Do global CO₂ emissions from fossil-fuel consumption exhibit long memory? a fractional-integration analysis

ABSTRACT

In this article we use an autoregressive fractionally integrated moving average approach to measure the degree of fractional integration of aggregate world CO₂ emissions and its five components – coal, oil, gas, cement, and gas flaring. We find that all variables are stationary and mean reverting, but exhibit long-term memory. Our results suggest that both coal and oil combustion emissions have the weakest degree of long-range dependence, while emissions from gas and gas flaring have the strongest. With evidence of long memory, we conclude that transitory policy shocks are likely to have long-lasting effects, but not permanent effects. Accordingly, permanent effects on CO₂ emissions require a more permanent policy stance. In this context, if one were to rely only on testing for stationarity and non-stationarity, one would likely conclude in favour of non-stationarity, and therefore that even transitory policy shocks have permanent effects. Our fractional-integration analysis highlights that this is not the case.

KEYWORDS:

• CO2 emissions
• long memory
• ARFIMA model

JEL CLASSIFICATION:

• C22
• Q54
• O13

Acknowledgements

We would like to thank Pedro G. Rodrigues for very competent editorial assistance. In addition, the first author would like to acknowledge financial support from FCT–Fundação para a Ciência e a Tecnologia (grant UID/ECO/04007/2013) and FEDER/COMPETE (POCI-01-0145-FEDER-007659).

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ The work of Barros, Gil-Alana, and de Gracia (2016), which came to our attention only after the completion of the first version of our article, has a substantial overlap with this article in terms of both the focus on identifying long-memory patterns and in dealing with global emissions and their major components. Significantly, our empirical results and their policy implications are in sharp contrast with those in that paper.

² See Joint Research Centre of the European Commission (2014) for a detailed comparison among different available measurements of CO₂ emissions.

³ It is important to highlight that the data set used in this article incorporates important updates and corrections on carbon emissions data after 1959. Compared to the earlier versions of this data set typically used in the previous literature, total carbon emissions were corrected downwards after 1980 with more pronounced corrections in the recent decades. From a disaggregated perspective, carbon emissions from solid fossil fuels and, to a lesser extent, liquid fuels and cement were corrected downward while emissions form gas fuels were corrected upwards.

⁴ Detailed test results are available from the authors upon request.

⁵ Results for the entire sample period, 1751–2014, are not substantially different, and are available from the authors upon request.

⁶ Because of the similarities in focus and approach between this article and Barros, Gil-Alana, and de Gracia (2016), we performed a wide variety of experiments attempting to replicate the results that were obtained in that paper. We considered different time horizons, previous versions of our data set comparable to the one used in that paper [see footnote 3], carbon emissions as well as carbon-dioxide emissions, estimates in levels and in logs, different structural breaks, and different statistical packages to generate our results including EViews and Stata. Despite all of our efforts, we were unable to replicate their results. Our only conjecture is that the different results are the product of different econometric procedures, a conjecture we were unable to test since we do not have access to the procedures used in Barros, Gil-Alana, and de Gracia (2016).

Additional information

Funding

The first author would like to acknowledge financial support from FCT–Fundação para a Ciência e a Tecnologia [grant UID/ECO/04007/2013] and FEDER/COMPETE (POCI-01-0145-FEDER-007659).