Pedro Martins (UNECA) | Structural change: concepts, data and methodologies

Copyright creative commons

15 September 2015.

Structural change is back in fashion. After a promising start in the mid-20th century – owing to the seminal works of Allan Fisher, Colin Clark, Simon Kuznets and Hollis Chenery – the topic was subsequently relegated to obscurity during the years of structural adjustment (a rather different concept!). It remained sidelined in the 2000s, when the attention of the research and policy communities was mainly devoted to the Millennium Development Goals and their focus on social outcomes.

Over the past five years, however, there has been a renewed interest in the study of structural change – partly due to concerns that recent growth patterns in developing countries are neither inclusive nor sustainable. The work of McMillan and Rodrik in 2011 did much to reignite the academic and policy debates. Structural change has also become part of the political lexicon and is increasingly captured in national and regional vision statements – for example, the African Union’s Agenda 2063 – as well as international policy agendas such as the forthcoming Sustainable Development Goals.

I have contributed to the topic with a comprehensive study assessing trends at the sub-regional level and a paper on Ethiopia. In this post, I provide a brief overview of the key concepts, data and methodologies that have been used in recent empirical studies.

Concepts

There is no universally agreed definition of structural change. In fact, many economists also refer to ‘structural transformation’, using the terms interchangeably. Judging by the way economists have tended to utilise the concept in practice, we can categorise existing perspectives into three broad groups: (i) ultra-narrow (production) focus, (ii) narrow (productivity) focus, and (iii) broad (socioeconomic) focus.

The first group assesses structural change merely in terms of shifts in the structure of production. Structural change happens when the economy shifts towards the production of goods and services associated with higher value added, which in turn stimulates economic growth. This usually entails a reduction in the weight of the agricultural sector in total output, and a concomitant increase in the share of industry and/or services. It is implicitly assumed that the market will automatically and efficiently facilitate any required reallocation of resources across sectors (e.g. capital, labour and land).

The second group evaluates structural change in terms of labour shifts from low-productivity sectors to higher-productivity sectors. This relocation of labour raises workers’ productivity, which contributes to accelerated economic growth. While the same sectoral patterns are expected, the explicit focus is on labour productivity rather than production alone. This stems from the observation that changes in the structure of employment often lag behind shifts in production. Many of the contemporary empirical studies fall in this category – such as McMillan, Rodrik and Verduzco-Gallo (2014), Roncolato and Kucera (2014), and de Vries, Timmer and de Vries (2015).

The third group goes beyond changes in the economic structure – such as production and employment – by also measuring changes in other aspects of society. For instance, structural change may entail a demographic transition (through lower fertility rates), changes in labour participation (through changing social preferences), and a spatial reorganisation of the population (through rural-urban migration). Given the interlinked nature of the process of structural change, it can be useful to consider as many dimensions as possible when conducting an empirical assessment. This is the approach followed in Martins (2014) and Martins (2015).

Data

The recent emphasis on structural change has led to a rapidly expanding body of theoretical and empirical work. Datasets with varying degrees of sectoral disaggregation and country coverage have been compiled. However, existing data sources present trade-offs that ought to be considered. I argue that the choice of data will depend on the purpose of the study – namely, that international sources are useful to carry out research at the regional and sub-regional levels, while national sources are better suited for country-level assessments.

International sources have the benefit of providing harmonised (and thus internationally consistent) data for a large number of countries. For instance, the International Labour Organization (ILO) ensures that the employment data it publishes are consistent with its definitions of employment and working-age population. The United Nations Department of Economic and Social Affairs (DESA) ensures that the output data reported by member countries are published in accordance with the System of National Accounts – an internationally agreed standard for the compilation of economic activity measures. These secondary sources provide output and employment data by sector with a high degree of comparability across countries. Moreover, their extensive country coverage enables highly representative assessments of structural change at the regional and sub-regional levels.

However, in-depth country-level assessments should be based on the raw data produced by national sources. This allows a greater focus on internal consistency and more flexibility when conducting the assessment. For instance, international sources often apply modelling procedures to fill data gaps in order to facilitate country comparisons – e.g. ILO’s Global Employment Trends (GET) database and the Groningen Growth and Development Centre (GGDC) database. However, these gains in comparability might come at the cost of distortions introduced by the modelling (or even harmonisation) procedures. This may not matter much when assessing aggregate trends – as biases are likely to partially cancel each other out – but may affect conclusions at the country level. In addition, international sources often rely on a subset of available data sources – e.g. labour force surveys and/or population censuses – which could be complemented by other sources. These factors may account for some of the discrepancies in the results for Ethiopia (see table below). Greater scrutiny of (all) available data sources – including a deeper assessment of data quality – and the ability to tailor the analysis to a country’s policy needs, in terms of both sectoral emphasis and time horizons, are paramount in producing more precise, relevant and up-to-date estimates on the pace of structural change.

Table: Comparison of results for Ethiopia

Study Period Overall growth Compound annual growth rate (%)
Output

per worker

growth (%)

Contribution from (%): Output

per worker

growth

Contribution from:
Within

sectors

Between

sectors

Within

sectors

Between

sectors

Martins (2014) * 1999-2005 17 87 14 2.7 2.3 0.4
2005-2013 65 76 24 6.5 4.9 1.6
de Vries et al. (2015) 2000-2010 55 61 39 4.5 2.7 1.7
McMillan et al. (2014) 1990-2005 32 21 79 1.9 0.4 1.5
McMillan & Harttgen (2014) 2000-2005 11 99 1 2.1 2.1 0.0

* The results have been updated by using the 2013 labour force survey.

There is also the issue of sectoral classification, which relates to the International Standard Industrial Classification of All Economic Activities (ISIC). Many countries are in the process of moving from ISIC revision 3.1 to ISIC revision 4, which creates a break in temporal comparability, since full correspondence is not possible between the two revisions. Only a meticulous country-specific investigation of the raw data can ensure that potential inconsistencies are minimised. This is particularly relevant for employment estimates.

Methodologies

Most studies on structural change are centred on the decomposition of labour productivity growth, which is typically measured by output per worker. This enables an assessment of the extent to which aggregate labour productivity has been driven by labour shifts across sectors vis-à-vis improvements within sectors – the latter being possible though skills upgrading, complementary capital investments, and/or increased organisational efficiency.

An alternative approach is to decompose output per capita growth rather than output per worker growth. This strategy enables an empirical assessment that is compatible with a broader view of structural change. In addition to evaluating the role of within-sector and between-sector productivity improvements, we are also able to assess the contributions of demographic change and the employment rate to economic growth. For instance, lower dependency ratios can generate a sizeable demographic dividend, while changing social preferences can – through economic inactivity – impact on employment rates, which in turn affect economic growth. Hence, this approach captures shifts in the structure of production, the structure of employment, the level of employment, and the size of the working-age population. In Ethiopia, demographic change accounted for about 10% of output per capita growth in 2005-2011, while a declining employment rate had a negative impact on economic performance. However, the latter was mainly due to young people staying longer in education, which is a positive development – especially if young people acquire skills that can boost labour productivity in the near future. In fact, these two dimensions are intrinsically connected. In order to benefit from a sizeable demographic dividend, African countries will have to scale up investments in human capital – as many Asian and Latin American countries have done in the past.

Conclusion

The recent proliferation of studies in this field has contributed to a better understanding of the pace and patterns of structural change in developing countries. Nonetheless, the findings emerging from the literature are sometimes ambiguous, owing to the use of different data sources, country samples, time frames, levels of sectoral aggregation, empirical methodologies, etc. This blog does not assess these discrepancies but provides a few thoughts on how to address some of the tensions arising from the different possible purposes of empirical studies in the structural change tradition.

 

Key references:

de Vries, Timmer and de Vries (2015) ‘Structural Transformation in Africa: Static Gains, Dynamic Losses’, Journal of Development Studies 51(6): 674–688.

Martins (2015) Sub-Regional Perspectives on Structural Change. CREDIT Research Paper 15/03, University of Nottingham.

Martins (2014) Structural Change in Ethiopia: An Employment Perspective. Policy Research Working Paper, WPS 6749. Washington, DC: World Bank Group.

McMillan and Harttgen (2014) What is Driving the African Growth Miracle? National Bureau of Economic Research (NBER) Working Paper No. 20077.

McMillan and Rodrik (2011) Globalization, Structural Change and Productivity Growth. NBER Working Paper No. 17143.

McMillan, Rodrik and Verduzco-Gallo (2014) ‘Globalization, Structural Change, and Productivity Growth, with an Update on Africa’, World Development 63: 11-32.

Roncolato and Kucera (2014) ‘Structural Drivers of Productivity and Employment Growth: A Decomposition Analysis for 81 Countries’, Cambridge Journal of Economics 38(2): 399-424.