Introduction
Despite the persistent analytical, statistical and policy focus on poverty in sub-Saharan Africa over the past two decades, the linkages between poverty and employment dynamics remain under-researched and poorly understood, especially in rural areas. Economic growth, education and health are usually seen as essential for poverty reduction. As Amsden (2010) has provocatively and lucidly argued, this agenda has led to ‘jobs dementia’ and a grassroots approach to poverty reduction that has generated poverty reduction fantasies with little support from historical evidence (see also Wuyts 2011). A striking imbalance emerges between the ever-expanding agenda on rural poverty and the little attention and evidence generated on rural employment. In particular, a cursory navigation of the vast literature on poverty in Africa and the official ‘poverty reduction strategies’ (notably reflected in a large number of PRSPs since 1999) gives clear indication that wage employment in rural Africa is generally ignored. As a result, our knowledge of rural wage employment and how it relates to rural poverty is exceptionally limited. Conventional wisdom and official data on rural employment in sub-Saharan Africa suggest the following ‘stylised facts’ (see, for critical assessments, Sender 2003; Sender et al. 2005; Cramer et al. 2008; White et al. 2006): (a) agriculture mainly consists of small ‘subsistence’ peasant farmers, most of them ‘poor’ (usually defined with reference to a monetary threshold of poverty); (b) rural inequality and stratification is not substantial, with the exception of highly unequal Southern African societies; (c) the non-farm economy, which is growing in significance, is thought to be more heterogeneous but dominated by self-employment; (d) as a corollary of these features, rural labour markets are regarded as thin or absent and rural wage employment, especially agricultural wage employment, as uncommon; (e) besides, cooperative/reciprocal labour exchange is often seen as more common than wage labour, partly reflecting supposedly greater degrees of equality, more widespread poverty and social capital in play rather than any form of pressured activity within structured (not equal) social relations.1
This paper examines this gap in knowledge with a particular focus on the collection of statistics on employment in rural Africa. The article thus explores the reasons why rural (especially agricultural) wage employment is so poorly understood and especially inadequately measured, and provides methodological alternatives to better capture the nature and dynamics of rural/agricultural wage employment. Finally, to illustrate these methodological lessons, the article summarises key themes and evidence emerging from micro-level studies that have attempted to overcome the main methodological barriers to a better understanding of rural labour markets in Africa. The article is organised as follows. The second section following this introduction briefly documents the relative neglect of rural wage employment in the literature and its relative invisibility in available statistics. There is a sin of omission and a sin of commission in both official statistics and conventional research, as will be argued in the following section, which also underpins large discrepancies in the significance of wage employment across African countries. The third section examines some of the main reasons for the apparent lack of development of rural labour markets and the associated scarcity of evidence of rural wage employment. It will be argued that there is a serious problem with the evidence available at the micro level from official statistics, and competing explanations will be reviewed. The fourth section will then illustrate some key themes emerging from micro-level rural labour surveys (where I directly participated) in which conventional data collection deficiencies were addressed and corrected. The final section will summarise some of the main points and will draw some tentative policy and research implications.
Neglected rural wage employment
The paucity of literature on wage employment in rural Africa is reflected in illustrations from Google searches. While the keyword ‘agricultural wage employment’ seems more commonly used, ‘rural wage employment’ only yields 305 hits in Google Scholar (for all years and regional domains) and 183 if one adds ‘Africa’ to the search. If we restrict the keyword to its use in titles (a more accurate representation of research that focuses on this topic) we get a paltry seven hits, only three discussing evidence on sub-Saharan Africa (SSA). A similar search with terms ‘rural wage labour’ and ‘Africa’ gives similar results with a few more hits. Anyone with an interest in wage employment in rural Africa and in agriculture faces two major challenges: first, finding enough secondary material that is up to date; second, finding reliable data series at country level. This is a simple and biased illustration of a sin of omission, i.e., the neglect of a topic that should deserve much more attention. There is of course a wealth of fascinating scholarly material on social differentiation and rural labour market formation in Africa (see Sender and Smith 1986; Swindell 1985; Kitching 1980; Ghai and Radwan 1983). Unfortunately, this kind of literature has followed a remarkable decline since the mid 1980s, partly mirroring the broader decline of agrarian political economy in African studies and the growing dominance of neo-classical economics and neo-populist approaches to rural development, which have largely focused research on smallholder producers, markets and agricultural reforms (see Oya 2011).
There is also a sin of commission. In fact, much academic and institutional literature on rural Africa, and indeed official published statistics on employment, posit that rural labour markets are either absent or very thin, especially for agricultural wage employment. Even emerging mainstream literature that acknowledges the importance of livelihood diversification and the rise of off-farm income sources also emphasises the ‘limited place of off-farm agricultural wage labour in rural African livelihoods’ (Barrett et al. 2005, 15). These views are echoed by official poverty reduction policies, where the place of wage employment is sorely neglected in favour of the usual focus on smallholders' incomes and productivity. Official statistics certainly tend to reflect an image where wage employment is marginal in much of rural Africa.2
The very low percentages of recorded wage employment in rural areas in fact underpin the low proportions recorded by aggregate employment statistics compiled by the International Labour Organization (ILO). Table 1 illustrates the contrast between SSA and other regions. In the former, the median percentage of people classified as ‘employees’ (wage workers in terms of their ‘main’ employment) is 20%, compared to 58% in Latin America and around 45% in Asian MICs (see also World Bank 2007; Gindling and Newhouse 2012; Sender et al. 2005). These aggregate averages however mask stark contrasts between African countries, as shown by Oya (2010, 6, Table 2). Contrasts on rural wage employment incidence are even more obvious. The WDR 2008 (World Bank 2007), for example, displays a table (Table 9.2) where a strong distinction can be drawn between SSA and other developing regions in terms of the proportion of people classified as wage-employed in rural areas (World Bank 2007, 205). First, in terms of agricultural wage employment, only 4% of adult men and 1.4% of women are reported to ‘mainly’ work as agricultural wage labourers in rural Africa compared with 22% of men and 11.4% in South Asia, respectively, or 21% and 2.3% respectively in Latin America (the table does not allow overlaps so people are classified only with respect to a notional ‘main activity’). Second, non-agricultural rural wage employment appears also significant in South Asia, Middle East and North Africa (MENA), East Asia and Latin America (between 15% and 31% for men, and up to 12% for women), in contrast with SSA, where this type of employment only applies to 9% of men and 3% of women. So, overall, about 13% of men and 4% of women in rural SSA are classified as wage earners, compared with around 38% of men and 14% of women in rural Latin America.
| Wage & salaried workers (employees) (%) | Total self-employed workers (%) | Share of vulnerable employment in total employment (%) | ||||
|---|---|---|---|---|---|---|
| Average | Median | Average | Median | Average | Median | |
| SSA (34 countries) | 28 | 20 | 70 | 80 | 69 | 79 |
| South Africa (data for 5 years in 2000s) | 83 | 82 | 17 | 17 | 13 | 13 |
| Asian Middle Income Countries | 43 | 45 | 57 | 55 | 54 | 53 |
| Latin America (6 countries) | 57 | 58 | 42 | 42 | 37 | 37 |
| Source: Author calculations from KILM (7th edition). |
| 1990–92 | 1995–97 | 2003–05 | |
|---|---|---|---|
| Developing regions | 2.6 | 2.7 | 2.7 |
| Northern Africa | 1.7 | 1.6 | 1.6 |
| Sub-Saharan Africa | 1.9 | 1.9 | 2.0 |
| Latin America and the Caribbean | 0.8 | 0.7 | 0.6 |
| Eastern Asia | 6.2 | 6.1 | 5.3 |
| Southern Asia | 3.0 | 3.2 | 3.4 |
| South-Eastern Asia | 2.7 | 2.8 | 2.6 |
| Western Asia | 0.9 | 0.9 | 0.9 |
| Source: FAOSTAT. | |||
| Note: this refers to population engaged in agriculture. |
As noted above, variations within SSA make aggregate statistics not particularly useful. Winters et al. (2008), one of the few recent studies focused on rural wage employment in developing countries, use information from three African countries: Ghana, Malawi and Nigeria. While their recorded average in the incidence of rural wage employment is also much lower than other non-African countries in their sample, there is a massive contrast between Ghana and Nigeria, on the one hand, where recorded participation in rural wage employment does not exceed 8%, and Malawi, on the other hand, where this reaches 39%.3 Part of this contrast can be traced to differences in coverage, since Ghana and Nigeria data would suggest that only ‘permanent’ wage workers were captured in these surveys.
What are the reasons for this apparent African ‘exceptionalism’ with regard to rural wage labour and for the big variation within Africa itself? Before we focus on key methodological challenges, which are the main aim of this paper, it is worth briefly addressing the conventional explanations for the lack of recorded evidence of rural wage labour in Africa.
Why is rural wage employment underestimated and misunderstood?
Conventional hypotheses
This section addresses possible explanations of the relative neglect and invisibility of rural wage employment in much of the literature and statistics on most African countries. First, some have argued that Africa is characterised by land abundance or high land/labour ratios and therefore by processes of accumulation without dispossession (Berry 1993). In addition, according to other analyses, in arid-to-semi-arid areas, very low population density, low agricultural labour productivity and abundant land combine to reduce both supply of and demand for farm wage labour (Barrett et al. 2005). Indeed, this may explain why a class of ‘pure proletarians’ has not emerged in rural Africa, especially in arid and semi-arid areas. But this is not the point. As Lenin (1899) showed when writing about capitalist development in Russia, ‘pure proletarianisation’ was never rapid or an inevitable accompaniment of capitalist development, since most workers still retained some land for long time periods.4 Available evidence on rural wage workers, as discussed in the final sections of this paper, shows indeed that most of them do have access to some land. Table 2 suggests that although labour/land ratios are lower in Africa than in Asia (see also Karshenas 2001) Africa is not distinct in terms of very low labour/land ratios as reflected by the figures for Latin America, the Middle East (North Africa and parts of Western Asia) and parts of Central Asia, where reported rural wage employment is generally much higher than in Africa. Moreover, population densities also vary within SSA, and the most densely populated countries like Rwanda or Burundi present figures of proportion of wage workers that are as low as 6–19%, of which half being farm workers (years 1996 and 2006) and 5.2% (year 2004) respectively at national level (Rizzo 2011, 7, for Rwanda and KILM–ILO for Burundi). Botswana, Namibia and Gabon with some of the lowest population densities, have some of the highest proportions of wage employment in Africa.
A second argument is the idea of ‘uncaptured peasantry’, associated with the resistance of rural people to proletarianisation and capitalism.5 This contrasts with the historical evidence from a wide range of capitalist economies, which shows that long periods of ‘resistance’ have been overcome, albeit unevenly and through a variety of factors, from dispossession, to the attractions of industrial and urban employment, to social differentiation driven by market compulsion and so on (Byres 2003). As countries develop and capitalism matures self-employment loses significance as the ‘standard employment relationship’ of firm-based employment settles (Schaffner 1993; Schultz 1990). This is partly reflected in a strong cross-country negative relationship between the reported incidence of self-employment and levels of income per capita globally and within Africa (Figure 1, and Gindling and Newhouse 2012). It is not the intrinsic capacity of peasants to resist but the overall level and speed of development that tends to correlate strongly with changes in employment status. There is some evidence that resistance existed and still exists, but even in the event of ‘subjective’ resistance (i.e., some people would prefer to be ‘independent’ producers), this does not necessarily translate into the possibility of total avoidance of market compulsion to work for wages (Bernstein 2010). Failure to resist the economic compulsion to work for wages may therefore be much more frequent than normally assumed, even though many poor rural Africans still cling to farming their tiny plots.
Third, very low wages (too similar to or lower than returns to family labour) could be another reason and, in contexts where poverty is pervasive and potential small-scale employers have little cash to spare, a constraint on wage labour supply. However, there is no convincing empirical evidence about this, partly because it is hard to find studies that systematically and rigorously compare wage rates with net returns to labour in own-account activities, especially farming (Kevane 1994; Sender et al. 2005). Research on wage employment cited below shows marked heterogeneity in wage working conditions and that, for many, even low wages are preferable to extremely volatile and low returns to self-employment in marginal small-scale farming (see also Palmer and Sender 2006).
Finally, an argument about the ‘lack of employers’ is common in surveys that ask overly general questions about rural employment opportunities. First, there is a problem of interpretation, i.e. that respondents frequently associate the idea of ‘paid employment’ (and inadequate translations of this term, as learned from our qualitative research) with formal sector jobs and generally stable job-holdings, while they dismiss the much more pervasive forms of casual wage employment as not constituting salaried/wage employment, whether it is for small or large producers. Second, while some of the casual labour performed for middle- and large-scale commercial agriculture may be underestimated despite evidence of the growing significance of these forms of production in Africa (World Bank 2007), large-scale capitalist producers are not the only employers of rural labour, since a wide range of jobs occur among smallholders and small and medium non-agricultural businesses. Therefore, the argument about weak labour demand depends much on what kind of wage labour demand is accounted for and how it is captured.
The weaknesses of conventional labour data collection
While the arguments briefly explored above may partially help us understand instances where the incidence of rural wage employment is genuinely low, another more plausible explanation is that, in fact, rural wage employment is poorly captured/measured by existing official data and even some micro-level research is done with inadequate, standardised methods and without sufficient attention to the peculiarities of rural labour markets in developing countries, especially in poor African economies (Cramer et al. 2008; Mwamadzingo 2003, 31; White et al. 2006). There are different reasons for the paucity and unreliability of data on rural wage employment and rural labour markets, which will be addressed in the following sub-sections. In a nutshell, the poverty of (rural) labour statistics in Africa can be traced to a combination of weaknesses and tendencies that have contributed to obscure existing realities of labour relations, namely, (a) the growing marginalisation and simplification of questions on labour and employment in nationally-representative household surveys in favour of the collection of data on consumption and other welfare indicators; (b) the problems in survey design for labour indicators, notably the inadequacy of some statistical conventions, definitions and survey practices, and the definitions and boundaries of the ‘household’.6
Shortage of labour force surveys and neglect of labour modules
Efforts to collect systematic, detailed and context-specific labour market statistics are generally rare in sub-Saharan Africa (Backiny-Yetna 2003). The scarcity of in-depth (rural) labour force surveys in Africa is striking, especially in comparison with Latin America and Asia (Sender et al. 2005; Mwamadzingo 2003). The frequency of labour force surveys (LFS) in sub-Saharan Africa is very disappointing and some countries have not had an LFS since the 1970s. The situation has not improved despite recently growing emphasis on employment and jobs in international agencies and governments.7 Labour force surveys, if carefully designed, are the main vehicle to collect reliable and sufficiently disaggregated information on employment, but have fallen off the table of priorities for statistical agencies in Africa, largely reflecting the priorities of international aid agencies that often contribute a very large proportion of budgets in African statistics agencies.
The World Bank plays a big part in this by putting most of its emphasis on income/ expenditure surveys, welfare indicators and multiple integrated household surveys since the 1980s. For example, the World Bank has funded many Living Standards Measurement Surveys (LSMS), including Household Income and Consumption Expenditure Survey (HICES), Core Welfare Indicators Questionnaires and Multiple Indicator Cluster Surveys (MICS), and provided much of the technical assistance on data collection (see, for an example, Grosh and Glewwe, 2000 and http://go.worldbank.org/IPLXWMCNJ0 ). The burden of income/expenditure surveys – in themselves extremely skill-demanding and time-consuming while also potentially affected by large measurement errors – and the effects of multiple donor agendas have led the trend towards organising ‘integrated’ large-scale household surveys in an attempt to save time and money so that more data are collected in a single shot. These surveys have multiple aims and scope, and often contain lengthy modules on issues such as health, education, social capital and community development. It is hard to believe that this has not happened at the expense of quality and depth in data collection, especially on employment. It is somewhat striking that the Bank has acknowledged the paucity of data on rural wage employment in the World Development Report 2008 (World Bank 2007) and generally huge gaps in labour statistics (World Bank 2012), without a mea culpa. In sum, the dominant poverty and MDG agendas since the 1990s have resulted in funding for employment-focused surveys being dried up, as many statistics agency officials have told us in the course of our research in Mozambique, Mauritania, Zambia and Senegal. Agencies like the ILO simply do not have the financial muscle to fund these efforts in data collection, and perhaps it is not their mandate to do so.
Problems in employment modules and employment-related questions in conventional household surveys
The paucity of labour surveys and data is compounded by the low quality of the data actually collected. One principal reason is that employment modules in conventional household surveys in Africa are generally weak, partly for some of the reasons stated above. The integrated and income/expenditure surveys promoted by the Bank and other donor agencies use employment modules that are small, very standardised and generally not adequate to capture the complexity and specificity of rural labour relations in developing countries. In fact, even surveys following standard ILO guidelines may not be able to capture complex labour relations, which are common in developing countries. As Standing (2006) argues, the ‘labour force approach’ has dominated the statistical agenda for labour indicators since the 1930s–1940s, when concern over mass unemployment was paramount. The resulting ‘trichotomy’ of ‘employed, unemployed and economically inactive’, which suits industrialised labour markets in high-income societies, is problematic for analyses of labour market and employment dynamics in low-income countries where the ‘standard employment relationship’ is much less well established.8 The standard dichotomy of self- vs. wage-employment is also very problematic. Wuyts (2011, 11) questions the significance and rise of ‘self-employment’ and argues that ‘under conditions of low productivity in the context of the uncertain environment of informality the character of wage labour, and the variety of forms in which it occurs, does not correspond to the conventional or “formal” definition of wage labour’. Furthermore, implemented surveys do not even follow the detailed guidelines produced by the World Bank itself on labour questions (see Grosh and Glewwe 2000, volume 1, 217–250) but focus instead on the details of consumption expenditure. As a result, as. Sender clearly puts it:
[In] most developing economies no efforts at all are made to collect time-series data on the wages of those employed in small-scale farm and non-farm rural enterprises, especially on the wages of those who are irregularly, seasonally, or casually employed. In most of these economies, in fact, there is no reliable data on the number of people or households that depend upon earnings in these types of employment; it is simply assumed that the rural poor are, or will become self-employed… (Sender 2003, 414)9
Second, the excessive reliance on standard questions with a seven-day reference period, which in contexts of strong seasonality, irregularity of activities and occupation multiplicity, introduces potentially damaging statistical biases. Asking a poor rural person what he/she has done in the last seven days is not particularly useful, given structural seasonality, occupation multiplicity and job instability.11 This also biases the responses to questions on ‘main occupation’ since different occupations have very different time patterns.
Third, very crude questions and distinctions on status in employment (with the main distinction between self-employment versus paid employment) may miss out on a range of economic activities which would be classified as paid employment. This applies especially to those of a casual nature that are particularly prevalent, precisely because of their diversity, irregularity and sometimes ambiguous nature. Despite the efforts by the International Conference of Labour Statisticians (that is convened by the ILO) to establish clear distinctions according to the International Classification of Status in Employment (ICSE-93) and, independently from this under the International Standard Classification of Occupations (ISCO-08), their applicability in rural areas of Africa is not always straightforward. The aim of ISCE-93 is to classify jobs ‘with respect to the type of explicit or implicit contract of employment of the person with other persons or organisations’, and the basic distinction is between paid employment and self-employment (under which employers, own-account workers and contributing family workers fall). ISCO-08, which focuses on the tasks and duties undertaken in the job, does not map occupations directly into status in employment.
However, the practice of many enumerators in large-scale data collection processes is often to assume that certain jobs/occupations ‘naturally’ belong to a particular status in employment (self-employment or paid employment). I have observed this in rural surveys where enumerators often assume a self-employment status to typically ‘informal’ occupations (e.g., in trade and transport), without the necessary probing. For example, most jobs listed in Table 3 (in rural Mauritania) were initially considered as ‘self-employment’ by enumerators almost automatically. Once conscientious probing was done, the proportion of paid employment within some of those categories significantly increased to the levels recorded in Table 3. Therefore, the use of supposedly well-defined standard labour statistical categories sometimes becomes a problem and a source of serious biases. Part of the problem lies in the standardisation itself, and the use of general standard questions across widely different contexts. Partly it lies in the extent to which standard labour categories reflect the labour force approach applied to data collection in the particular context of advanced capitalist countries (Standing 2006). Arguably, the notion of ‘self-employment’ may carry different meanings in different places unless a detailed explanation is provided in each case. In fact, as cogently argued by Wuyts (2011), many informal workers usually regarded and classified as ‘self-employed’ simply constitute another form of wage labour that does not conform to standard conventions (see also Breman 2006). Conceptually, they should not be seen as self-employed because they do not own the means of production, their job is highly insecure, and they earn a residual income after the entrepreneur (owner of the means of production) is paid the agreed (rather imposed) profits. This situation is very common in non-farm employment but some may argue that even smallholder contract farming often constitutes another form of ‘disguised wage labour’ especially for the poorest producers (Clapp 1994). Based on more than 10 years of experience in training enumerators for rural labour surveys, I come to conclude that the distinction between self-employment and wage employment in rural settings remains one of the biggest challenges in rural labour survey designs. Enumerators easily make mistakes and have an inclination to classify many activities as ‘self-employed’ without properly probing the central questions of ownership and control over means of production and labour decisions.
| Activity | Monthly salaried(%) | Daily or piecework wage labour(%) | Commission or profit share(%) | Own-account with own means of production(%) |
|---|---|---|---|---|
| Working in palm tree/date cultivation | 0 | 43 | 14 | 43 |
| Mason/brick maker | 0 | 100 | 0 | 0 |
| Trader/shopkeeper | 12 | 0 | 10 | 78 |
| Small itinerant trader | 0 | 0 | 5 | 95 |
| Processing agricultural commodities or homemade food for sale | 14 | 9 | 0 | 77 |
| Teacher (koranic) | 65 | 35 | 0 | 0 |
| Hairdressing | 0 | 0 | 0 | 100 |
| Artisan (weaving, dye, tailor, etc.) | 0 | 0 | 3 | 97 |
| Fishing | 0 | 14 | 0 | 86 |
| Source: Author elaboration from Rural Labour Market Survey in Mauritania. Also GIRM and World Bank (2007). |
Fourth, in many parts of rural Africa there are stigmas associated with casual wage employment, especially the most exploitative and degrading forms, which are thus easily under-reported or unreported altogether. Many of the local words used to describe these occupations indeed reflect the social stigma associated with them. Kibarua in Swahili speaking countries (especially Tanzania) is in fact a derogatory term that is used for most forms of casual manual wage labour; it contains reminiscences of slave forms of labour (Mueller 2011, 33). Hill (1968) reported how men in rural Nigeria typically considered wage work for neighbours as a source of shame, as being a sign of desperation. The terms ganho-ganho in Mozambique and ganyu in Malawi, used for various forms of casual labour, are also intrinsically associated with low social status and despair (Bryceson 2006). In addition, patriarchal barriers to women's entry into local wage employment, particularly for demeaning casual jobs, may lead to important biases in reporting women's participation in rural labour markets (Sender 2003; World Bank 2007, 204). These studies show that qualitative research is very useful in helping researchers reveal and understand some of these biases and stigmas, which are hard to capture in quantitative surveys. Ignoring these stigmas and not acting upon them through careful wording and probing can introduce very substantial biases in the estimation of people in wage employment and in the correlation between poverty and status in employment. Therefore, at best, the officially recorded share of ‘employees’ (wage employment) is probably a very low boundary for the total number of workers who engage at least occasionally in wage labour.
Fifth, the terminology is also important. Some analytical and statistical categories may be difficult to comprehend by enumerators and respondents, especially when terms like ‘salary’ or ‘wage’ are associated with formal sector ‘well remunerated’ employment such as civil servant jobs, or teacher jobs. This implies that any question that uses these terms is likely to yield biased responses about participation in local wage labour markets. In general, notions of ‘employment’, ‘gainful activity’ or ‘remunerated activity’ are not unproblematic in rural African contexts (Bardasi et al. 2010, and for urban areas see Luebker 2008).12 Many activities may not yield cash income and some forms of wage labour paid in kind (with food) may typically be regarded as a source of ‘help’, whereas the mention of ‘salary’ or ‘wage’ depending on the local terms may be automatically associated with regular stable wage employment. As Hussmanns et al. (1990, 256) argue, ‘the concept [of employment] is complex and interviewers’ or respondents' own subjective understanding of terms like “economic activity” or “work for pay or profit” may differ from what the concept intends to include'. Besides, there are many specific local terms for various forms of wage tasks which change from place to place. If enumerators don't use these locally recognisable terms and instead rely on literal translations, agricultural wage employment may be seriously under-reported.
Rather than just asking respondents if they have been ‘employed’ during the reference period, it is thus important to operationalise the concept of employment through questions that are easily understood by respondents. For example, in one of the micro-studies referred to in the next section (in Mauritania), the test of alternative methods to estimate labour force participation rates led to disparate results especially for one village. If a question about ‘remunerated employment’ was used (for people aged over 14), the resulting ‘employment-to-population ratios’ were very low: around 50% for two villages and only 26% for the most problematic village, precisely the one where more economic dynamism had been observed by the team. Instead, when ILO best practice was applied and respondents were asked to report and list all the economic activities that brought in some income or that occupied their time on the basis of an activity list (‘did you do any of the following?’ etc.), employment-to-population ratios increased significantly, in the more problematic village (to 70%) and the others (to 85%), more in line with reported labour force participation rates in rural Africa. The results were then triangulated with in-depth qualitative research to understand relevant concepts of labour participation and respondents' perceptions of the meaning of ‘employment’. Qualitative evidence is indeed crucial to help in the design of adequate quantitative labour surveys.
Recent experimental research carried out by the World Bank has provided additional compelling evidence about the sensitivity of labour market statistics, especially labour participation rates and wage employment, to survey design (Bardasi et al. 2010; Dillon et al. 2012). Thus Bardasi et al. (2010), after testing designs with shorter vs longer employment modules, find significant differences across survey designs and conclude that:
Our findings suggest that both types of survey design decisions have statistically significant effects on labor statistics. These effects are largest on the measure of labor force participation, but also exist for weekly hours of work, daily earnings, main activity, and type of work… Using the short questionnaire lowers female labor force participation and also affects the distribution of workers across sectors, lowering the share of paid employees among the employed. (Bardasi et al. 2010, 31, emphasis mine)
Defining households and their economic boundaries
Another source of bias derives from the definition of sampling units, i.e., ‘households’. In a context of significant personal mobility, the definitions of ‘household’, its ‘residents’, and of the economic ‘boundaries’ of the village or ‘community’ are problematic. A residential definition of the household,13 typical of large-scale representative household surveys, may fail to capture very relevant household members who do not reside (or only sporadically) in the main residence but who may be significant contributors to the expenses of the households or dependants from its income sources (Beaman and Dillon 2012). ‘Footloose labour’ constantly on the move in search of jobs, or workers sleeping in work dormitories (also known as labour camps) and temporary accommodation next to their employers' premises are routinely missed out in conventional household surveys, because of how the primary sampling units and principal respondents are defined and found (Sender et al. 2005; Breman 1996; Akresh and Edmonds 2010). Cramer et al. (2011, 18) discuss in detail the unreliability of standard sampling frames, lists of households provided by village chiefs and warn about the danger of missing out large groups of people without a fixed residence, who live in temporary accommodation and are highly dependent on casual wage labour, ‘leading to a serious underestimate of the degree to which poor rural households depend on income derived from wage employment’. An alternative to this, as articulated in much of the research cited in the following section and especially in Cramer et al. (2011), is to rely on fresh censuses of ‘residential units’ where individuals (related or not) sleep (which would include ‘conventional’ households too) and collect information on all ‘economically-liked’ individuals so as not to miss out relevant household roster members who may never or only rarely reside in the observed unit. All this requires the complementary use of qualitative techniques to inform the choice of research sites and the definition of residential units in each specific context.
Empirical illustrations of rural labour markets in Africa
This section presents few key selected themes from a number of micro-level studies in which I was involved, in order to briefly illustrate how alternative survey methods and conventions help us capture the incidence, complexity and various dimensions of rural wage employment.14 Due to space constraints, findings are selected on analytical and methodological grounds following issues discussed in previous sections. In particular, the findings presented here concern: (a) the nature and incidence of rural wage labour relations in diverse African contexts; (b) the diversity of forms of rural wage employment; (c) the significance of fragmentation and segmentation in rural labour markets in Africa, as well as the effects of migration on segmentation; and (d) the gendered nature of rural labour relations.15 The studies in question include:16
• A very large rural wage employment survey of over 2600 wage workers and their households in rural Mozambique in 2002–2003, designed to have an over-representation of female labour (almost half the sample) to capture gender aspects of rural labour market participation and to cover a wide range of farm and non-farm activities in rural areas in three populous provinces of Mozambique (see Cramer et al. 2008; Oya and Sender 2009; Sender et al. 2006).
• A study done in rural Mauritania in 2004–2005 with the Government of Mauritania to examine in greater depth the nature of village labour markets and rural livelihood diversification. This work was based on three in-depth village studies, representing three of the most distinct agro-ecological and socio-economic contexts in semi-arid Africa (GIRM and World Bank 2007; Oya 2007; Pontara 2010).
A key feature of these studies was their use of mixed methods and their avoidance of the key biases discussed in the previous section. While the focus was on the design and implementation of quantitative sample surveys, in-depth qualitative research was carried out both before and after running the employment surveys. Careful qualitative scoping research was critical to better design the survey instruments (questionnaires and notebooks) and the sampling process. Then ex-post qualitative research shed light on aspects that could not be captured through the quantitative surveys, particularly on barriers to women's participation in the labour market (Oya and Sender 2009), employment histories, abuses at the workplace and in the household, and various aspects of labour market segmentation.
Absent/thin rural labour markets? The significance of wage labour in rural Africa
A key lesson from years of field research on rural wage employment is that methodologically innovative, context-sensitive surveys of rural employment with a specific focus on rural labour markets and aided by qualitative research manage to capture a variety of forms of rural wage employment – including the most demeaning and oppressive forms – relatively accurately. In the studies listed above, there was direct and indirect evidence of the quantitative and qualitative significance of rural wage employment and rural labour markets in different agro-ecological and socio-economic contexts. The studies were not designed to obtain nationally representative data but sampling was relevant to the research questions and carefully justified according to clear analytical criteria (see Cramer et al. 2008 and 2011 for more elaboration); and qualitative evidence corroborated the significance of wage employment and tendencies towards greater reliance on rural labour markets among generally very poor respondents.
So, do rural/village labour markets exist? Is labour hiring common or only a marginal occurrence? Perhaps one of the toughest tests was provided by the study of villages in remote parts of Mauritania where conditions are similar to those usually associated with absent or thin labour markets in the view of Barrett et al. (2005) and Binswanger et al. (1989). And the test showed that in fact even in such contexts labour hiring is much more common than expected, a finding that is consistent with previous studies of labour relations in other parts of rural semi-arid Africa (see Kevane 1994, for example; GIRM and World Bank 2007; and Table 4). Moreover, the implication is that labour hiring in rural Africa is not simply limited to rare instances of large-scale commercial plantation agriculture (only one case in the Mauritania study), but rather a fairly common phenomenon in smallholder farming and across different socio-economic groups engaged in a variety of activities, as Table 4 shows. As a result, the diversity of forms of wage employment in rural areas is striking (Cramer et al. 2008; Oya 2007).
| Socio-economic group | Percentage of sample (%) | Socio-economic status – ratio to average asset index | Percentage of households hiring in labour (%) | |
|---|---|---|---|---|
| 1 | Landlord or prosperous farmer, employer | 3 | 1.86 | 100 |
| 2 | Middle peasant farmer (mostly hiring in labour) | 12 | 1.55 | 92 |
| 3 | Small peasant farmer (both hire-out and hire-in or no hire at all) | 32 | 0.70 | 29 |
| 4a | Agric wage labourer and some farming, mainly wage employed (poorest) | 7 | 0.68 | 14 |
| 4b | Landless agricultural labourer, only wage employed | 5 | 0.86 | 0 |
| 5a | Trader/transport/artisan, employer (wealthier) | 6 | 1.56 | 100 |
| 5b | Trader/artisan/transport petty, self-employed (poorer) | 12 | 0.82 | 13 |
| 6a | Salaried (non-agriculture), wage employed and employer | 5 | 1.45 | 48 |
| 6b | Non farming, non-agricultural wage employed | 6 | 0.91 | 50 |
| 6c | Farming, self-employed and non agricultural wage employed | 12 | 1.01 | 48 |
| ‘Capitalist’ classes (1, 2 and 5a) | 21 | 1.60 | 98 | |
| ‘Labouring’ classes (3, 4, 5b and 6) | 79 | 0.84 | 49 | |
| of which, mostly self-employed (3 and 5b) | 44 (of total) | 0.73 | 25 | |
| of which, mostly wage-employed (4 and 6) | 35 (of total) | 0.97 | 35 | |
| Total | 100 | 1.00 | 41 | |
| Source: Author elaboration from Rural Labour Market Survey in Mauritania. Also Oya and Pontara (2008) and GIRM and World Bank (2007). | ||||
| Technical note: Total sample of 200 households in three villages representative of three different rural contexts in Mauritania. |
Data on participation in rural labour markets in the Mauritanian and Mozambican studies (both through questions on hire-in and hire-out) therefore suggested that official statistics on rural wage employment underestimate its extent for many of the reasons explored in the sections above (GIRM and World Bank 2007; Pontara 2010; Cramer et al. 2008; O'Laughlin 2002; Wuyts 2011).
Diversity in rural wage employment, segmentation and power
Quite apart from the quantitative significance of rural wage labour, especially for some of the poorest segments of rural Africans, exploring the observed heterogeneity of rural labour markets and wage employment patterns across and within countries was a central aspect of these studies. Despite the generally irregular and poorly remunerated nature of most rural wage jobs, wage workers in rural Africa undergo a wide range of experiences and situations, so, for example, a clear-cut relationship between poverty and participation in rural labour markets cannot be easily established (see also Winters et al. 2008 and Table 4 in this article). One reason is the very fact that a ‘pure’ landless proletariat only dependent on wages is hard to find. Most rural people hold a variety of jobs, including own-account farming, though a majority of wage earners are more dependent on food purchased in the market than on own production. Relative dependence on wages is thus consistent with a range of self-employment activities. These activities are always common especially in contexts of strong seasonality, risk and where jobs are usually available on an irregular basis. For very poor wage-earning workers self-employment is often ‘residual’ and a reflection of remnants of resistance to proletarianisation and the unpredictability of employment.
Another important reason and indeed a stylised fact emerging from the studies mentioned above is the high degree of labour market segmentation, manifested in the diversity of rural and village labour markets in Africa and variation in working conditions. The studies covered very different regions, from arid to semi-arid to sub-tropical, spanning a wide range of crops (groundnuts, cereals, tea, cotton, tobacco, horticulture, etc.) and activities (farming, trade, artisanal crafts, transport, domestic service, construction, etc.). Evidently some of the differences found in working conditions and wages have much to do with the specific activity, task and crop involved. In other words, sector/activity segmentation is an important feature of rural labour markets. In agriculture, significant differences in wages and working conditions (including forms of payment) between crops and by task, were observed particularly in Mozambique, but also in Mauritania, where labour-intensive crops, especially in horticulture, normally commanded higher wages for equivalent tasks than staple cereals. Reasons for these crop-specific features could be the type of employer involved in each crop and the significant differences in productivity across crops, locations and among employers, which result in differential ability to pay higher wages.
Of course, segmentation along activities and tasks partly emerged because of a range of barriers to entry into these different activities, often defined in terms of skills and ‘aptitude’ (as in the frequently mentioned superiority of women in doing ‘careful’ work like tea plucking), but frequently also socially and culturally determined, especially for what regards specialisation along gender lines. Barriers to entry and skill specificities also correlated with class ‘locations’ (of the kind proposed in Table 4), in the sense that the poorest rural wage workers were often restricted to a narrower range of low-skill, very poorly paid and casual manual occupations. In fact, domestic workers, casual (especially female widowed) agricultural workers, and petty vendors working for other traders were among the poorest workers in our samples in Mozambique and also in Mauritania.
The studies also highlighted the importance of employers' discretion and power relations at the workplace as correlates of varying working conditions. For similar occupations and tasks, the Mozambique survey found significant differences in payment methods and wage rates across employers. Strikingly, some of this evidence pointed at discretion used by a single employer to discriminate among his own workers (see Table 9 in Cramer et al. 2008), but, overall, a pattern emerged whereby smaller, resource-poorer employers (e.g., small-scale farmers and small traders) would offer worse working conditions in comparison with larger-scale, more technologically dynamic and productive employers (usually large plantations, sometimes foreign-owned, featuring greater crop specialisation and strong links with global markets).17 Therefore, a scale bias operates in rural labour markets, and especially agricultural wage workers seem better off in larger-scale, more organised faming units (Cramer et al. 2008; Sender et al. 2006). This is far from surprising since the levels of productivity of these farms are often superior to smaller-scale counterparts so they can afford to pay higher wages. They are not inherently ‘nicer’ to their workers. Moreover, larger-scale, especially foreign-owned, agribusinesses are much more exposed to monitoring and inspection by local authorities, trade unions, non-governmental organisations (NGOs) and different advocacy groups with an interest in labour conditions and globalisation. They have more to lose if working conditions in their business are intolerable. In contrast, smaller-scale farmer–employers often fall out of the radar of agencies and institutions that monitor conditions of wage employment. In fact, the assumption that small farmers only use family labour often precludes any serious consideration of what is happening to their casual workers.
The studies described in this section were not limited to workers with similar characteristics (i.e., very poor, low-skilled and performing manual agricultural tasks), but surveyed a range of occupations and jobs, which allowed to consider a wide continuum of working conditions – from very bad to relatively good (or ‘decent’) jobs – in rural areas. At the heart of this continuum of heterogeneity is the extent to which rural jobs are more determined by either pull (incentives) or push (distress) factors. Not all rural jobs are examples or manifestations of ‘trickle-down’ in a simple virtuous sense. Nor are all coping mechanisms of last resort. In all the studies the poorest workers, who normally also corresponded to the poorest quintiles in the national household representative surveys in terms of simple asset or other welfare indices (see Sender et al. 2006; Cramer et al. 2008), generally depended on casual wages in jobs performed for local neighbours, either in farming or personal services. Many of these jobs were objectively and subjectively (i.e., in workers' perceptions) purely ‘distress’ activities. They were seen as ‘last resort’ and partly contributing to growing local inequalities and acceleration of differentiation (see also Bryceson, 2006, on similar occupations in Malawi). Some of the worst jobs could also be found in non-agricultural occupations. Reardon (1997) finds that for a range of micro-studies, non-farm incomes can be five times higher than farm wage incomes, but fails to offer more detailed disaggregation within non-farm occupations. Cramer et al. (2008), in contrast with Reardon's (1997) findings that on average non-agricultural jobs are much better paid than agricultural employment, generally disaggregate non-agricultural jobs and find examples in the rural non-farm economy that are reserved to the poorest segments of the rural population, with a strong gender and age bias (i.e., mostly girls and sometimes young boys). These include domestic workers, hairdressing apprentices, market porters, street vendors, fare collectors and so on.
At the other end of the continuum, some of the better agricultural and non-agricultural jobs could be seen as mechanisms of livelihood improvement and ultimately escape from poverty. This is particularly the case for some women wage workers in Mozambique (see below). Workers, particularly women, with access to better jobs were indeed also more ‘empowered’ and confident in joining collective action (strikes) and unions wherever these were present (Sender et al. 2006). Sometimes the availability of better job opportunities in their local area meant an opportunity to delink from highly exploitative relations with other local employers who hitherto effectively exerted their monopsonistic power to impose very bad working conditions.18 Overall, however, much of the research undertaken in these countries showed the weakness of collective action and organisation for poor rural wage workers, both in terms of the lack of presence of unions and the weakness of government institutions dealing with the monitoring of labour law implementation (e.g., the General Inspectorate of Labour in Mozambique).
More generally, access to ‘better’ jobs in agriculture and non-agriculture (on longer contracts, with higher wage rates and generally more non-wage benefits) was mildly correlated with supply-side characteristics, such as education levels and skills acquired through previous employment experience (see Cramer et al. 2008, especially Tables 12–14). However, the fact of being in an area characterised by more dynamic agriculture, more investment and tighter labour markets made a crucial difference. In other words, favourable individual and household characteristics are not enough. Labour demand is a key factor, which underscores the critical importance of incentives to create nodes of accumulation to increase labour demand especially for better jobs that may be accessible to poorer workers (Oya and Sender 2009). In any case, even the scope for improving the quality of ‘good’ rural jobs in Mozambique was significant. In fact, median wage rates for ‘better’ jobs in Mozambique were still below the national statutory minimum wage and most of these jobs were of an irregular or seasonal nature.19 Very few rural wage workers were employed on a permanent basis and most, therefore, were not protected by existing labour legislation, as employers often manipulated the length of contracts to avoid applying labour laws (this included many of the foreign large-scale employers). In sum, the heterogeneity of rural wage jobs reflects context-specific patterns of employer–worker relationships, and result in a continuum between distress and livelihood improvements, which can be best captured with a lot of fine-tuning in survey design and judicious use of qualitative evidence.
Rural labour market fragmentation and migration
As noted above, an important form of segmentation is location, which is also due to the uneven distribution of wage labour demand in agriculture and non-farm activities. The literature on rural labour markets suggests that the village is indeed a relevant unit of analysis, i.e., that rural labour markets are fragmented by village, because high transaction and mobility costs, exacerbated by poor transport infrastructure, impede a smooth articulation of rural labour markets and constrain labour flows (Mduma and Wobst 2005; Kevane 1994; Rao 1988). This feature is likely to be more common in poorer rural areas than those where agricultural commercialisation and modernisation have proceeded more rapidly. Our research in Mauritania offered some evidence in support to this hypothesis. Wages and working conditions varied, especially between villages – wage rates acting as local ‘norms’ regardless of relative labour abundance or shortages or how competitive rural labour markets may seem at first glance, as reported in many contexts worldwide (Turnham 1993).20 These local village-specific ‘norms’ often applied to the poorest segments of the wage-earning classes, indicating collusive behaviour among the few, more powerful, village employers who offered some occasional employment on their farms, houses and petty businesses. The compliance with these ‘norms’, especially in rural Mauritania, also reflected the entrenched personalised power relations at village level, where these payments were often portrayed by their employers as ‘help’ or even as alms, part of their duties towards the local poor, in a typical manifestation of employers' paternalistic practices and discourses, and often not far from forms of labour bondage.
However, in more dynamic contexts, especially where linkages with markets were stronger and where ‘outside’ investors, national or transnational, appear, there can be a breakdown of local wage ‘norms’. This was observed in rural Mozambique in 2003, where the typical wage rate paid by local farmers to casual workers for various agricultural tasks (MZM 10,000 per day – US\(0.43) was below that offered by newcomers (around MZM 21,000 or US\) 0.90, i.e., more than double the conventional local rate), more preoccupied with attracting sufficient workers than with respecting local wage ‘norms’.
The presence of migrant labour also contributes to segmentation by creating fractions of working classes paid and treated differently despite not being very different in terms of skills, experience and productivity. Migrant labour is often used by agricultural employers as a tool to depress wages and create different classes of workers, thereby undermining collective action (Breman 1996; Standing 2006).21 This can happen for a wide range of employers, including large-scale agribusiness, as the history of sugar plantations in Mozambique shows (O'Laughlin 2002; Head 1980). The fast development of tobacco farming in Central Mozambique also owed much to inward-migration, this time of both employers and former farm workers from Zimbabwe (Hammar 2010), with the irony, as it was found in our research, that many of the latter were actually Mozambican return migrants, having previously moved to Zimbabwe during the Mozambican civil war.
In Mauritania, the effects of migration on rural labour markets were nonetheless contradictory. Whereas in one of the villages, close to the border with Senegal, it was usually reported that migrant workers from Senegal worked for lower wages and depressed wages below local ‘norms’, especially in rice production and horticulture, in the oasis-farming environment of another very different village, daily wages increased with the seasonal wave of migrants coming to work on date harvesting (GIRM and World Bank 2007). Most of these migrants were coming from other regions and sometimes from towns, so their reservation wages were higher than those of local, very poor workers. Their skills and the importance of timing in these operations meant that local employers would compete for migrant labour and offer higher rates or better conditions (e.g., better meals and accommodation in what is normally a very hostile environment, especially during the Guetna season, devoted to date harvesting). In addition, the Mauritanian study also gave evidence of a virtuous circle of causation between out-migration and employment creation, which in turn fuelled the use of migrant labour in agriculture, particularly in areas with access to irrigation and more labour intensive crops. Thus remittances from urban-based relatives were often used to expand agricultural operations, irrigate fields and therefore increase the demand for labour, which, interestingly, often originated from different villages or even across the border from Senegal.22 In sum, migration does contribute to rural labour markets segmentation, albeit in a somewhat contradictory variety of ways, depending on local circumstances, relative labour shortages, local power relations and the scale of migrant labour.
Gendered rural labour markets
Perhaps more straightforward is the segmentation produced by gender relations and the nature of women's participation in rural labour markets. The research in Mozambique (see Oya and Sender 2009) paid particular attention to rural women wage workers. A striking, but perhaps not surprising finding was that a large purposive sample of rural wage workers (especially agricultural workers) found that a very high proportion (around 40%) of women workers were divorced/separated or widowed, i.e., living in de facto female-headed or female-dominated households. This finding is consistent with research in other parts of the world (see Dreze and Srinivasan 1988; Kabeer 1997) and previous research in Tanzania (Sender and Smith 1990). The interpretations of such findings relate to the dichotomy distress vs. emancipation mentioned above and point to the complexity of gender relations, patriarchy and labour market participation in developing countries. In the research carried out in Mozambique, it was clear that patriarchy and paternalistic control were significant determinants of women labour supply in rural labour markets. This happened in two ways. First, husbands and fathers were preventing women from engaging in any form of wage labour outside the household in an imposed form of resistance to ‘proletarianisation’. Second, they prevent access to particular types of jobs, especially in large-scale workplaces where contact with many other men was particularly feared. In other words, women's labour market participation and access to particular jobs were constrained and shaped by patriarchal power and by the bargaining of women within existing ‘patriarchal bargains’ (see Kandiyoti 1988).
Reproductive stories, and especially childlessness or lack of sons, could also be determinants of relationship break-up and consequent women's engagement in wage employment to survive. However, one should avoid excessive determinism about these relations as a variety of patterns was observed that could defy generalisations. In a similar vein, despite the fact that women workers were generally discriminated against in terms of the type of job and working conditions in comparison with men after controlling for education and age, it is also true that the sample of women wage workers was quite heterogeneous and reflected the extent to which women in rural Mozambique were not necessarily locked into low-quality highly exploitative agricultural jobs (Oya and Sender 2009). As noted above, the particular location patterns of rural wage labour demand determined the opportunities available and provided options for women to emancipate through access to more regular and better paid wage jobs. This underscores the importance of macroeconomic and sector policies to boost demand for unskilled (female) rural labour.23
Conclusions
This article has explored the gap in knowledge and statistical evidence about rural wage employment in Africa. It has been argued that evidence on rural wage employment is extremely scarce, fragmented and of dubious quality for most of sub-Saharan Africa. The article has presented some arguments and hypotheses as to why this is the case and suggested ways of overcoming the methodological challenges of collecting data on rural labour markets, based on extensive field research experience in several African countries. The last section presented illustrative evidence of the kinds of findings and themes that arise from carefully designed and context-sensitive rural labour surveys based on mixed methods, but primarily on quantitative sample surveys. These themes and data are critical for an understanding of the political economy of poverty, accumulation and employment in rural Africa. Understanding rural poverty and agrarian change requires an understanding of labour dynamics, the complexity of which can only be captured with serious attention to the quality of survey design, in terms of sampling procedures, definitions, coverage and questionnaire design. The research experiences described in this article demonstrate that the use of qualitative research to improve quantitative data collection in this field is essential. Lessons from the micro-studies briefly presented here and many other attempts to capture the complexity of rural labour market dynamics should be taken seriously by statistics agencies and funding bodies. The World Bank's own work with survey experiments (e.g., Bardasi et al. 2010; Dillon et al. 2012; Beaman and Dillon 2012) is a good step in this regard. In fact, more survey experiments are needed to help improve labour market statistics in developing countries and adapt conventional categories to the needs and constraints faced in those contexts. A challenge is whether ideological fetters will continue to obstruct a pro-poor statistical agenda that centres on labour. The myth of self-employment and independent agricultural producers is so generalised and entrenched among academics, practitioners and aid officials (Wuyts 2011) that it is hard to envisage a dramatic change in focus. In an era still dominated by a mixture of market neo-liberalism and pro-peasant neo-populism, the place of rural wage labour in African studies and statistical agendas remains in doubt. Nevertheless, political economists interested in rural and agrarian dynamics in Africa are urged to contribute to boost the hitherto marginal(ised) work on employment and rural labour relations.
Note on contributor
Carlos Oya is Senior Lecturer in Political Economy of Development, Development Studies Department, School of Oriental and African Studies, University of London. He has published on various aspects of agrarian change, agrarian capitalism, rural poverty and labour markets, mostly on Africa, particularly in Mozambique, Senegal and Mauritania, from an agrarian political economy perspective. Current research interests also include the impact of certified standards on agricultural wage work, as well as the political economy of aid and policy space in Africa.