Identification of how economic development affects energy use through a natural experiment

ABSTRACT

This paper empirically investigates how economic development (GDP) affects energy use. It deviates from previous studies in two important ways: (i) it explicitly allows the effect to be heterogeneous across regions, and (ii) it acknowledges that changes in GDP, i.e. GDP growth, can be endogenous and controls for this by exploiting deaths from natural disasters in 2002 and 2003. This instrument incorporates a natural experiment caused by the unusually severe heat wave that swept past several European countries in the summer of 2003. Based on data from 120 countries, results indicate that the marginal effect of the GDP rate is 66% higher when using 2SLS compared to OLS and that there is substantial heterogeneity across countries. Energy use is negatively correlated with the GDP level suggesting that there may be an Energy Kuznets Curve, but unexplained heterogeneity at high income levels suggests that there is still more to learn about the GDP-Energy relationship.

HIGHLIGHTS

• Investigates empirically how GDP growth affects energy use.

• Utilises the 2003 European heat wave as an instrument for GDP.

• Results show the relationship is heterogeneous and OLS is downward biased.

• Indication of Kuznets effect, i.e. a stronger relationship for less developed countries.

KEYWORDS:

• GDP
• energy use
• natural experiment
• natural disaster

JEL CLASSIFICATIONS:

• O13
• O44
• Q54

Acknowledgements

For comments on earlier drafts, the author is indebted to Raimund Bleischwitz, Matthieu Glachant, Florian Flachenecker and participants of the SINCERE Shanghai seminar in October 2015. All remaining errors are entirely mine.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 These agreements are ambitious in scope and increasing energy efficiency is only one part of what they cover.

2 Al-Mulali et al. (2013) find that the use of renewable energy sources does not in itself weaken the relationship between GDP and energy consumption. The future demand for energy, and how the relationship between GDP and energy will evolve, is uncertain. For example, the relationship will weaken as a result of technological development (e.g. improved energy efficiency), but societies are also subject to a general electrification process and some sectors, such as mining, is likely to increase their use of energy in the future (e.g. Rötzer, N., & Schmidt, M., 2020). Continuing climate change and economic development may stimulate the demand for energy further (van Ruijven, De Cian, and Wing 2019).

3 At its meeting on the 22 May 2013 the European Council identified low security of supply as one of the most important challenges related to the economy-energy nexus.

4 Further empirical studies that claim there is a circular relationship between GDP and energy consumption include Kalimeris, Richardson, and Bithas (2014), Lee (2006), and Mishra, Smyth, and Sharmac (2009).

5 For a review of mismeasurement in macroeconomics, see Orphanides et al. (2000). Events that happen in the middle of the year have a lower risk of being registered in the wrong year. Even more importantly, the consequences of macroeconomic events are often delayed so if the event takes place at the end of the year, it is a higher probability that the consequences are registered in both the current and the following year, making causal analyses more complicated.

6 A source of why environmental preferences vary is cultural differences (Wilhite et al. 1996).

7 Empirical investigations of the consequences of sudden deaths of politicians and head of states suggest that they lead to market unrest, e.g. drop in financial value, reduced sales and more limited access to credit (Faccio and Parsley 2009; Iqbal and Zorn 2008).

8 Examples of such aid programmes are the EU Solidarity Fund (set up in 2002), The U.S. Department of Agriculture Disaster Assistance Programme, the Australian and Victorian Governments’ Recovery Assistance Package after the Victorian Bushfires in 2009.

9 See Panizza and Presbitero (2014) for a more detailed analysis of how OLS is affected by reversed causality, and in particular about the roles of the signs of the two relationships.

10 Oil rent is defined by the World Bank as the difference between the value of crude oil production at world prices and total costs of production.

11 The sources referred to by Henderson (2005) state that “no more than 2%” of household in the European Union had air conditioning installed at the end of the 1990s.

12 The four control countries are Austria, Czech Republic, Poland and Denmark.

13 We take the average of this variable during five years because of several missing values and unexpectedly high within country variation.

14 It has been claimed that higher quality institutions increase investments, which increase growth (Gwartney, Holcombe, and Lawson 2006).

Additional information

Funding

Financial support was received from ANR, [grant number ANR-14- ORAR-0001].