Contents lists available at ScienceDirect Journal of Natural Gas Science and Engineering journal homepage: www.elsevier.com/locate/jngse Stochastic convergence in US disaggregated gas consumption at the sector level Mehdi Abid a,b,∗ , Mohsen Alimi c a College of Humanities and Administrative Science, Jouf University, Saudi Arabia.Lamided, University of Sousse, Sousse, 4023, Tunisia b College of Humanities and Administrative Sciences, Jouf University, Jouf 2005, Saudi Arabia c Higher Institute of Computer Science and Management of Kairouan, Avenue Khemais El Alouini, Kairouan, 3100, Tunisia ARTICLEINFO Keywords: Convergence Natural gas consumption Unit root tests Seasonality United StatesJEL classification: C22 Q40 ABSTRACT In order to assess how seasonality afects disparities in natural gas consumption among sectors, this paper aims to study the pattern of convergence in natural gas consumption in a sample of 11 sectors in the United States between January 1973 and February 2017. In addition to the full sample, the existence of convergence is also examined in fve subsets of sectors: residential, commercial, industrial, transport and electric power. By using various types of unit root tests, empirical results provide signifcant support for the convergence of disaggregated natural gas consumption across sectors in the United States. Another important fnding of this paper is that natural gas consumption, despite being convergent, is very persistent. 1. Introduction Natural gas has been the most important source of fossil energy since the 1970s thanks to its economic and environmental benefts. Yet, the consumption of natural gas in the United States (US) is seasonal. Seasonality refers to periodic 1 fuctuations in consumption of natural gas at sectors levels. But, it's not surprising that seasonal trends speci- fcally impact the disparities in natural gas consumption intensity across sectors. For example, in summer, when demand is lower, natural gas is in- jected into underground storage facilities so that it can be consumed during the winter months when demand reaches its yearly maximum. In 2016, natural gas was the most used energy after oil: technolo- gical advances were constantly improving the efciency of its extrac- tion, transport and storage techniques, as well as the energy efciency of natural gas-powered equipment. Consumption grew more rapidly after World War II thanks to the rise of pipeline networks and storage systems. This fossil energy source has considerably increased over the past 40 years, going from 1241.5 MMm 3 (billion cubic meters) in 1974 to 3.470 MMm 3 in 2015, an increase of 36.68% (compared to 26.4% for oil over the same period) (EIA, 2017). In 2015, natural gas met one quarter of global energy demand. During the same year, ten countries accounted for more than 60% of global demand. The main consumer countries were the United States (22.8%), Canada (2.9%) and the 25 European Union countries with nearly 1003.5 MMm. 3 But, the real question is: Just how strongly does the seasonal efect infuence the stochastic convergence in US dis- aggregated Gas consumption at the sector level? Recently, several studies have often found evidence of stochastic conditional convergence of energy consumption among states in the U.S. Payne et al. (2017). But the literature in this area remains open for studies aiming at testing stochastic conditional convergence at the sector level for diferent countries. Mishra and Smyth (2017) stated that sector level studies are considered necessary because fndings of sto- chastic conditional convergence of energy consumption at national or state level potentially hide signifcant diferences across sectors, such as the seasonal efect. Thus, in order to better understand the impact ofthe seasonal efect on natural gas consumption, we study, within a time https://doi.org/10.1016/j.jngse.2018.10.002 Received 9 June 2018; Received in revised form 4 October 2018; Accepted 5 October 2018 ∗ Corresponding author. College of Humanities and Administrative Science, Jouf University, Saudi Arabia.Lamided, University of Sousse, Sousse, 4023, Tunisia. Tel.: +966 557 763 423. E-mail addresses: abid.mahdi@yahoo.fr (M. Abid), mohsenalimi2000@gmail.com (M. Alimi). 1 For monthly data, the seasonality period is usually 12 because there are 12 months in a year. 2 See Durlauf and Quah (1999) and Islam (2003) for a detailed literature review. 3 Numerous studies circumvent the problem of seasonality by utilizing seasonally adjusted data after eliminating seasonal factors by applying, for example, for the monthly data series the factorization of the operator of the seasonal diferences Δ 12 =1-L 12 . However, as indicated by Depalo (2009), this solution is sub-optimal because, during the adjustment process, a lot part of interesting information on data dynamics which is mainly contained in the peak and valleys phases will be totally lost. Consequently, we represent the seasonal data in their raw format, but we test the efects of seasonality in the data. Journal of Natural Gas Science and Engineering 61 (2019) 357–368 Available online 11 October 2018 1875-5100/ © 2018 Elsevier B.V. All rights reserved. T