New Data Sheds New Light on Global Electrification Patterns
by Michaël Aklin
Increasing access to electricity is key to improving the welfare of millions of households across the world. Fans and A/C provide much needed relief in periods of intense heat. Reading after sunset is decidedly easier and healthier with electricity than with kerosene. Reliable electric power allows to charge phones, listen to the radio, watch TV, and keep food and drugs cold. Yet about one billion individuals lack proper access to electricity.
Implementing effective policies is imperative to reduce this number. But to design good policies, decision-makers must be able to rely on good data. As it turns out, reliable data on electrification are hard to come by. The leading source of data on household electrification rates across the world is undoubtedly the International Energy Agency’s World Energy Outlook, which itself relies on the World Bank’s Global Tracking Framework (GTF). This is where the problems start.
First, the World Bank’s dataset only begins in 1990. This means that we cannot learn from countries that successfully tackled electrification challenges earlier. Valuable insights are therefore lost from developing and emerging countries – countries that would, presumably, have much to offer in terms of experiences. This includes important cases like China and Vietnam.
Second, the World Bank solved the missing observations problem (years for which no data are available) by simulating them. Simulation, when done transparently, can be helpful. But the GTF has two problems. One is that the statistical models used by GTF to generate simulated values are opaque. As a result, analysts and researchers cannot readily reproduce these data and assess their uncertainty. Another problem is that while GTF flags which values are simulated and which ones aren’t in its in its reports, the online databases make no such distinction. Thus, there is no easy way for researchers to exclude simulated observations from their analysis.
As a result, S.P. Harish, Johannes Urpelainen, and I undertook one of our most ambitious projects (graciously funded by IGC): creating a new global dataset of household electrification at the total, urban, and rural levels. This dataset had to obey two principles. First, only reliable sources such as country censuses would be used. Each observation would need to be documented in a transparent manner. Second, the dataset would seek to go back in time as far as possible – including the pre-1990 period.
After several years of work, this dataset is now available online and described in a recent article published in Energy Policy (gated, ungated). We found data for 124 developing and emerging countries, going back to 1960 for several of them (and sometimes even earlier!), for a total of 1,035 observations.
Our new data illustrates some of the GTF’s shortcomings. An analyst relying on World Bank data would substantively underestimate how much electrification rates have improved over the past decades. This person would find that household electrification rates improved at a rate of 1% per year. Using our new data, we find that yearly growth reached 1.3% – a difference of 30%! The difference is even starker when considering rural electrification. We find that it increased at a rate of 1.5%, whereas it is estimated to increase at a rate of only 0.98% with World Bank data. These are not trivial differences. Inaccurate data paint a misleading picture.
Even though our data provide a more optimistic picture of global electrification patterns, we are of course not claiming that the problem is overblown. Our aim is different: it is to highlight that good data are necessary to identify critical opportunities and rank priorities. Kandeh Yumkella, the UN Secretary General’s Special Representative for Sustainable Energy for All, said “What you measure determines what you get. That is why it is critical to get measurement right and to collect the right data.” We agree and this is why we are keen to see researchers and practitioners work toward providing better data for enlightened analysis.
Assistant Professor of Political Science
University of Pittsburgh