the task set. For example, on a large joint venture farm in Nampula Province growing sisal and cotton, day workers were paid MT31,347 a day for a set task (cutting ninety-eight piles of sisal leaf), but for clearing weeds in sisal fields the task rate was MT21,577 (for three ‘lines’ of weeding). Thus, among large-scale employers, even when a ‘daily rate’ applied, this could vary depending on the specific task involved. Table 7 gives some idea of the latitude for setting differing wage rates for weeding.

Table 7

Daily pay for weeding by selected large and mid-scale employers in Nampula (Meticais)

Notes: ^a For tough weeding; ^b for lighter weeding.

Source: MRLS 2002/3.

On the larger farms in the sample, with hundreds of casual workers at peaks of the agricultural cycle, employers often paid a bonus for work well done or for regular attendance. Although some respondents only performed an occasional day of ganho ganho (casual) labour on these farms, employers appeared to prefer offering longer (unwritten) contracts, typically for one to three months, to ‘daily’ labourers. For example, on one farm the employer offered MT5,000 per day on top of the daily wage, payable on the last day of an agreed span of work and conditional on the employee having turned up regularly. In many cases it was left to field supervisors to judge whether a task had been done satisfactorily or whether a bonus had been earned.

Local farmers/privados rely much more heavily on task-based payment systems. The tasks their workers have to perform are even less clearly defined than on the large farms and require widely varying amounts of time and effort to complete, making it extremely difficult to calculate the wage received per day or hour.Footnote ²¹ One important advantage of task-based payment systems is that they allow privados to incorporate the labour efforts of additional female and child labour without having to contract (or pay) these workers directly. Thus, a woman who has difficulty completing the task set by her employer will bring along her children and female relatives to ‘help’ with the work.

Payment methods are determined not only by the size and ownership characteristics of farms, but also by the crop and the farming operation. An analysis of 733 responses to open-ended questions concerning agricultural tasks (mainly from the sample in Nampula province) suggests the following patterns: work in cultivating maize, cashew and rice is almost always paid by task or piece-rate; where tobacco, sisal and horticulture are the main crops, the proportion of more regular monthly-paid workers is much higher than for other crops; tasks like weeding and harvesting are more likely to be paid on a piece-rate basis, whereas clearing and sisal stumping are usually paid on a daily or monthly basis.

A CLASSIFICATION OF JOBS AND PAYMENT METHODS

The analysis above has shown that rural Mozambicans experience a complex range of methods and rates of payment for wage labour. A crude dichotomy between privileged ‘labour aristocrats’ and all other rural workers cannot capture this reality. The MRLS allows for a more nuanced, although still rather simple, taxonomy of wage employment. Thus, Table 11 identifies five main types of wage employment obtained by rural Mozambicans, using a mixture of quantitative and qualitative criteria. The classification contrasts the characteristics of workers who are relatively (un)successful in the labour market, in terms of earnings and some aspects of working conditions (e.g. access to trade unions and compensation for working overtime).

Table 11

Better and worse jobs – payment methods and rates

Note: ^a SD=Standard deviation.

Source: MRLS 2002/3.

Table 11 shows that some (770) workers enjoy access to ‘good’ jobs that guarantee a relatively decent and more or less regular flow of income. A similar number of workers (708) only succeeded in finding casual or very low-paid (‘bad’) jobs. The five types of employment identified in the table are not mutually exclusive. For example, the second is a fraction of the first. Categories ‘bad 1’ and ‘bad 2’ are also partly overlapping, and a few workers with access to ‘good’ jobs also perform some of the ‘bad’ jobs on the side.

Although the median wage of all the agricultural workers classified as having ‘good’ jobs is still below the statutory minimum, the ‘good 2’ category workers do receive 13% more than ‘good 1’ category workers, and over 25% more than other monthly-paid unclassified workers. Moreover, these agricultural workers are more likely to have access to trade union representation than workers in any other type of job (Table 12). However, the benefits they derive from union representation are not clear. One or two of the biggest agricultural employers stated that in the past unions were combative and even aggressive; of late they have only been ‘helpful’. This was confirmed by other employers (and by provincial union officials) who said that union officials either never visited farms or that they did visit from time to time but would only do so if they could get a lift from the farmer – in other words, one major constraint on union officials organising on farms is the lack of transport facilities. Clearly, the disorganisation of unions – chiefly the Sindicato Nacional de Trabalhadores Agro-Pecuários e Florestais (SINTAF) – and the failure of both government and international donors to invest in increasing their capacity, allow for the high degree of employer discretion in setting wages and their composition in terms of money wages, benefits and payments-in-kind.

Table 12

Employment tenure and other work conditions, by types of job

Source: MRLS 2002/3.

Among monthly-paid workers, the 145 domestic servants receiving less than the minimum wage are particularly disadvantaged, since half of them earned below MT150,000 per month. The median daily rates of the agricultural workers employed on a casual basis (ganho ganho), whether they are classified as having a ‘bad 1’ or ‘bad 2’ job, are consistently lower than the daily rates for any other type of worker.

Moreover, workers with bad jobs were more likely to be paid in kind (usually with food) than any other worker (Table 11). The most common substitutes for money wages reported in interviews with large farmers were dried fish, sugar, soap, maize or cassava flour, and capulanas (the cloth wraps worn by women). For example, a woman worker might work for four days to earn a capulana ‘worth’ (according to the farmer) MT35,000. Obviously, precise estimates of an imputed money wage (or the employer's wage costs) are difficult when payments are made in kind. The lack of precision appears to increase employers' control over the terms on which they acquire labour.

In all types of rural wage employment, job tenure appears to be insecure. Even for those workers who have a ‘good’ job, half report spending twelve months or less in their present job. More than half of the workers with the worst jobs (bad jobs 1 and 2) only manage to find twenty days or less of wage work per year. An increase in the number of days per year when they can find employment would have a dramatic impact on their standards of living.Footnote ²⁹

Small-scale employers and especially Mozambican-owned small farm enterprises are unlikely to offer ‘good’ jobs to their workers, while almost two thirds of workers employed by foreign agricultural investors enjoy good jobs in agriculture (Table 13). Many workers with relatively ‘good’ jobs suffer from employment conditions that are below statutory minimum standards. For example, about half of them do not receive any compensation for working overtime (Table 12).Footnote ³⁰ Nevertheless, they are more likely than workers with ‘bad jobs’ 1 and 2 to be provided with housing, meals and credit by their employers.

Table 13

Workers by employer categories and job types

(% of workers within each employer category)

Source: MRLS 2002/3.

The amount and quality of these non-wage benefits, as revealed in interviews with large farmers and in the quantitative surveys, is variable and discretionary. Benefits can be and are withdrawn, whenever the employer feels this is appropriate. Moreover, almost all the women captured in the MRLS were denied access to the most basic non-wage benefits. Thus, only about 3% had paid maternity leave; less than 4% of all female wage workers were given paid holidays by their employers and less than 10% were given paid sick leave or any medical benefits.