Data and stylized facts

Examining the effect of natural resources on urbanization in developing countries is not straightforward, as data on both are not always readily available. For our analysis, we assembled data for 116 countries from 1960 to 2010.7 We exclude from our study Europe and the Neo-Europe, as they were already highly urbanized in 1960. We focus on countries that belonged in the "developing world" in 1960. Of those developing countries, we exclude those that are smaller in area than 500 square kilometers (e.g. the Seychelles) and countries that do not have sufficient data over the entire period (e.g. North Korea, Central Asian exCommunist countries). Full details of our criteria are available in the Web Data Appendix. The 116 countries we are left with belong to four areas, as defined by the World Bank: Asia (N = 27), Latin America and the Caribbean (LAC, N = 26), the Middle-East and North Africa (MENA, N = 17) and Sub-Saharan Africa (SSA, N = 46).

We use United Nations (2011) to obtain the urbanization rate (%) for the 116 countries for every ten years from 1960 to 2010.8 For the purposes of our analysis, we define natural resource exports to consist of fuel, mineral, cash crop and forestry exports. We use World Bank (2013) and USGS (2013) to estimate the share of fuel and mineral exports in total merchandise exports (%) for each country-year observation. Two issues with the World Bank data set are that there are many missing observations and that the value of fuel and mineral exports is sometimes purposely underestimated in national accounts data, especially in more corrupt countries. We correct the World Bank data set using the very informative annual country reports of USGS (2013). This allows us to estimate the "true" contribution of such exports to these economies. We use FAO (2013) to obtain exports of cash crops and forestry for the same observations.9 Using these data, we define countries with resource exports to GDP of more than 10 % as "resource exporters" (60 countries) and those with less than 10 % as "non-exporters" (56 countries).11

With these data in hand, we can look at several simple correlations. At first glance, resource-exporters and non-exporters do not appear to be significantly different in terms of the income/urbanization relationships. Figure 1 plots the urbanization rate in 2010 against log GDP per capita in 2010. As can be seen, there is a consistent positive relationship across both samples. In all cases, higher urbanization rates are associated with higher income per capita. But despite the apparent consistency in Fig. 1, there are distinct differences in the relationship of industrialization and urbanization between resource-exporters and non-exporters. Figure 2 plots the urbanization rate in 2010 against the share of manufacturing and services in GDP for 2010, a crude proxy of industrialization. For non-resource exporters (the black circles), the relationship of urbanization with manufacturing and services is clearly positive and quite consistent within the sample. Increasing urbanization is associated with increasing industrialization. But for resource-exporters (the gray squares), there is no meaningful relationship of urbanization with manufacturing and services in GDP. In general, these countries have much higher urbanization rates at any given level of manufacturing and services in GDP than non-exporting countries. They have urbanized without necessarily industrializing, and the suggestion of Fig. 2 is that natural resource exports may offer an explanation for this pattern.

To this point, our definition of industrialization includes both manufacturing and all services. Reasonable alternatives would be to include only those services that are potentially tradable or to ignore services completely. It is not immediately obvious how to disaggregate services in a useful way.12 But the general correlations hold if we plot urbanization against manufacturing plus FIRE ("finance, insurance, real estate and business services"), or if we plot urbanization against manufacturing alone.13

These differences can be seen more clearly in Fig. 3, in which we plot the relationship between urbanization and industrialization separately for resource exporters and non-exporters. Here we see that the non-exporters line up neatly with the typical intuition regarding urbanization and industrialization. In comparison, for resource exporters, there is no tendency for urbanization rates to be associated with industrialization. However, Fig. 4 shows that for these same countries, urbanization exhibits a strong positive correlation with their level of natural resource exports.14

It is useful to consider patterns within specific regions of the world, as they display the correlations more starkly. Latin America and the Caribbean (LAC) and the Middle. East and North Africa (MENA) are relatively more urbanized than Africa and Asia, with urbanization rates about 80 and 60 % in 2010 respectively. In both Africa and Asia, the urbanization rate was only 10 % in 1950, but it is now around 40 %, as high as in developed countries after the Industrial Revolution. Asia and the LAC region offer examples of urbanization with industrialization, as can be seen in the left-hand panels of Fig. 5, which display a clearly positive association of urbanization with the share of manufacturing and services in GDP. Exceptions include Mongolia and Venezuela, which heavily depend on resources.

In contrast, Africa and the Middle East offer many examples of urbanization without industrialization. Figure 5 shows in the right-hand panels that urbanization in these regions is positively associated with the importance of natural resource exports in GDP, but there appears to be no relationship of urbanization with the share of manufacturing and services in GDP. Sub-Saharan Africa is a particularly interesting example. Its manufacturing and service sectors are relatively small and unproductive (McMillan and Rodrik 2011; Jedwab and Osei 2013); in 2010, employment shares in industry and services were 5 and 29 % for Africa, but 15 and 35 % for Asia, and Asian labor productivity was 1.9 and 2.3 times higher in industry and services, respectively (World Bank 2013). However, Africa has urbanized to the same level as Asia over the last half-century. The most urbanized countries export natural resources: oil (e.g., Angola, Gabon and Nigeria), diamonds and/or gold (e.g., Botswana, Liberia and South Africa), copper (e.g., Zambia) or cocoa (e.g., Ivory Coast).

An important point to make is that urbanization in resource exporters is not driven by meaningful shifts of labor into urban areas to work in the resource sector. Mineral resources are highly capital-intensive, and their production creates very little direct employment. Angola's urbanization rate was 15 % before oil was discovered in the 1960's, but it was 60 % in 2010. The country now has an urban population estimated at 11 million people. But although oil accounts for over 50 % of GDP, this sector employs only about 10,000 workers. Botswana has a similar urbanization rate to Angola, and while the diamond sector accounts for 36 % of GDP, it only provides employment for 13,000 people. Oil exports account for 50 % of GDP in Saudi Arabia, one of the most urbanized countries in our sample, yet the mining sector only employs 1.5 % of its urban workforce.15 Mining accounts for 1 % of total urban employment on average in our sample of resource exporters. Minerals and other "point source" resources do not directly drive urbanization. Neither is urbanization driven by direct employment in the production of cash crops or forest products. These are generally produced in rural areas and mostly contribute to rural employment.

While it is too crude to simply equate urbanization with non-agricultural work, nonetheless there is a strong correlation between the two. Using the most recent data available between 2000-2010 for 42-83 countries, we find that 91 % of agricultural employment is in rural areas while 67 % of non-agricultural employment is in urban areas. Rural areas have about 30 % of their employment in non-agricultural activity, but in urban areas this is 92 %. Overall, neither mineral nor cash crop employment is capable of explaining the correlation of resource exports and urbanization.16 We thus have prima facie evidence that resource exports drive urbanization, but not through direct employment effects. In the following sections we will first demonstrate empirically that the positive effects of resources on urbanization are in fact a robust feature of the data, and additionally that resource-led urbanization generates different types of cities than more traditional industrial-led urbanization.

Questions:

1. what data do the authors use to answer their research question?

2. How many observations do the authors have?

3. Are these data particularly well-suited to answer the research question?

4. Any data limitations, such as missing variables or missing observations?