Despite recent surges in economic growth, Africa has not yet been able to replicate the rapid and sustained gains in incomes as previously seen in East Asia. In search of an explanation, Rodrik (2013 and 2016) inspired a literature that focused on the role of structural transformation in the growth process—notably the insufficient shift of workers from low-productivity jobs in agriculture into high-productivity work in industry. In contrast to the East Asia experience where workers moved out of low-productivity agriculture into high-productivity jobs in industry. In Africa, labour was shifting not into industry but rather into services. In fact, as a share of GDP in Africa was peaking at a much lower per capita income historically than in other regions—leading to what Rodrik termed “premature deindustrialisation.”
The emerging role of services in productivity growth
Manufacturing was seen to be special because it was subject to rapid technological change, economies of scale, and exporting, and low-income countries with manufacturing sectors on average tended to exhibit more rapid growth in labour productivity irrespective of other conventional determinants of growth. These writings pointed out that Africa’s growth, though sometimes comparable to rapid growth in other regions, was not the consequence of an expanding manufacturing sector, but rather growth within agriculture. Although Africa sustained nearly two decades of positive real economic growth prior to the pandemic, the premature deindustrialisation literature attributed this growth to less sustainable factors, such as positive weather and demand shocks in the agricultural sector.
Manufacturing has been a crucial propellant of economic growth because it exhibits “unconditional convergence”—that is, the tendency for economies with initially lower levels of sectoral productivity to experience faster growth toward the global frontier, independent of conventional determinants of growth such as institutions, socioeconomic indicators, and governance. An average country with a booming manufacturing sector could experience high growth rates, even in the absence of exceptional institutions, demographics, and geography. Unlike other sectors, manufacturing allows poorer countries to channel capital at the margin towards investment in manufacturing to drive productivity growth. Since agriculture and services were seen not to provide drivers of high productivity, a pattern of agriculture-to-services structural transformation would likely foredoom Africa to slow growth.
More recent studies, however, have challenged the idea that services are unlikely to drive productivity growth. Nayyar et al. (2021) undertook arguably the most ambitious and comprehensive study that shed light on the power of services to drive productivity. Others such as Kinfemichael and Morshed (2019), and Enache et al (2016) similarly found that Africa’s labour reallocation from agriculture into services was not merely a shift towards subsistence livelihoods with a new face but symptoms of a more powerful transformation taking place: highly productive services sectors starting to act a whole lot like manufacturing.
In Africa and elsewhere, structural transformation began changing in the 1990s and accelerated into the 2000s. This new form of structural transformation has favoured formal-sector firms in tradable sectors that are at once highly productive and job-generating. Much of the early premature deindustrialisation literature relied on panel datasets that did not capture this new development. However, recent research like Diao et al. (2021) and Rodrik (2022) have advanced models that broaden the definition of growth-driving sectors to include unspecified activities beyond manufacturing, which they refer to as “modern sectors”.
Unconditional convergence in new activities?
So, what are these modern sectors? In recent years, new data sets have become available for a large sample of countries that allows us to assess unconditional convergence across a broad range of activities beyond formal manufacturing. Using new data, Figure 1 demonstrates how productivity growth is correlated with levels of productivity by sector for more than 57 economies in the world, representing all world regions and income groups. The horizontal axis represents the value added per worker in the base year while the vertical axis shows the growth rate in labour productivity. The steeper the trend line in a panel, then the more rapid the average increase in productivity for the initially low-productivity countries relative to the initially high-productivity countries, indicating a stronger unconditional convergence in the given sector. Strong unconditional convergence means that economies with low value added per worker are likely to exhibit higher productivity growth in the coming decades.
Unconditional convergence is captured by the parameter “β” which is the slope of the regression line between the axes controlling for decade fixed effects. This measures how countries with lower productivity catch up to those with higher productivity over time and "β" is a parameter that shows the speed of this catch-up, considering the time period of observation. In sharp contrast to agriculture, the figure clearly shows that unconditional convergence is present in a diverse array of sectors beyond manufacturing, including construction, and transport services. Unconditional convergence is strongest in financial services and business services, which are the cornerstones of what Nayyar et al (2021) call “global innovator services”. In fact, the convergence in global innovator services is stronger than in aggregate manufacturing (like all of the sectors in the data, aggregate manufacturing includes both formal and informal activities). The strong convergence of services sectors more broadly suggests that these other modern sectors offer potential pathways to growth.
Figure 1 includes two trend lines (and two βs) that show the relationship between the axes including and excluding Africa from the global data. This exercise allows us to understand convergence dynamics in Africa. While we find that unconditional convergence is generally stronger when Africa is excluded, there are generally not significant differences between the slopes of the two lines and there are many sectors for which β is large and negative even when Africa is included.
Figure 1: Unconditional convergence in labour productivity by sector
Notes: The vertical axis shows the compound annual growth in real value added per worker for one-digit ISIC sectors by country using the three periods: 1990-2000, 2000-2010, and 2010-2018. The horizontal axis shows baseline labour productivity values in the given base year (1990, 2000, or 2010). The β coefficient is the slope of the regression of the vertical axis on the x-axis controlling for decade fixed effects. The sample includes 57 economies from all world regions and all income groups. Source: Heitzig, Newfarmer, and Page. (Forthcoming 2023). "From De-Industrialization to Job Creation: New Perspectives on African Growth." In: "New Pathways to Job Creation and Transformation in Africa: The Promise of Industries without Smokestacks." The Brookings Africa Growth Initiative.
What sets Africa’s structural transformation apart?
To compare Africa’s pattern of structural transformation with that of other regions, we decompose Africa's productivity growth into two parts: within-sector growth, or the fruits of using inputs more productively in a given broad sector, and between-sector growth, or the fruits of labour reallocating from less productive sectors to more productive ones. Our approach follows the methodology developed by Fabricant (1942) and subsequently popularised by de Vries et al. (2015), Diao et al. (2019), Nayyar et al. (2021), and others. The analysis uses the Economic Transformation Database which contains data for employment and value added for 12 sectors for 1990-2018. The database covers Rwanda, Uganda, Zambia, Ethiopia, Kenya, Ghana, Senegal, South Africa, Botswana, Burkina Faso, Cameroon, Lesotho, Malawi, Mauritius, Mozambique, Namibia, Nigeria, and Tanzania.
Between 1990-2018, sub-Saharan Africa’s productivity growth (as measured in a sample of 18 countries with available data) though lower than global leaders in East Asia and South Asia, was higher than in comparable countries in the Middle East and North Africa, and Latin America (Figure 2). Sub-Saharan Africa’s pattern of structural transformation is unique among the other regions due to its distinctive near even split between within-sector productivity growth and between-sector productivity growth. Agriculture has been the key driver of within-sector productivity growth, but services, manufacturing, and other industrial sectors also contributed. Most remarkably however, these data show that services have played a more significant role than manufacturing in both within-sector and between-sector growth in every world region, suggesting that the services-led growth is not unique to sub-Saharan Africa.
Figure 2: Decomposition of labour productivity by region and sectoral group, 1990 - 2018
Notes: Sectors from the Economic Transformation Database were divided as follows: Construction, mining, and utilities were classified under “Industry”, and trade services, transport services, business services, financial services, real estate, government services, and other services were classified under “Services”. Agriculture and manufacturing are defined by their ISIC Rev. 4 codes “A” and “C” respectively. Source: Economic Transformation Database, Groningen Growth and Development Centre and UNU-WIDER. Design inspired by Nayyar et al. (2021).
Pre-mature deindustrialisation, services, and unconditional convergence: Eight countries
While manufacturing has not played the same pivotal role in Africa that it did in East Asia, the sector has expanded in absolute size in virtually all countries even though its share in GDP has not increased on average. However, the services sector expanded more rapidly. A review of eight sub-Saharan African countries illustrates trends in manufacturing and services – and illuminates the interactions between the two sectors.
Part of the reason manufacturing has not been for Africa what it was for early industrialisers is because of technological dualism, when the capital-intensive technology inputs needed to compete in global markets tend to produce large productive firms with relatively little employment growth. Recent research from Tanzania and Ethiopia suggests this is precisely what is occurring in Africa. Two main types of manufacturing firms are in operation in these two countries: large, capital-intensive firms with exemplary productivity growth but relatively little job creation, and small labour-intensive firms accounting for large employment growth but weak productivity growth. These two types of firms coalesce to form an expanding though often informal manufacturing workforce with limited productivity growth except for in a few high-performing firms. For many African economies, this trend has coincided with manufacturing’s share in output falling even while its employment share has risen (Figure 3).
Figure 3: Evolution of the manufacturing sector, 1990 (smallest dot) to 2018 (largest dot)
Notes: In these eight sub-Saharan African countries for which data is available, the share of manufacturing in output has fallen even though the share of people employed in the sector has risen. Source: Groningen Growth and Development Centre. 2022. Economic Transformation Database.
Meanwhile, services in the same countries have expanded with considerable productivity growth (Figure 4). The sectors at the heart of the convergence in services are high-skilled tradeable sectors like IT, financial services, and business services. The services sector has married rapid job growth with productivity growth. This is evidenced by its rising share in employment to a greater degree, and output. All eight of these countries are to the left of the 45-degree line, suggesting that the services sector is more productive (or substantially more productive in some cases) than the economy-wide average.
Figure 4: Evolution of the services sector, 1990 (smallest dot) to 2018 (largest dot)
Notes: The services sector includes sector normally classified under services (trade services, business services, financial services, real estate, and transport service) as well as construction. Source: Groningen Growth and Development Centre. 2022. Economic Transformation Database.
As economies grow into the potential offered by services, this trend does not preclude growth in manufacturing. To the contrary, the manufacturing sector remains a source of productive jobs in Africa. However, as economies grow and become more complex, so too do the linkages between services and other sectors, becoming more intertwined.
Implications for policy and further research
As African policymakers design their policies to promote economic growth and transformation, they need to be mindful that focusing solely on industrial policy promoting manufacturing would miss the opportunity to secure productivity gains in other sectors of which services are perhaps the most important. These include tourism, business services, financial services, telecommunications, and e-commerce. In addition, looking for ways to expand commercial agriculture into high value-added crops, particularly for export, offers obvious growth opportunities in land rich Africa. Finally, even sectors once thought relatively impervious to technological change, such as construction and transport, can play an important role in absorbing labour and driving productivity growth.
These patterns also raise provocative questions for research: what country-specific activities within agriculture and within services hold the greatest promise for rapid productivity growth for a country, how do they fit into growth patterns within African countries, and what policies can be adopted to promote them? Some studies have used qualitative methods to begin exploring these questions for example, by using differences between formal firms and informal firms in a given sector or between export activities and domestic production as proxies for converging activities within highly heterogeneous sectors such as agriculture, retail services and construction. Perhaps most important, is the need to identify the policy amenable “conditional” factors such as governance, investment climate, and social inclusion measures, which, when combined with high productivity job creation in the converging sectors, produce sustained increases in incomes for African countries. East Asia’s structural transformation was driven in large measure by very high domestic savings and investment rates, a challenge still confronting Africa.
Technical details for productivity decomposition
For the productivity composition, we use the Groningen Growth and Development Centre and UNU-WIDER’s Economic Transformation Database (ETD), a 12-sector database that covers more than fifty economies across the world, including 18 from Africa. The expanded coverage provides us with a more accurate picture of employment and the dynamics of structural transformation across the region. We apply the canonical productivity decomposition developed by Fabricant (1942) – and subsequently applied by Nayyar, et al (2021), de Vries et al. (2015), Diao et al. (2019), Dieppe and Matsuoka (2021), and others – to the ETD to understand how the drivers of productivity growth in sub-Saharan Africa compare to those in economies from other regions. Specifically, we define the change in productivity
is the employment share of sectoral group i (i.e. services, manufacturing, other industry, agriculture) at time t, m is the number of sectoral groupings, and a base period measuring k years before t. The term
be thought of as the share of productivity change derived from within-sector changes, while
is the share derived from between-sector reallocation of labour. Because structural transformation in sub-Saharan Africa, like predecessor economies, has begun with a shrinking employment share in agriculture, it is helpful to reformulate the rightmost term in terms of the gains with respect to agriculture, making use of the fact that
To find average annual productivity gains, divide both sides of the equation by
This step will allow us to measure what share of the productivity growth over the period is due to within-sector growth in services, industry, manufacturing, and agriculture, and which is due to labour reallocation from agriculture to services, industry, and manufacturing.
Technical details for measuring labour productivity convergence
The analysis on convergence that appears in the text utilises a sample that combines the ETD with sectoral value added and employment data from the OECD's Structural Analysis Database. The resulting sample used for analysis is comprised of 80 countries that represent all world regions and all income groups. There are many ways to measure unconditional convergence in the literature. This post uses the following regression to calculate beta convergence (β):
is the compound annual growth in real value added per worker
for sector i in country c over the previous decade and the estimated value of β1 indicates the speed of convergence by sector.
Dic is a vector of binary controls for each decade in the dataset. The predicted values of
represent the orthogonal component of growth. Note that because the ETD has data from 1990-2018, the final decade included in the analysis is only eight years: 2010 to 2018. The first two decades used in the analysis are from 1990-2000 and 2000-2010.
 This annex is a preprint from an annex that will appear in a book expected to be published later in this year with the following attribution: “Heitzig, Newfarmer, and Page. (Forthcoming 2023). "From De-Industrialization to Job Creation: New Perspectives on African Growth." In: "New Pathways to Job Creation and Transformation in Africa: The Promise of Industries without Smokestacks." The Brookings Africa Growth Initiative.”