In popular and contemporary debates, ‘poverty’ and ‘Africa’ are often presented as intimately linked. Yet, this direct association is relatively recent.Footnote 1 In the 1970s, when A.G. Hopkins wrote his Economic History of West Africa, the myth he first had to dispel was that of ‘Merry Africa’.Footnote 2 By contrast, in 2009, he reflected on how Africa was depicted as a continent of poverty, as illustrated by the ‘Make Poverty History’ campaign of 2005.Footnote 3 Economic analysis of African economic growth and poverty is based on the underlying assumption that growth is a chronic failure, and poverty is persistent. This stands in contrast to the work of Africanist historians, who have long argued that peasant producers and workers were relatively prosperous between 1940 and 1970.Footnote 4 Nor is mid-twentieth-century prosperity completely unknown to economists. François Bourguignon and Christian Morrison found using a large database of compiled household surveys that ‘in 1950, 12 percent of world inhabitants with incomes less than half the world median lived in Africa. By 1992, 30 percent did’; yet these changes in poverty over time have had little influence on how the history of poverty in Africa by numbers is told.Footnote 5 At one level, something as basic as the historical vantage points play a role here. Recent events loom large and frame research questions. Therefore, although relative prosperity in Africa was commonplace until the 1970s, it has proven easy to lose track of such comparative data points given the dominance of ideas such as the ‘African growth tragedy’ or the ‘lost decades’ of the 1980s and 1990s.Footnote 6
Another main contributing factor to the ahistorical approach to poverty is the limited availability of quantitative evidence on poverty. Dominant research questions are motivated by the availability of data points, and what constitutes relevant data points has been limited in the mainstream economic analysis to World Bank surveys of poverty.Footnote 7 Once we widen the scope of evidence, however, the approach to understanding poverty in Africa changes considerably.
A longstanding complaint among economic historians is the failure to see the potential of the African economic past to inform contemporary economic analysis.Footnote 8 Patrick Manning warned in the 1980s that ‘future work in African economic history will be conditioned fundamentally by the limits on available data, and by the success or failure of projects to generate or systematize further data’.Footnote 9 While Manning did not subscribe to a narrow interpretation of data, it seems that in this case he was referring to datasets that are easily accessible to scholars, and can therefore stimulate further work.Footnote 10 In the same survey article he suggested that the key reason why a longer view of African economic history did not inform economic analyses was precisely the lack of such datasets.
The key contribution of this article is to present new data that has been published and collated in the past decade and to discuss how it changes our interpretation of poverty in Africa in the twentieth century. Before doing so, the paper briefly assesses the conventional scholarly consensus about poverty rates and levels in sub-Saharan Africa since the 1980s. This is important because this literature incorrectly overrates the quality of contemporary evidence and underrates the relative quality of historical evidence. This false appreciation of relative data quality has, amongst other things, meant that the history of poverty by numbers is very short, and more generally the lack of long time-series data on growth and poverty has caused what Gareth Austin has called a ‘compression of history’.Footnote 11 The second half of the paper discusses the responses of economic historians of Africa to the economic literature which has highlighted some historical factors shaping contemporary outcomes. First, however, it is worthwhile to diagnose the knowledge problem that has shaped the approach of economists to the history of poverty in Africa.
HISTORICAL AND CONTEMPORARY DATA ON POVERTY AND GROWTH
The availability and presumed reliability of poverty numbers and economic statistics have shaped the historical analysis of poverty and growth in sub-Saharan Africa. Quantitative economic histories are written using time-series data on economic growth. In this case, ‘year 1’ is typically considered to be 1960, and hence economists have focused on the postcolonial period.Footnote 12 This temporal focus is due to the fact that most databases, such as the World Development Indicators maintained by the World Bank, only contain data starting in 1960. While reporting of National Accounts in industrialised countries goes back to before the Second World War in many cases, a near global systematic coverage only became available starting in the 1960s.Footnote 13 The first compendium that contained GDP data for all countries published by the World Bank, the World Bank Atlas, was published in 1966. It contained two estimates for each country: its population and its per capita gross national product in US dollars, both for 1964.Footnote 14 However, in conventional histories of poverty by numbers, using the dollar-a-day metric, ‘year 1’ has been 1990. The dollar-a-day definition of global destitution made its debut in the World Bank's 1990 World Development Report.Footnote 15 The Povcalnet database, maintained by the World Bank, collects the poverty survey data, dating back to as early as 1985 for Côte d'Ivoire. Generally, however, the baseline of 1990 has been considered the starting point since there are very few data points before the 1990s, and also because the United Nations chose 1990 as the baseline for the Millennium Development Goals.Footnote 16 Both 1960 and 1990 are, of course, artificial starting points, but data availability has tended to restrict scholars to this time frame, and therefore limited what kinds of questions can be investigated.
The narratives of economic growth in African economies would change dramatically if we reject 1960 as the starting point of the history of economic growth in Africa. Notions such as ‘chronic growth failure’ and the ‘bottom billion’ would not hold if we dispensed with this short time horizon.Footnote 17 Similarly, narratives of trends in living standards in Africa are shaped by the current configuration of what constitutes poverty knowledge at the World Bank. The short history of poverty by numbers in Africa is misleading because it fails to contextualise the history of poverty in the 1990s and 2000s within a longer-term perspective.Footnote 18
The dearth of immediately available data on trajectories of poverty has led researchers to take short cuts when studying long-term historical roots of poverty in Africa. Rather than carefully outlining the trajectory of economic growth or poverty over past centuries, the economic literature is content to treat current-day GDP per capita as a stand-in for a lack of growth in the past. Thus, according to this method, correlating contemporary income distribution with some explanatory variable in the past is sufficient to explain the historical roots of Africa's relative poverty.
This ‘new African economic history’, as A. G. Hopkins coined it, was claimed to be ‘causal history’ by James Fenske. Fenske contended that ‘practitioners of causal history focus on identifying causal historical relationships, which sets them apart from the qualitative, “old” economic historians and from much of the “new” economic history’.Footnote 19 Presenting new econometric history as a causal history and setting aside old, non-econometric history as non-causal amounts to a blanket statement of methodological superiority. Rather, the methodologies of economics and history are both necessary to grasp the causes of poverty.Footnote 20 Privileging one methodology over the other results in misconceptions, leading us down a road where we actively ignore knowledge. One prominent example of primarily qualitative research is John Iliffe's work, The African Poor: A History, which unearthed historical causes of poverty.Footnote 21 Iliffe traced and defined the different forms that poverty has taken through time and space in Africa. In this interpretation, poverty is multidimensional and thus defies simple quantification and advanced econometric testing.Footnote 22 Further, Iliffe identified two types of poverty: structural poverty and conjunctural poverty, arguing that each type of poverty has different causes and trend-lines. Conjunctural poverty relates to time and place and is rooted in economic shocks such as famine and war. On the other hand, structural poverty relates to an economy's systemic poverty, a figure that varies with deeper changes in the political economy. This nuanced understanding of poverty is lost in the quantitative methods used by scholars of the ‘new economic history of Africa’. Iliffe used historical observations of poverty in Africa to identify plausible patterns of causality but did not apply tests of statistical causality to his theories. This is a pragmatic solution to dealing with the nature of the evidence. If the usefulness of evidence is determined not by what it tells us, but by how it can be statistically tested, then scholars risk discarding credible information.Footnote 23
Iliffe highlighted the inadequacy of numerical sources as one of the key obstacles to the study of African poverty. Cautioning against an overreliance on oral traditions and generalised descriptions, Iliffe suggested that records of poverty in documents such as missionaries’ letters, travellers’ journals, and administrators’ reports might be the next best alternative.Footnote 24 To that list one could add that information can be obtained through the meticulous study of population censuses from the colonial period.Footnote 25 Rhiannon Stephens has also shown how oral sources can complement linguistic and archaeological evidence.Footnote 26 Citing a lack of written data, especially prior to 1860, Stephens uses a mixture of these sources (and others) to help reveal the social, political, and economic history of the African poor. Such endeavours will become less and less accessible to us if our study of poverty is limited by and to numbers.
The study of historical trajectories of growth and poverty in sub-Saharan Africa has been limited by the databases. First of all, there has been a failure to appreciate the multidimensionality of poverty, with today's reductive approaches which commonly use the term ‘poverty’ as a shorthand for a very specific monetary measure, such as the World Bank's world poverty headcount. In principle, the poverty headcount is a universal measure of how many people (either in absolute terms or as a proportion of the population) have insufficient income to cover a basket of subsistence goods.Footnote 27 Of course, if the poverty problem had been conceived as a relative measure, reflecting in-country inequality, it would have produced a radically different global discourse on poverty.
The use of numbers in social sciences is usually justified because it adds rigour and reliability, but on closer inspection the poverty numbers are often not as credible as they are presented to be. The global numbers are based on small or non-existent samples at the local level. Household surveys covering a few thousand households are conducted infrequently and in only some of the countries.Footnote 28 Indeed, the ability to collect consistent survey data on poverty is constrained by problems with both design and implementation.
Using current methodology assessing poverty levels in any given country would take approximately one million US dollars and two to three years from initiation to completion.Footnote 29 Once the survey data is available, the first step is to use foreign exchange rates to express the income of one country in the currency of another. However, differences in domestic prices of non-tradable goods cause a divergence in purchasing power parity (PPP).Footnote 30 To achieve PPP, one needs to adjust for the fact that one dollar goes a lot further in a country like Burundi than it does in a country like Belgium. This entails a complicated process of collecting local prices, weighting them appropriately, and finally constructing a basket of goods and services for each country in order to compare their costs against one another.Footnote 31
Since 1970, there have been eight rounds in the International Comparison Program that collects the price data needed for the PPP conversion, with the most recent in 2011. Participation from African countries has been limited in each of these rounds. In 1970 and 1973, only Kenya participated. In 1980, Kenya was joined by Malawi and Zambia, while the number of participating countries increased to cover more than half of the continent in 1993.Footnote 32 Nevertheless, only 11 of 48 sub-Saharan African countries have completed regular Living Standards Measurement Studies. Even if one were to widen the data to include other types of poverty surveys, as of 2012, only 25 of forty-eight countries had at least two surveys available for tracking poverty trends over the past decade. Of the twenty-three remaining countries, there is not enough information to determine the direction of change. For five countries representing 5 per cent of the sub-Saharan African population, there is no data for measuring poverty at all.Footnote 33 Beyond the limited availability of data, there are questions of sampling, survey design, and other errors in measurement. These factors all demonstrate the need to question the assumption that recent poverty numbers are of such quality that they merit the very prominent role they have in motivating research questions and shaping conventional wisdom about poverty.Footnote 34
Just as the mainstream study of poverty excludes alternative sources of information from consideration, so too does it shorten the history of poverty that is being analysed. A number-centric and particularly a monetarily-focused perspective on the history of poverty in Africa is especially limiting in that it explicitly leaves some information out. This is true from the perspective of econometricians, for whom the availability of quantitative evidence represents the boundary of investigation. Thus, one way of driving the literature forward is by expanding the quantitative horizon backwards.
EXTENDING THE HISTORY OF POVERTY BY NUMBERS: NEW EVIDENCE
Because historical numbers on poverty are scarce, many of the ‘historical’ papers by economists often use evidence from a very recent past.Footnote 35 However, a ‘renaissance of African economic history’ has stimulated a wave of new archival research aimed at measuring poverty and living standards particularly for the colonial period.Footnote 36 Observations are still preliminary and suffer from many limitations. Nevertheless, whereas the previous section has argued that our knowledge on poverty by numbers in the recent postcolonial period is overrated, the following will suggest that our knowledge of trends in ‘poverty’ or, at the very least, ‘living standards’ in the colonial period might be underrated.
a) Living standards and real wages
Real wages series calculate the cost of goods that workers consumed over a given time period and compare them to contemporary nominal wages, thus giving us a measure of the real change in living standards. Price data is compiled for a ‘bare-bones’ subsistence basket of goods — the absolute minimum daily requirements for an adult male.Footnote 37 Historical prices are compiled from a variety of sources to calculate how much the basket would have cost at any given time, and compared to nominal wages received by unskilled labour. Standardising income and needs in this way allow for a comparison of well-being across both time and distance. Frankema and van Waijenburg used this approach in their analysis of living standards in eight British African colonies from 1880 to 1940.Footnote 38 Specifically, they argued that real wages exceeded subsistence levels for this period and that wages in a selection of the colonies studied actually exceeded those in major Asian cities during the same time span. Further, the authors highlighted that a divergence of paths has occurred not only between Africa and comparable Asian locations but also within the African continent. They pointed to the different paths that led to present-day poverty and challenge the argument that colonial institutions and path dependency are at the root of poverty in African countries.Footnote 39 The central argument in the literature on colonial legacies is that the impact of colonialism set these countries on a path towards failure, and that political pressures made sure that they remained on this particular development path.
Frankema and van Waijenburg's paper specifically addressed what they perceive to be a central claim of the economic literature, namely that African poverty has been persistent over time because ‘structural impediments to … growth’ referred to in their paper's title.Footnote 40 Economists have suggested many root causes of the lack of growth, such as the impact of colonialism, the disease environment, or the slave trade, but the common assumption behind each of these is that Africa has failed. Yet, real wage data presented by Frankema and van Waijenburg paints a far more positive picture: former British colonies showed persistent growth in urban real wages from the 1880s into the 1960s. In fact, in some periods and some locations this growth was very rapid. In British West Africa annual growth rates typically averaged, but sometimes exceeded, 2 or 3 per cent. In East and Southern Africa, there was growth throughout, though at a lower rate of between 1 and 2 per cent annually.
Using similar evidence, though not the same subsistence baskets, Bowden, Chiripanpanhura, and Mosley compiled wage, poverty, and real wage data for three settler economies (South Africa, Zimbabwe, and Kenya) and three peasant economies (Ethiopia, Uganda, and Ghana), and compared long term trends in economic trajectories and levels of living standards.Footnote 41 Their main contribution is to challenge the thesis by Acemoglu, Johnson, and Robinson that higher levels of European settlement led to more productive institutions that in turn predicted better economic outcomes today. Bowden, Chiripanpanhura, and Mosley demonstrate that on the contrary higher levels of poverty, and in particular lower wages, were observed in the settler economies. One issue that has not been dealt with explicitly in the literature is whether low wages should be interpreted as a potential avenue for growth, or whether relatively high wages in Africa might be interpreted as an obstacle to economic progress.
Their investigation shows that real wages stagnated in the settler economies, whereas in the peasant economies (Uganda and Ghana) real wages rose markedly from the 1910s through the 1970s. These datasets allow for a long-term historical reading of poverty in Africa, raising questions about viewing poverty as a longstanding, structural problem in African societies.
The new data points are chiefly drawn from the colonial period. However, the main argument made here is also supported by Rönnbäck's analysis of pre-colonial West Africa.Footnote 42 Rönnbäck provided estimates of living standards using wages from commercial records in pre-1900 Gold Coast. He used the same methodology, calculating welfare ratios (nominal wages divided by subsistence basket cost) for canoemen, and concludes that living standards were similar to other parts of the world. Borrowing again from Iliffe, it may be more appropriate to view the high incidence of poverty in modern Africa as a conjunctural problem from the 1980s onwards, coinciding with the appearance of the much more influential World Bank databases.
b) Anthropometrics and living standards
Another method used to measure poverty draws on anthropometrics. Specifically, researchers have used mean adult height as a measure of well-being. Stunted adult height is a result of the deprivation of essential needs in childhood. Conditions of childhood poverty have a ceteris paribus effect of stunting growth, though commonly height reflects the physical quality of life more generally, incorporating childhood nutrition, housing, sanitation, and medical care.Footnote 43 Anthropometrics use average population height in order to pre-empt the critique that genetics impact height. When one aggregates to the population level, the influence of genetics is cancelled out, especially when comparisons are made over time within one population.Footnote 44 At the macro-level, it is suggested that the genetic pool can be assumed to be fairly constant given the relatively short period of time on which such studies typically focus.Footnote 45
The use of average height as a measure for quality of life has a variety of advantages. First, because height is an outcome (rather than an input, like income) of well-being, this method does not discriminate between different varieties of input (avoiding the need to standardise consumption baskets) or make any assumptions about basic minimum requirements. The primary point of interest is whether individuals’ nutritional and medical conditions were sufficient to meet their needs. Further, height measures are applicable to the diverse socioeconomic conditions of developing countries (hunters, pastoralists, subsistence farmers, urban labourers) and thus facilitate comparison across groups. Mean height is also sensitive to inequality due to the diminishing returns of nutritional input on height, redistribution from rich to poor will raise the mean population measure. Finally, height data allows for statistical insights into historical periods for which other measures of poverty are lacking.Footnote 46
Such methods are generally restricted to relative measures of well-being. Specifically, mean adult heights are compared across different birth cohorts. If one cohort has a greater mean height than another from the same location measured during an earlier period, well-being is assumed to have increased through an improvement in overall health, nutrition, and living conditions. The most readily available height records prior to regular survey data are typically found in documents associated with the slave trade or colonial armies. Of course, these samples are not random, and they may not always have been representative. A preference for physically fit (taller) slaves could skew the sample mean, and military enlistment was often voluntary, paid, and subject to minimum height requirements.
Given such concerns, it is important for researchers to have an intimate familiarity with their dataset and with the dynamics of the region(s) being analysed. Scholars have thus paid considerable attention to demonstrating that their datasets are either representative of the broader population or using econometric techniques that adjust the samples to make them representative. For example, in analysing Dutch recruits born in Ghana and Burkina Faso between 1800 and 1840, Austin, Baten, and van Leeuwen have suggested that because the soldiers in their dataset were mostly purchased from slave owners, bias associated with volunteering was unlikely.Footnote 47
Further, they found no evidence of a height premium and thus dismissed the idea that a sample of slaves would be taller on average than the general population from which it was drawn. This addresses two different, but related, concerns. First, that slaves from genetically taller regions would be overrepresented in the population and second, that a higher price (and therefore demand) would be associated with taller slaves from within any given population. In response to these concerns they draw on Eltis's research that found a ‘small or even negligible’ height bias between the height of slaves and that of the populations from which they were drawn.Footnote 48 Related to the first issue of regional bias, Eltis found no difference in price for slaves from relatively taller and relatively shorter regions — which would have been expected if there was a greater demand for taller slaves. With regards to the second issue of a height bias within a given population, the authors point to a normal height distribution among slaves from all regions. This supports the conclusion that no height bias existed as a skewed distribution would have been expected if one was present. The authors conclude that ‘slave raiders, and victorious armies, had an interest in capturing everyone who could move.’Footnote 49 After correcting for the effect of regional origin and minimum height requirements, Austin, Baten, and van Leeuwen compared average heights in Burkina Faso and Ghana with those in other regions of the world and found that these African countries are further behind the global norm today than they were two hundred years ago.
Meanwhile, Austin, Baten, and Moradi's analysis of regional inequality in Ghana from 1880 to 2000 found that living standards improved during colonial times and then regressed after the economic crisis of the 1970s.Footnote 50 Their data came from a variety of periods and sources. First, attestation forms from the Gold Coast Regiment facilitated the estimation of average heights for those born between the 1880s and 1920s. The authors used regression analysis to establish mean heights that were more representative of the population. They then compared these averages with data taken from the 1987/88 Ghana Living Standard Survey and Demographic Health Surveys (DHSs) from 1988, 1993, 1998/1999, and 2003 in order to evaluate the development of regional differences in Ghana over time.
Moradi followed a similar methodology in further refuting the notion that contemporary poverty is rooted in colonial legacies.Footnote 51 Also making use of colonial army records, he argued that Ghana and Kenya experienced significant improvements in health and well-being through the middle decades of the twentieth century as measured by average population height. Moradi compared improvements in this metric with data from other developing countries to demonstrate that Ghana and Kenya were outperforming comparable countries internationally between the 1920s and the 1970s.
It is worth emphasising that the observed increases in mean heights in populations living in regions and colonies that were experiencing increases in export revenues from the late nineteenth century into the 1970s provide strong corroborative evidence that underlines the positive historic trend observed in other indicators. Taken alone, one might see export growth and be concerned about trends in other ‘invisible’ sectors. If one sees increases in real wages, one might still ask about sampling and distribution. These findings together suggest a more positive and lasting impact of the ‘cash crop revolution’ than previously thought. However, work on distribution and income inequality is still in progress, and we will learn more as additional datasets on wages from Francophone and Lusophone Africa become available.Footnote 52
CONCLUSION
That international datasets on growth and poverty have shaped the literature and conventional wisdom is understandable. The perceived credibility and comparability of international numbers are mutually reinforcing forces. They are also far more accessible than the sources that recent revisionist researchers have used as their evidentiary base: colonial ledgers, commercial accounts, and army recruiting forms. Some of this newly employed evidence such as the colonial blue books, are now available online, contain standardised information, and are thus easier to use than before. One reason they have not been extensively utilised previously is that there was perhaps a justifiable worry that an exclusive reliance on such sources would essentially retell the story of progress purposefully created by colonial administrators, and thus it is with serious caveats that one writes African history from the imperial archives.
On the other hand, the lack of data from prior to 1960 has arguably led to a ‘compression of history’ which is equally skewed. Correlations between present-day poverty and historical causal factors have had a strong methodological appeal in economic studies. The apparent absence of extended time series data made it easier to write as if little or nothing happened between some long-ago point in time and the present day.
Of course, the quality of many historical numbers that we have on poverty is poor. Yet judged against the sweeping claims derived from current-day databases, the rigour in the historical data presented in this paper does not compare so unfavourably. When historians are faced with such methodological bias from the discipline of economics it may be tempting to respond by ignoring numerical evidence and favour alternative sources of knowledge. However, one should be careful not to draw a line in the sand between quantitative and non-quantitative studies of poverty or between quantitative methods and historical analysis in general. The dominant narrative of poverty in Africa is number-centric — it does not have to be that way. It is striking now to remember that not so long ago it was generally accepted that poverty was multidimensional, and even that the poor had a say in defining what constituted poverty — as exemplified by the World Development Report and the work of Chambers in the 1990s.Footnote 53 Such nuanced understanding was largely swept aside as the dollar a day measure was codified as ‘poverty’ as part of the Millennium Development Goals in 2000.
Partly conditioned by the availability of data, the overwhelming emphasis in the mainstream economic literature has been to explain the relative poverty of African economies vis-à-vis the West and other emerging regions. This has only been possible through the appearance of imagined or stylised facts about Africa. The first derives from the economic growth literature written from the 1990s onwards that focused on explaining slow growth in Africa. The imagined event that arose from this was that Africa had experienced a chronic failure of economic growth. The use of average growth rates in the postcolonial period (1965–1995) and the brevity of the time series utilised combined to mean that periods of rapid expansion in the 1950s, 1960s, and 1970s and again from the late 1990s into the 2000s and 2010s were ignored and explanations for growth in Africa were unduly coloured by the downturn that took place in the 1980s.Footnote 54 Such a perspective would not pass the test of time with inclusion of datasets — or at the very least strong qualitative information — documenting growth at least since the 1910s.Footnote 55 Instead of this more long-range account, the frame of ‘chronic failure’ was embodied in the phrase ‘the Bottom Billion’, and a first level of conflation arose: in the literature in the 2000s the task of explaining slow growth was gradually conflated with the task of explaining low GDP per capita ‘today’. Of course, at some level, low income ‘today’ could mean no growth ‘yesterday’ but that is a testable — though yet unproven — hypothesis. It has been shown to be a shaky foundation for a literature purporting to explain the historical roots of poverty. The second level of conflation was made possible by approaching poverty as ‘absolute deprivation’ so that high levels of poverty today is conflated with low GDP per capita today.
Some of the evidence reviewed in this article points to more recent stylised facts. One is that the ‘gap’ in living standards between African and non-African economies was likely lower a century or two ago. Corroborating evidence indicated that the experience of poverty was not a typical African phenomenon five decades ago. Thus, enriching the database with new historical data points offers other comparative frames than the ones prevalent in contemporary literature. Moreover, the study of longer-term trajectories in growth, poverty, wages, and other metrics capturing living standards indicates a long trend of improvement in most African economies from the 1910s through the 1970s.
This raises the possibility that the 1980s, and the narrative of African poverty that emerged from that decade, constitutes a historical anomaly. Such a comparative historical approach may also offer an alternative perspective on the recent ‘Africa rising’ narrative. First, there is nothing phenomenally new about growth in African economies, and second, the current period of growth and poverty reduction may compare unfavourably to that of the 1960s or even the colonial period.Footnote 56
