Variable selection in the analysis of energy consumption–growth nexus

Abstract

There is abundant empirical literature that focuses on whether energy consumption is a critical driver of economic growth. The evolution of this literature has largely consisted of attempts to solve the problems and answer the criticisms arising from earlier studies. One of the most common criticisms is that previous work concentrates on the bivariate relationship, energy consumption-economic growth. Many authors try to overcome this critique using control variables.However, the choice of these variables has been ad hoc, made according to the subjective economic rationale of the authors. Our contribution to this literature is to apply a robust probabilistic model to select the explanatory variables from a large set of potential candidates in the case of the US from 1949 to 2010, not only for an aggregate analysis but also for a sector analysis. The results highlight the critical role of public spending and energy intensity in the explanation of growth. Furthermore, since the study reveals different explanatory variables for each sector, it indicates the importance of policy decisions specifically aimed

Energy consumption in the US reconsidered. Evidence across sources and economic sectors

“…This small contribution may explain why there is no causality between renewable energy sources and economic growth. Camarero et al, [15] show a broad overview of the control variables used in the literature with the aim of measuring the relationship between energy consumption and economic growth. These variables include: employment, energy prices, government spending, gross fixed capital formation, real money supply, energy intensity, energy efficiency, business sector productivity and exports.…”

“…However, the inclusion of control variables is not without problems. First, the choice of these variables has been ad hoc, made according to the subjective economic rationale of the authors [15]. Additionally, a recent survey by Narayan and Smyth [18] about this literature warns of the trade-off that necessarily emerges using the bivariate model, which is susceptible to omitted variable bias, and using a multivariate approach has an associated risk of overparameterization of the model, which contributes to estimation error [18].…”

Multi-Output Conditional Inference Trees Applied to the Electricity Market: Variable Importance Analysis

“…Large consumers could benefit by efficiently managing their energy consumption, which could potentially lower their costs. Governments, on the other hand, are interested in a more accurate forecasting process for price and demand in order to implement new environmental regulations and structure changes applied to the energy market, or integrating different electric markets [3], which could eventually have a positive impact on their economic growth [4].…”

Re-assessing causality between energy consumption and economic growth

“…Over recent years, several authors have attempted to overcome this problem by including additional variables in the causal analysis of the relationship between energy consumption and growth. However, as pointed out by [ 19 ], these additional variables have been selected on a rather ad hoc basis and the results on causality may be influenced by variable selection bias. Related to this problem of selection bias and variable omission, [ 20 ] point out that the absence of a prior theoretical model may cause the causality test to deliver mixed results.…”

“…Related to this problem of selection bias and variable omission, [ 20 ] point out that the absence of a prior theoretical model may cause the causality test to deliver mixed results. To address the lack of statistical motivation when choosing the control variables for the causal analysis, [ 19 ] apply a robust Bayesian probabilistic model to select the explanatory variables to be considered in the causal analysis of the relationship between energy consumption and economic growth. This approach allows for the evaluation of the posterior probability of including in the model a control variable selected from a large group of possible candidates; to the best of our knowledge, it is the first time in the literature that a robust variable selection method has been applied for this purpose.…”