The Influence of Built Environment and Socio-Economic Factors on Commuting Energy Demand: A Path Analysis-Based Approach

Energy is one of the main concerns in this century because of its scarcity and also due to its greenhouse gas (GHG) emissions. Cities are responsible for more than three-quarters of global pollution (Rogers, Gumuchdjian 2008). As a direct consequence of the increasing use of fossil fuels, climate change and its impact have already been observed since the past century with global warming, the melting of the ice pack and the disappearance of several plant and animal species. And if no action is taken, the impact of climate change will continue to be a disaster for the Earth and for humanity. According to the IPCC (2007), even if measures against climate change are taken into consideration by countries, there will be an increase of 0.1°C for each decade, which means that we expect a minimum temperature increase of around 1° by 2100. Sustainable development is introduced in international law to be the purpose around which all countries across the world have to act to reduce the impact of climate change (Boutaud 2005). The main orientation of sustainable development is to assess the countries’ development based on indicators (Idem). This approach helps the decision makers to take necessary actions to reorient the development of their regions so as to be more sustainable. Several summits were organised to gather countries around a set of actions aimed at reducing fossil energy consumption. Algeria, as a signatory of the Kyoto protocol, has launched several actions to reduce its consumption of fossil energy, such as inventories of energy consumption (MEAT, IPCC 2001, MATEC 2010) and the enactment of law 99-09. These set out several actions, among which we found PNME, the national plan of energy mastering, which has as its main goal reducing energy consumption in the four main sectors consuming fossil energy: residential buildings, transport, industry and the agricultural sector (APRUE 2007, 2014, 2015, 2017). For the residential sector, several actions were proposed like constructing 600 housing with high energy performance (HPE) in different climate zones of the country (Boukarta 2019).

In Algeria, the transport sector alone could account for >30% of the final energy consumption, which means that this sector represents a real potential for reducing fossil energy consumption and the resulting GHGs (APRUE 2014). To address this objective, it is necessary to identify the driving factors behind the need for motorised travel. This paper seeks to explain energy consumption resulting from commuting in relation to the built environment (BE) and the socio-economic (SE) characteristics of households. In developing countries, studies aimed at identifying effective factors that could explain the energy consumption caused by commuting are rare; in Algeria, this study is the first. Commuting is the only type considered for this study, as it is characterised by its regularity and large share of total transport volume. According to our literature review, the other modes of travel such as shopping or leisure require different methods for their characterisation (Boukarta 2019). This paper also proposes a comparison of the impact of the driving factors identified between developing countries and the developed ones.

In Algeria, as in many countries around the world, new urban areas are emerging every year with less built density and less functional diversity. It is therefore to be expected that the use of private cars for commuting and energy consumption will increase in the future. An assessment of the energy consumption generated by commuting would help policymakers and planners to design more energy-efficient neighbourhoods. This paper proposes a method to quantify energy consumption in urban areas and investigates the main factors explaining energy consumption due to commuting.

The second section presents a review of the scientific literature with the objective of identifying the research gap and the main factors that have an impact on commuting. In the third section, the developed method is presented and in the following section, an application of the method in the urban area of the city of Djelfa is demonstrated and discussed in the light of other studies.

The literature review aims at identifying the conceptual model, the scale of the study, the type of transport, the data sources and sampling, the method of analysis adopted in several articles and the results of each study.

The literature review is based on original papers indexed in the Scopus and Web of Science databases. A total of 27 articles are reviewed and summarised in Table A1 in Appendix. The literature review allowed us to identify the following findings:

In contrast to developing countries where transport studies are rare, Western countries, led by the United States and the Nordic countries are the forerunners in the field of transport studies.

It appears that both BE and SE factors are important for explaining modal share, but some researchers have emphasised the importance of one over the other. In the United Kingdom, Stead (2001) emphasises the importance of SE factors in explaining car modal share over BE factors. In the United States, Ewing and Cervero (2002) found a greater impact of BE factors in explaining car modal share.

As observed in this literature review, few studies attempt to answer the question of how the results from developed to developing countries can be transferred. This paper aims to shed light on this aspect by identifying the impact of household SE parameters and those of the BE in explaining commuting and related energy use. In addition, this paper seeks to test the hypothesis that household SE characteristics play a structuring role in shaping commuting energy consumption in the same way as factors in the BE. A comparison between the impact of SE factors and the BE would help explain the importance of both in explaining energy consumption.

The path analysis-based approach allowed us to identify 15 variables explaining the energy consumption generated by commuting in the urban area of the municipality of Djelfa. Ten driving factors have an increasing effect on energy consumption and five variables tend to reduce it (Fig. 7).

Fig. 7

We found in the first position the factors increasing the commuting energy consumption. The home-to-work distance has a direct effect of 0.66 and an indirect effect mediated through the round-trip frequency −0.076 (−0.28 × 0.27). The total effect is 0.584 per standard deviation (SD), that is, for each additional kilometre, energy consumption increases at 52.64 kWh annually. Almost the same results have been obtained by other authors. In Belgium, Van Acker and Witlox (2010) have found that distances travelled by car increase by 2.6% for every 15 minutes more in the time spent travelling to work. In the same country, Marique (2013) has found that job diversity within a 10-km radius is associated with a −0.421 decrease in energy consumption per person per trip (kWh per person). Calabrese et al. (2012) have observed in Massachusetts that car travel distances to work decreased by −0.3917 for every additional 10 km of distance to work.

The number of cars comes second, with a direct effect of 0.45 and an indirect effect via home-to-work distance of −0.079 (−0.12 × 0.66). The total effect is 0.379 per SD (177.31 kWh), that is, for each additional car per household, energy consumption increases by 116.87 kWh × person⁻¹ × year⁻¹. A similar effect was identified by Cervero and Radisch (1995), Kitamura et al. (1997) and Pan et al. (2009), who have found that the number of cars per adult reduces the modal share of non-motorised travel by −0.7798, −0.387 and −3.652 times, respectively.

The third most important effect is the round-trip frequency, with a direct effect of 0.27 per SD, that is, for each additional round-trip frequency, energy consumption increases by 93.87 kWh × person⁻¹ × year⁻¹. The model also shows that the frequency of trips is a mediator of the effect of the type of housing (individual or collective) and the frequency of buses. This means that the frequency of trips increases in single-family dwellings where the bus frequency is lower. The number of bus rotations has an indirect effect mediated by commuting distance of 0.284 (0.43 × 0.66) and commuting frequency of −0.0325 (0.43 × −0.28 × 0.27). The total effect is 0.249, that is, for each bus rotation, energy consumption increases by 79.40 kWh × person⁻¹ × year⁻¹. Indeed, increasing the number of bus stops would increase the travel time. Pan et al. (2009) have found that the choice of car as a travel mode increases by 1.03 times for each additional minute and Chen et al. (2007) have observed that the number of bus stops per trip increases car use for travel by 7.1%.

Distance to the centre comes fifth with an indirect effect mediated by built-up density 0.05 (−0.33 × −0.15), by home-to-work distance 0.218 (0.33 × 0.66) and by round-trip frequency −0.025 (0.33 × −0.28 × 0.27). The total effect is 0.241 per SD, that is, for each additional kilometre of distance to the centre, energy consumption will increase by 61.35 kWh × person⁻¹ × year⁻¹. Distance to the centre is positively correlated with distance travelled by the private car. Holden and Norland (2005) have obtained almost the same result in Oslo. For each additional kilometre of distance to the centre, the energy consumption generated by motorised travel increases by 108 kWh × person⁻¹ × year⁻¹. Similarly, Naess (2010) has shown a positive association between distance to the centre and energy consumption generated by car travel.

Profession comes sixth with a direct effect of 0.165 per SD. In other words, liberals tend to consume 55.726 kWh × person⁻¹ × year⁻¹ more than public-sector employees. Occupation could be considered as a proxy for income; in Algeria liberal employees have a higher income than public-sector employees. Income is the most common determinant in mobility studies. Several studies have found a positive correlation between income and VKT; a higher income leads to more motorised mobility. In our study, the income effect acts indirectly through the number of cars by 0.193 (0.43 × 0.45) and through the home to work distance by −0.034 (0.43 × −0.12 × 0.66). The total effect is 0.161 per SD, that is, for every additional 15,000 Da (≈103€), energy consumption increases by 30 kWh × person⁻¹ × year⁻¹. Van Acker and Witlox (2010) have found that income is positively associated with car travel in Belgium. Households with an income above €3,100 travel 13% more by car than households with a lower income. Khan et al. (2014) have observed that income could explain up to 6.21% of car trips in Seattle, while Dieleman et al. (2002) have found a smaller impact in the Netherlands (2.9%).

Holden and Norland (2005) in Oslo, Norway, have noted an increase in energy consumption due to motorised mobility of 2 kWh/persons for every 1,000 Norwegian Kroner or 100€ × person⁻¹. And Næss (2010) found that income is positively correlated with VKTs and explains an additional 1.41 km (Beta = 0.07) for every 1,000 Yuan (127€). While for the same income, and for trips to the centre, Naess (2014) has observed a weak positive correlation between VKTs and income in Copenhagen, 0.0045 km more for every additional 1,000 Danish kroner (134€). The author explains this weak correlation by the location of housing. Households with higher incomes tend to live in the centre, while households with lower incomes tend to live on the outskirts near public transport.

In the eighth place comes the number of floors per urban block with an indirect effect through the number of cars per household of 0.117 (0.26 × 0.45) and through the home-to-work distance of −0.0206 (0.26 × −0.12 × 0.66). The total effect is 0.097, that is, for each additional storey per urban block, energy consumption increases by 16 kWh × person⁻¹ × year⁻¹. The number of floors per urban block is a proxy for the distance-to-centre factor. Indeed, the highest buildings are located in the periphery of the municipality of Djelfa.

In the ninth place is the occupancy rate per dwelling, which has an indirect effect mediated by home-to-work distance of 0.0858 (0.13 × −0.66) and by round-trip frequency of −0.0098 (0.13 × −0.28 × 0.27). The total effect is 0.076, that is, for each additional household member, energy consumption for mobility increases by 7.187 kWh × person⁻¹ × year⁻¹. The occupancy rate per dwelling or the size of a household in most studies is negatively correlated with motorised mobility. Ding et al. (2017) have found that household size reduces car travel by up to −0.56 in Baltimore and Cervero and Radisch (1995) have shown positive associations between household size and increasing use of public transport modes, cycling and walking rather than driving in the United States.

Similarly, Ma et al. (2014) in China have observed that household size has a negative impact on chain travel. While in Seattle, Khan et al. (2014) have found that household size has a positive impact (by 25%) on distances travelled by car. The researchers explain this positive correlation by the presence of young children dependent on their parents’ mobility. Other researchers have found that household size has no impact on the volume of car travel (Chen et al. 2007; Manaugh et al. 2009). Chen et al. (2007) have observed that the effect of household size is more pronounced for households with children under the age of six.

In the last position and with the smallest effect on energy consumption generated by commuting is the type of housing which has an indirect effect mediated: by built density of −0.099 (0.66 × −0.15), by number of cars of 0.135 (0.3 × 0.45), by home-to-work distance of −0.024 (0.3 × −0.12 × 0.66), by round-trip frequency of −0.0378 (−0.14 × 0.27) and finally by profession of 0.391 (0.23 × 0.17). The total effect amounts to 0.016 per SD only, that is, an occupant residing in an individual housing tends to consume 5.67 kWh × person⁻¹ × year⁻¹ more than a person living in a collective housing. The type of housing can also be considered as a proxy of the income, as living in a single-family house is often possible for households with a higher income.

In the cluster of driving factors with a reducing impact on energy consumption, we find built density at the top of the ranking with a direct effect of −0.15 per SD, that is, for each additional +1 value of built density, energy consumption decreases by −35.46 kWh × person⁻¹ × year⁻¹. The same conclusions have been found by Brownstone and Golob (2008) in California. A reduction in built density for the same household composition could increase vehicle miles travelled by up to 1,171 miles. Researchers explain these results by the fact that car ownership is less frequent at higher built density. Van Acker and Witlox (2010) have also found that the car modal share tends to decrease in areas with high built density. They also found that car ownership is also negatively correlated with car modal share. In contrast, in Norway, Holden and Norland (2005) have found that built density is not significantly correlated with VKTs.

The age of the respondent comes second, with a direct effect of −0.12 per SD, that is, for every 10 years increase in the age of the respondent, the energy consumption for commuting decreases by −19.66 kWh × person⁻¹ × year⁻¹. Our results are in line with several studies. In much of the scientific literature, age is considered to be a factor that reduces the energy generated by car travel (Handy et al. 2005; Pan et al. 2009; Van Acker, Witlox 2010). Similarly, in Montreal, age was found to be positively correlated with proximity to work (Manaugh et al. 2009). While in Hangzhou, age only explains −0.013 km × day⁻¹ of VKT (Naess 2014) and +0.088 km × day⁻¹ in Copenhagen (Idem). The author explains this opposition of the age effect by the lifestyle effect. Educated young people in Copenhagen tend to live in the centre, while the opposite is true in Hangzhou. In New York, Chen et al. (2007) have also found a positive correlation with VKTs (0.01). The authors explain this correlation by the fact that older people tend to use cars more than younger people, as car ownership is higher among older people. Holden and Norland (2005) have found that age in Oslo has no statistically significant effect on VKTs.

Bus frequency in the residential environment comes third, reducing energy consumption by an effect of −0.054, mediated by round-trip frequency −0.027 (−0.10 × 0.27) and by profession −0.027 (−0.16 × 0.17). In other words, for each additional bus frequency in the residential area, energy consumption for mobility decreases by −6.13 kWh × person⁻¹ × year⁻¹. The distance to the national road comes second to last with a small effect of −0.037 mediated by built density. For each additional kilometre of distance from the station road, energy consumption decreases by −9.41 kWh × person⁻¹ × year⁻¹.

Indeed, the national road runs along the city from north to south and divides it into two parts: east and west. A large volume of transport passes along the national road, which is characterised by the large width of its carriageway and the presence of several bus stops, which gives it a structuring role in the city of Djelfa. The distance to the national road in the city of Djelfa is a proxy for the distance to public transport. Indeed, the greater the distance to public transport, the greater the use of the car for travel. Van Acker and Witlox (2010) and Calabrese et al. (2012), Khan et al. (2014) and Ding et al. (2017) state that doubling the distance to public transport increases the distance travelled by car by 3.4%, 3.04%, 13.2% and 4.7%, respectively. Ding et al. (2017) have also found that public transport use drops by −11.4% if the distance to public transport is doubled.

Education level comes last, with a small effect of −0.016 per SD, mediated by built density. In other words, for each additional level of education, energy consumption decreases by −2.67 kWh × person⁻¹ × year⁻¹. The impact of education level on VKTs is variable in the scientific literature, ranging from a positive, negative and non-significant correlation. Feng et al. (2013) in China and the Netherlands, Marique (2013) in Belgium have found that each additional SD of the level of education generates 23.4%, 18.1% and 16.8% more distance travelled by car respectively. Calabrese et al. (2012) have found a negative correlation between the education level and car travel distances (Beta = −0.0817). In the Netherlands, Dieleman et al. (2002) have observed that people with a higher level of education tend to prefer car travel to public transport by 4.3% and that the level of education explains only 0.2% of the distances travelled by car. Other research has shown that education level has no statistically significant effect on the energy generated by car travel in Oslo (Holden, Norland 2005) and China (Naess 2010). In general, the impact of education level has a small or insignificant impact in explaining VKTs.

To draw a general conclusion, we have observed that SE factors explain 281 kWh × person⁻¹ × year⁻¹, while BE factors explain 164 kWh × person⁻¹ × year⁻¹. The BE has an impact that appears to be lower than SE factors in explaining commuting energy. This finding is similar to that of Stead (2001), who has found that variation in travel patterns is often due more to SE reasons than to characteristics of the BE.

The environmental impact caused by fossil fuel consumption and the resulting GHGs is increasingly being taken into account by developed countries, and recently by developing ones. The comparison of our findings with those obtained by the authors who considered other cities in developed countries showed almost similar effects. This would lead us to the conclusion that the current knowledge developed through studies in developed countries is potentially transferable to developing ones, subject to seeing what further research in developing countries would yield. The impact of driving factors on the energy consumption of commuting allows us to draw some interesting conclusions. Increased built density, proximity to workplaces, more bus stops and functional diversity in new urban extensions would reduce commuting energy consumption. Similarly, car use for home-to-work trips, round-trip frequency and car ownership would decrease with a reduction in home-to-work distance. Our results suggest a structuring role for the SE characteristics of households in explaining the energy consumption of commuting. The SE characteristics of households are directly related to the quality of the BE, and there appear to be disparities in the distribution of working area density. The new urban areas are less dense and almost totally residential. This implies a greater distance to reach the working areas which are concentrated in the centre. Urban development policy should take into account functional diversity when designing new urban extensions to reduce the modal share of the car, GHG emissions and energy consumption.

Other mobility modes such as shopping, leisure and home-to-school mobility could be considered for future research. Our results could be used in a parameterised tool that could help policymakers and urban planners to design less energy consuming neighbourhoods and could also raise awareness among commuters to be more environmentally friendly.