Essays on macroeconomics in Africa

This thesis comprises three self-contained essays on topics surrounding economic growth and macroeconomic performance in sub-Saharan Africa (SSA). The first two address aspects of measurement error in macroeconomic data from SSA: the first essay (Chapter 2) explores the potential of nighttime luminosity as a means of improving GDP estimates, the second (Chapter 3) assesses the potential impact of changes in the data on empirical results from time series analysis. The third essay (Chapter 4) investigates exchange rate dynamics in Zambia, and how these pass through to consumer prices. The first essay (Chapter 2) explores heterogeneity in the nexus between nighttime light emissions and economic activity, a relationship that is increasingly exploited with the aim of reducing measurement error in growth estimates of countries where the data is considered weak. I show that the elasticity between nighttime lights and economic activity varies substantially across countries, and that this variation has a systematic component which could have implications for empirical results. Using the elastic net method in order to isolate the relevant factors, I find that 55% of the variation can be explained by observable factors. When tracing economic growth in Africa since 1992 using luminosity and accounting for the explained share of parameter heterogeneity, I find no evidence of an 'African Growth Miracle' as described by, e.g., Young (2012). However, I do find evidence that countries that recently revised their GDP figures (like Nigeria and Ghana) had a tendency to report inflated growth rates for recent years. This is consistent with Jerven (2014)'s hypothesis of purely 'statistical growth'. The second essay (Chapter 3) explores the inconsistencies across different versions and sources of national accounts data (three versions of the Penn World Table and the World Development Indicators), and their impact on macroeconomic inference from time series analysis. I use the statistical framework developed by Juselius et al. (2014) and assess the robustness of their conclusions regarding the long-run impact of foreign aid on economic growth in 36 SSA countries that are generally considered to have low statistical capacity, and where sources on macroeconomic variables sometimes disagree strongly. The results of this exercise are mixed: When I apply the Cointegrated VAR models precisely as developed by JMT to the new datasets, results remain robust for about two thirds of the countries. Once I re-specify the time series models based on the respective data (as different data will sometimes imply different lag lengths, cointegration ranks, etc.), using the same statistical methodology as JMT, this often leads to more substantial changes in the conclusions. The third essay (Chapter 4) investigates the dynamics between the exchange rate and consumer prices in a resource-rich setting in a case study of Zambia. Using a combination of short-run sign- and zero-restrictions, I identify relevant global and domestic shocks in a structural VAR (SVAR). The results suggest that the pass-through of the exchange rate to consumer prices (ERPT) depends greatly on the shock that originally caused the exchange rate to fluctuate. While, for instance, the price of copper is the most important driver of the exchange rate, the fluctuations caused by it tend to affect prices only moderately (an ERPT of ca. 7%). On the other hand, exchange rate fluctuations caused by monetary shocks come with a much higher pass-through of up to 25%