Exploring Nigerian Auxiliary Generator Use with Variable Length Markov Chains with Covariates
Speaker: Fabrice Rossi, Ceremade
Coauthor: Hugo Le Picard
Abstract
While Nigeria is the first Sub-Sahara economy, its centralised electricity network is very unreliable and is down approximately 45 % of the time. Business and wealthy households use auxiliary power generators to compensate for these blackouts. The aggregated capacity of these auxiliary generators is estimated between 10 and 15 GW, roughly three times more than the actual available capacity of the centralised grid. Generators are a large economic burden and induce health and environmental issues, hence their use should be discouraged. However, usage patterns of these generators remain unstudied which prevents estimating the effects on their use that could be induced by different types of improvements of the centralised power grid. We discuss in this presentation preliminary results obtained on a unique data set of 68 generators installed in Nigeria companies and recorded for several days together with the state of the centralised power grid. Each generator is represented by a binary time series (on and off) recorded with a time resolution of ten minutes. The state of the portion of the power grid to which each company is connected is represented in a similar way. To analyse such binary time series, we rely on Variable Length Markov Chains (VLMC). These are parsimonious Markov chains that can combine short and long memory components. This is particularly adapted to model generator use patterns that are independent from the status of the centralised network, owing to its unreliability. We also consider a recent extension of VLMC to situations where a time series can be influenced by its own past but also by external covariates. This extension is well adapted to capture generator use patterns that depends only on external factors (the state of the centralised network) or on both external factors and the past status of the generator itself. Using a BIC based automatic model selection, we show that all types of dependencies are present in our generator data set. Using simulations generated from the obtained models we assess to what extend they can be used to study the potential effects on usage patterns of improvements of the quality of the centralised grid.
Electrification and Deforestation in Côte d'Ivoire: a spatial econometric analysis
Speaker: Alpha Ly, LEDa
Coauthors: Raja Chakir and Anna Creti
Abstract
This study analyses the links between electrification and deforestation in Côte d'Ivoire. First, we assess the alignment of night lights intensity data with the official electricity coverage statistics, which are available only at the regional level. Then, using panel data on night lights intensity, we investigate the relationship between electrification and deforestation in greater detail, focusing on a fine resolution at departmental level. In this analysis, we take into account both spatial autocorrelation and individual heterogeneity. Our findings reveal that electrification has an overall positive impact on deforestation with a direct positive impact in electrified localities and a negative indirect impact on neighboring ones. This empirical evidence, contrasting with prior findings on developing countries, carries significant implications for environmental policy and sustainable development efforts.