My colleagues Nori Tarui, Takashi Yamagata and I set out originally to explore the EU emissions trading system but discovered that the data we had hoped to find wasn’t available to us.
Around the same time, Covid-19 spread across the globe. The virus brought lockdowns and a massive reduction in economic activity – with the shock to the global economy predicted by the International Monetary Fund in their World Economic Outlook, published in April 2020, to be both faster and more severe than during either the 2008 financial crisis or the Great Depression.
We began to wonder what effect the pandemic might have on global consumption of fossil fuels and CO2 emissions.
When investigating the relationship between energy consumption and growth, many studies employ a single country analysis, adopting vector autoregressive models and vector error correction models, due to the relatively small number of observations (annual data since 1960).
These models capture the dynamic interrelationship among a set of variables, permitting in the case of the latter the underlying variables to have a long-run common stochastic trend. They are suited for modelling a small number of variables, with the focus on the temporal rather than spatio-temporal dimension, the latter allowing to capture interlinkages across variables of a number of individuals, be it countries, regions etc.
Recently, studies using panel data analysis have emerged in the literature. However, they also use a relatively small number of observations across time. These panel data studies permit country heterogeneity only in a limited way without controlling for global common shocks or spatial dependence, despite both of these featuring prominently in the recent panel data econometrics literature.
Our work adopts a global modelling approach that can address all these issues.
For reliable estimation of the associated large dimensional model, we required a sufficient number of time-series observations, a minimum of at least around 80 observations, across countries that cover a major part of the world economy. A precise number in the latter case is not appropriate as this also depends on the relative sizes of the countries. If the countries in the sample are all of the small open economy type, usually a larger number of these is required for the reliable application of the methodology.
The use of quarterly data across a reasonably large number of countries (appropriately adjusted for temperature and seasonal effects) allowed for this, as well as for incorporating a richer set of dynamics into the modelling process than would have been permitted with annual data.
Which data did we access from the UK Data Service?
How our approach to the data was innovative
As we were able to access quarterly data across a reasonably large number of countries, we were able to produce [adopt] a global modelling framework that captures complex spatial-temporal interdependencies across countries, appropriate for assessing the international propagation of the economic impact due to the coronavirus spread, which was the main focus in our work.
The modelling approach can be summarised in a two-step procedure where in the first step, small‐scale country‐specific models are estimated conditional on the rest of the world. These models feature domestic variables and weighted cross‐section averages of foreign variables. In the second step, the individual country models are stacked and solved simultaneously as one large global model.
This two-stage approach allows to analyse complex spatial-temporal interactions in the global economy and other data networks where both the cross‐section and the time dimensions are large, dealing with the curse of dimensionality in a theoretically coherent and statistically consistent manner.
As part of our analysis and following the IMF pattern, we divided the countries into ‘advanced economies’ and ‘emerging market and developing economies’ (for short, ‘emerging economies’) as per table 1.
Table 1. 32 countries and associated groups within the sample.
|Advanced Economies||Emerging Economies|
What were the key messages from our research?
Our research suggests that fossil fuel consumption and CO2 emissions are likely to return to their pre-crisis levels, and even exceed them, within the two-year horizon (2020-2021) despite large reductions in the first quarter following the coronavirus outbreak (Q1 2020).
Our analysis indicates that more robust growth is anticipated for emerging economies as opposed to for advanced economies.
Recovery to pre-crisis levels of fuel consumption and CO2 emissions is expected even if another wave of pandemic occurs within a year.
The bottom line is that the initial reduction in fuel consumption and CO2 emissions during the early part of the COVID-19 pandemic does not provide countries with a strong reason to delay climate change mitigation efforts.
How our research could be useful for the research community going forward
Assessing the economic impact of the coronavirus pandemic clearly requires a global modelling approach that accounts for cross-country dependencies and dynamic macroeconomic effects; but even more generally nowadays, within a globalised world, economic issues need to be considered from a global perspective.
It is our hope that the global modelling approach adopted in this work, which adequately captures such dependencies and effects, will be further adopted within the energy field to address key economic issues related for example to the Paris Agreement targets, among others.
Our work proves that there is scope for the use of quarterly energy data with such a modelling framework across a reasonably large number of countries, allowing for richer dynamics to be included and more reliable estimation compared to the limited number of observation available with annual data.
Moving forward it would be beneficial to the research community if more quarterly energy related data could become available.
L. Vanessa Smith is Senior Lecturer at the Department of Economics, University of York. Her main research interests lie in panel data econometrics, Global VAR modelling, and estimation of large covariance matrices.