Financial Implications of Car Ownership and Use: a distributional analysis based on observed spatial variance considering income and domestic energy costs

This paper presents a new perspective on assessing the ﬁ nancial impacts of private car usage in England and Wales using novel datasets to explore implications of motoring costs (principally Vehicle Excise Duty and road fuel costs) for households as part of the overall costs of their energy budget. Using data from an enhanced version of the Department for Transport ‘ MOT ’ vehicle test record database, combined with data on domestic gas and electricity consumption from the Department for Business, Energy and Industrial Strategy (formerly the Department of Energy and Climate Change), patterns of car usage and consequent energy consumption are investigated, and the costs of Vehicle Excise Duty and road fuel examined as a proportion of total expenditure on household direct energy consumption. Through the use of these new datasets it is possible to analyse how these vary spatially and in relation to levels of median income. The ﬁ ndings indicate that motoring costs are strongly regressive, with lower income areas, especially in rural locations, spending around twice as much of their income on motoring costs as the highest income areas.


Introduction
With increasing digitisation of vehicle records, new opportunities are being afforded to researchers interested in exploring car usage at the level of individual vehicles. In particular, periodic vehicle safety and emissions inspections are providing a fruitful source of new data. Globally, these tests are becoming increasingly common, taking place in all 27 EU Member States, 32 States in the US, and at least 17 countries in Asia (Cairns et al., 2014;Chatterton et al., 2015). Data from these tests are being put to a range of uses, including understanding spatial patterns and elasticities of car ownership and usage (Moyce and Lloyd, 2013;Reardon et al., 2016;Yeboah et al., 2016), understanding geographical patterns of vehicle emissions (Chatterton et al., 2015), relationships between vehicle usage and urban form (Diao and Ferreira, 2014), implications of future city growth on travel and associated greenhouse gas emissions , issues of environmental and energy justice (Chatterton et al., 2016a) and the potential positive and negative impacts of pay-per-mile vehicle insurance (Ferreira and Minikel, 2013).
In this paper, we explore the financial implications of car use by combining annual data from around 30 million vehicles from the UK vehicle inspection ('MOT' Ministry of Transport) test with accompanying registration data on the location of the registered keeper of the vehicle. We use this to calculate costs of Vehicle Excise Duty (VED) (an annual vehicle tax in the UK) and fuel costs at both a per vehicle and an aggregated area level (around 700 households). We then place these costs in the context of domestic expenditure on electricity and gas use by using energy consumption data from 24.5 million electricity meters and 21 million gas meters (DECC, 2014). While much previous work has looked at motoring costs longitudinally, particularly with respect to price elasticities of road fuel (e.g. Dargay, 2007, Goodwin et al., 2004, in this paper we look instead at how expenditure on motoring varies spatially and in relation to levels of median income. This places the work more in line with previous work on household expenditure (for example, Dresner and Ekins, 2006;Brand and Boardman, 2008;Druckman and Jackson, 2008;Thumin and White, 2008;Gough et al., 2011;Buchs andSchnepf, 2013a, 2013b;Hargreaves et al., 2013). However, this existing body of work generally has no, or very limited, spatial detail as it tends to be based on limited sample survey data, predominantly the UK Living Costs and Food Survey (formerly the Family Expenditure Survey and National Food Survey) which has an annual sample size of 6000 households in the UK per year. We present the work here as an important complementary perspective to these survey based approaches. Whilst our datasets present (near) universal information on vehicle and energy usage, we are cognisant of a number of limitations of this approach. First, due to both the size and security considerations of the datasets used, it is necessary to undertake analysis predominantly on the basis of data that is spatially aggregated (albeit over relatively small and socially homogenous areassee below). Second, the motoring costs that we are able to base our assessment on are those that are dependent specifically on vehicle characteristics and usage, rather than costs such as insurance which are dependent heavily on the characteristics of the driver. Due to this second point, in this paper, our examination of expenditure has focused predominantly on VED and fuel costs. These are important as they are relatively inflexible and are the motoring costs most directly influenced by national taxation policy, therefore reflecting political decisions. Additional work has been carried out that has provided estimations of vehicle depreciation costs as well as the proportion of motoring costs used through travel to work. These have been presented elsewhere (Chatterton et al., 2016b).
Initially, this paper sets out the general costs of motoring from survey based work, before establishing the political history of both VED and fuel duty. This context is important for understanding the longstanding tension between viewing automobility as either a luxury or a necessity, and the impacts this has on what are considered to be appropriate taxation structures. The overall methodology is then described before setting out a number of different analyses. These are: relationships between VED and fuel costs, first at the level of individual vehicles and then as household averages at an areal level (including by level of urbanisation); relationships of VED and fuel costs to income and between road fuel costs and domestic energy costs; and finally looking at the proportion of income spent on these costs. There is then a discussion and conclusion section which explores the implications of the findings within the context of current and future mobility and energy policy.

Costs of car ownership
The costs of running a car are made up of fixed annual costs (VED, MOT test fee, insurance etc.), sporadic costs (repair and maintenance), fuel costs and, greatest of all, depreciation. The overwhelming impact of the balance of these costs is that "annual average cost per mile decreases as the annual mileage increases and is frequently perceived as merely the cost of fuel" (RCEP, 1994: Box 7 C). Fig. 1 shows the average annual household costs of car ownership by income decile calculated from the UK Living Costs and Food Survey (LCFS) (ONS, 2012). These vary in total from £660 for the lowest income decile, to £7649 for the highest. The proportion of this that is spent on fuel varies between 32.3% for the highest decile and 42.6% for the second highest decile (36.6% overall), given that purchase costs are included. The living costs survey accounts for VED (and motoring fines) as a subsection of 'Licences, Fines and Transfers' alongside Stamp Duty for house purchases. Although the overall section is split by income decile, no such split is available for VED and motoring fines separately, so in Fig. 1 these have been allocated proportionally according to the whole section. The overall average VED paid is £156 per household. The LCFS accounts for the cost of a vehicle in terms of purchase price, which is calculated as an average over all the households (although not every household purchases a vehicle each year). Another common way of reflecting this cost is in terms of depreciation (the annual reduction between the purchase price and the resale value). This has been estimated at around 15% per year (CarsDirect, 2013), and was estimated, in 1994, to represent 42% of average annual vehicle costs (RCEP, 1994). This compares with between 21% and 35% (average 29.4%) for purchase costs in the LCFS for 2011, as shown in Fig. 1.
To illustrate the difficulties in calculating the full costs of car ownership, which extend beyond the costs outlined above into a range of non-direct and non-monetary costs, it is worth considering Lynn Sloman's analysis from her book Car Sick: "The typical car owning, Briton today devotes nearly 1,300 hours a year to his or her car. It takes him over 500 hours to earn the money first to buy the car and then to pay for petrol, insurance, repairs and parking. He spends another 400 hours every year sitting in his car while it goes and while it waits in traffic jams. More than 250 hours are devoted to a myriad of small tasks associated with a car: washing it, taking it to the garage for repair, filling it with petrol, looking for the car keys and walking to the car, de-icing the windscreen in winter, and finding a parking space at the end of every trip. Finally, he has to work about 100 hours every year to earn the money to pay the extra building society interest because he has chosen a house with a garage rather than one without. All in all, the typical British car driver in 2005 devoted three and a half of his sixteen waking hours to his car. For this time, he travels a little less than 10,000 miles per year. His average speed is less than 8 miles an hour roughly the same as the speed at which he could travel on a bicycle." (Sloman, 2006, p1-2).
A highly detailed spatial analysis might also consider the impact of local policies on motoring costs, such as residential parking, workplace parking levies, low emissions zones, congestion charging and so forth. However, as already stated, this paper does not attempt to consider the full costs of car ownership and use, but focuses specifically on VED and fuel cost, representing around 40% of total car costs (according to LCFS figures) and constituting the proportion of costs that national level policy has direct control over. We describe these briefly below.

Vehicle Excise Duty
Taxation of motor vehicles was first introduced in the UK in the 19th Century under the Customs and Inland Revenue Act 1888 which extended the definition of 'Carriage' from "any vehicle drawn by a 'horse or mule, or horses or mules', to 'embrace any vehicle drawn or propelled' upon a road or, tramway, or elsewhere than upon a railway, by steam or electricity, or any other mechanical power". Key issues that have surrounded VED from the start have involved issues of fairness and equity as well as questions over the appropriate purpose of the tax. As early as 1909, there were objections to the imposition of the tax. In a House of Commons debate around the introduction of a graduated VED based on horsepower, Mr Joynston William Hicks, then Conservative MP for Manchester North West stated: "I hold that a motor car has now become almost a necessity, that it is very largely a commercial vehicle, not used, it is true, for carrying goods in that sense, but used by doctors and travellers, and by many people for other than purely pleasure purposes. In that sense I do not think a motor car can be classed as a luxury, and, therefore, should not be taxed as such." (Hansard, 1909, p.652). He goes on to provide informative figures on the ownership and usage of vehicles: "In 1905…there were 2732 motor cars of the average value of £374. Therefore the fashion is not so very luxurious after all. A very large proportion were small power cars. In 1906 the motor cars travelled 44, 352,000 miles, and there were only 16 accidents." (ibid. p.653). He then proceeds to set out a range of arguments around vehicle taxation as relevant today as they were then, including whether it is reasonable to charge a flat rate for access to the roads, whether the funds raised should be ring-fenced for road maintenance, what justification there may be for charging motor cars but not horses (for which we may now read bicycles), and whether the tax should be graded on the basis of size, engine size/power or the amount of dust (i.e. pollution) resulting from them.
A comprehensive history of UK VED is provided in Butcher (2015), but key changes to the basic framework established at the start of the 20th Century are set out in Table 1. Up until 1992, in addition to VED, the UK had a 10% car purchase tax, but this was ended as part of plans to increase fuel duty with the 'fuel duty escalator' (see below). However, a new graded first year rate of VED was introduced in 2008 (see Table 1). Current rates of VED are shown in Table 2. Exemptions currently exist for a number of vehicles under the Vehicle Excise and Registration Act 1994, in particular "electrically propelled vehicles" and "light passenger vehicles with low CO 2 emissions (i.e. that the emissions figure for the vehicle does not exceed 100 g/km)". Disabled people are also exempt, however it has not been possible to account for this within this study. It is also worth noting that for people who do not wish to or cannot afford to pay VED in an annual lump sum, options to pay monthly by Direct Debit or for only six months increase costs by 5% and 10% respectively.

Fuel costs
Fuel costs are comprised of two main elements: basic costs of fuel and taxation. In the UK, fuel duty for petrol and diesel (and biofuel equivalents) is one of the highest in the world at £0.5795 per litre, with standard rate VAT added on top (OECD, 2013). Between January 1990 and October 2015, this resulted in the total tax being paid on a litre of petrol comprising between 53% and 86% of the total pump price. When adjusted for inflation, petrol prices have increased by only 18% overall between October 1990 and October 2015 (from £0.85/litre to £1.08/litre), however there have been significant price spikes, with a maximum in April 2012 when petrol costs reached a 2015 equivalent of £1.47/litre (see Fig. 2).
Between 1992 and 1999, in a move towards an increasing tax on use rather than ownership, the UK government introduced the 'fuel duty escalator'. This was an annual increase in fuel duty above the rate of inflation. Initially it was a 5% per annum increase, and then from 1997 it increased to 6% (Potter and Parkhurst, 2005). The initial intention of the Conservative government was to double the price of fuel at the pump in order to a) encourage manufacturers to develop more efficient vehicles, b) discourage non-essential car-use, and c) provide a more even playing field for public transport (Gray et al., 2001). The tax allegedly went largely unnoticed for most of the decade due to falling oil prices in real terms. However, as prices began to rise in 1999 its effects started to become more apparent, particularly for road haulage companies, leading to campaigns to abolish it. Even after it was abolished in November 1999, increases in the price of crude oil led to continuing price rises at the pump resulting in campaigns to reverse the historic increases and, eventually, to the UK-wide fuel protests of September 2000 (Lyons and Chatterjee, 2002;Santos and Catchesides, 2005;.

Methodology
Through analysis of vehicle characteristics (year of registration, engine size, and fuel type) and the annual distance driven, it has been possible to estimate both annual VED and fuel costs for every private vehicle in Great Britain, including cars, minibuses, vans (<3.5 t) and two and three wheeled vehicles, and to consider these figures in association with income data. Due to limitations of available data on income, we have only performed the analysis for England and Wales. This analysis has focussed on 2011 in order to utilise UK Census data from that year Table 1 Key changes to UK VED since the start of the 20th Century. Year Key change to Vehicle Excise Duty (VED)

1889
VED introduced for four-wheeled motor vehicles by the Customs and Inland Revenue Act 1888 1896 Locomotives on Highway Act 1896 introduces additional duty of 2-3 guineas based on vehicle weight. 1909 The Central Road Fund was created to make roads self-financing based on new graduated tax on cars based on horsepower and a new tax on imported oil. 1937 The ending of hypothecation of the motor vehicle taxes so that it went into the Consolidated Fund and was no longer used specifically for highway maintenance. 1978 Labour proposals to end VED and replace it with a 20 pence-per-gallon increase in fuel duty.

1980
A decision by the Conservatives to retain VED in order to enable the maintenance of a vehicle register for "the police and vehicle control". 1990 The Institute for Fiscal Studies produced a report on Environmental Taxes (Pearson and Smith, 1990) which explored different ways in which taxes could be levied to cover environmental externalities. It concluded that a move from taxing ownership to taxing use could cut fuel consumption by 8%, but that there would be potential negative effects on rural populations and other locked-in high mileage users (though this might be countered in the long-run by a shift to more efficient cars) and possible inflationary effects on the economy by increasing the costs of goods moved by road.

1994:
A Royal Commission on Environmental Pollution report into Transport and the Environment (RCEP, 1994) favoured covering external costs through charges related to use rather than ownership.

2001:
Following announcements in the 1999 and 2000 Budgets, a new graded VED system was put in place. For all new cars registered from August 2001 onwards, a VED banding system was established based primarily on CO 2 emissions but with "a discount for 'cars using cleaner fuels and technology and a small supplement on diesel cars to reflect their higher emissions of particulates and other local air pollutants" (HM Treasury, 2000, para 6.62). Cars registered between 1973 and 2001 were split into two groups: smaller engine cars <1,200cc and larger engine cars. Cars registered before 1973 were classed as 'classic cars' and exempted under the 1998 Budget.

2002
New VED rates were applied to motorcycles to encourage their use in the place of cars, especially for commuting purposes due to lower impacts on the environment and congestion.

2006
A new seven band (A-G) VED system, still with petrol/diesel differentials was introduced, however the top rate only applied to vehicles registered from March 2006 onwards. Separate rates were introduced for vans (pre-and postÀ2001, and with a separate rate for Euro 4 vehicles). Cars preÀ2001 were split into two bands based on engine size above or below 1,559cc.

2010
Thirteen different CO 2 bands were created "to strengthen the environmental signal" along with differential first year rates of VED (HM Treasury, 2009).

2008
A new graded first year rate of VED was introduced: zero for cars emitting <130 g/kmCO 2 , equivalent to the standard rate for vehicles emitting between 131 and 160 g/ kmCO 2 , and £950 over this amount.

2015
For cars registered after 1st April 2017, first year rates will vary according to the CO 2 emissions of the vehicle, varying from £0 to £2000 across 13 CO 2 bands. A flat Standard Rate (SR) of £140 will apply in all subsequent years, except for zero-emission (i.e. electric) cars for which the SR will be £0 (rather than the current 100 g/km CO 2 threshold). Cars with a list price above £40,000 will attract a supplement of £310 on their SR for the first 5 years in which the SR is paid (HM Revenue & Customs, 2015). Through emphasising CO 2 emissions at the point of purchase, the new regime supposedly puts more pressure on manufacturers to reduce emissions. Controversially, the funds will again become hypothecated to establish a new 'road fund' (previously abandoned in 1937).
T. Chatterton et al. Transport Policy 65 (2018) 30-39 (since the UK Census only occurs every ten years). The basic principles of this analysis are set out in detail in Chatterton et al. (2015) but are summarised below. Further to that analysis, the MOT test record dataset has been 'enhanced' through the addition of a number of new parameters that have been acquired through a UK vehicle stock table from the Driver Vehicle Licensing Agency (DVLA). In particular, the DVLA data allows the linking of each vehicle to the Lower-layer Super Output Area (LSOA) of the registered keeper (a relatively socially homogeneous area of, on average, around 700 households, and 1600 persons); the CO 2 emissions (available for approximately 68% of vehicles, i.e. the vast majority of those registered after 2001 when it is relevant for the VED calculations); as well as an indication as to whether the vehicle is registered by a private individual or a corporate entity. This last parameter has allowed us, for the purposes of this analysis, to investigate only privately owned vehicles. Also, the provision of data from the DVLA stock table has allowed the identification and tracking of vehicles less than three years old. For the purposes of this analysis, the fields of interest from the MOT/DVLA dataset are: LSOA of registered keeper, date of first registration, MOT test class, fuel type and engine size. The analysis is done for all LSOAs in England and Wales, and unless stated otherwise, where figures for vehicle costs or fuel use are given per household, these refer to only those households with cars.
Following a modified version of the methodology set out in Wilson et al. (2013), an estimate of annual distance (km) travelled has been calculated for each vehicle. For vehicles without a valid MOT test in the base year (2011) due to being less than three years old, the annual distance has been estimated by taking the odometer reading at the first (post-2011) test and averaging this between the date of the test and the date of first registration. Then, using the methodology from Chatterton et al. (2015), the fuel economy (litres/100 km) has been calculated for each vehicle and a CO 2 rating (g/km) calculated for those vehicles which do not have an official CO 2 emissions banding from the DVLA data. Where any vehicle does not have complete data for a field, this has been infilled with an average value for the other vehicles from that area. Where vehicles do not have a valid fuel type, these have been classified as petrol.
Then, on the basis of MOT test class, registration date, engine size and CO 2 emissions, each vehicle has been placed in a VED class and assigned an annual VED rate according to the categories set out in Table 2. On the basis of the annual km travelled, fuel economy and fuel type, the annual fuel consumption and cost for each vehicle was then calculated. The latter was based on 2011 average prices of £1.33 per litre for standard unleaded, £1.39 for diesel and £0.73 for LPG (DECC, 2012). In the absence of prices from DECC or other UK sources on the cost of CNG as a road fuel, this has been set to £0.54, based on the LPG: CNG cost ratio obtained from the US (USDoE, 2015). For electric vehicles, a figure of £0.033 per km has been used based on an average 2011 domestic electricity price of £0.141 per kWh (DECC, 2015) and an 80kw Nissan Leaf using the NextGreenCar fuel cost calculator. 1 Costs have been allocated to households, and households with cars, using 2011 Census data about the numbers of each in each local area. Income data has been used from Experian estimates of median income (Experian, 2011).  VED per kilometre are £0.159/km. Fig. 4 shows maps of average household expenditure on VED and road fuel (for those households with cars). The left hand two maps are scaled in deciles. Urban areas stand out particularly sharply on these maps because, even though households without cars have been excluded, in these areas those households that have cars still tend to own fewer vehicles than in rural areas, leading to much lower average per household costs. This may be because there is less need for cars due to greater accessibility of services and/or better public transport provision, or it may be due to prohibitive factors such as higher on-street parking charges or significantly higher property prices for urban properties with off-street parking. These latter are, however, examples of costs that we cannot account for in this analysis. The bivariate plot on the right allows the identification of areas of high VED/low-medium fuel costs which are mainly suburban areas on the periphery of London and the Home Counties. This combination is likely to denote areas of greater wealth but lower mileage vehicles (potentially strongly correlated with levels of rail commuting). In general, rural areas are particularly characterised by high VED and high fuel costs. Areas with lower VED but high mileage appear to be more prevalent in the north of England and in Wales. Fig. 5 shows differences in expenditure on road fuel between urban and rural areas. It uses the UK Office for National Statistics Urban-Rural categorisation (Bibby & Brindley, 2013) which groups areas into classes (A: Major Conurbation, B: Minor Conurbation, C: City and Town: D: Rural Town and Fringe and E: Rural Village). It is evident that, in general, urban areas (A/B/C) lead to lower expenditure on road fuel and rural areas (D/E) spend significantly more on road fuel with a gradual increase as areas become more rural. The plots are Tukey style box and whisker plots created using R software (R Core Team, 2012) and where notches of two plots do not overlap there is 'strong evidence' that the two medians differ (Chambers et al., 1983, p. 62)).   6 shows average household expenditure on VED (for households with cars) in relation to median household income at the LSOA level (Experian, 2011). The plots indicate a significant increase in outlay on VED with increasing income. In the left-hand plot, there is a notable downward spike where there are lower household VED costs at lower incomes. Comparing this to the right-hand plot, it is evident that these are tending to occur in the second, third and fourth income quartiles. Fig. 7 shows average household expenditure on road fuel (for households with cars) in relation to median household income at the LSOA level. This indicates that although there is a tendency for expenditure on fuel to increase with income, this is not nearly as strong as for VED (R¼0.30 as opposed to R¼0.57 for VED). Of note in the scatter plot are some areas that stand out with low income/low fuel costs, and high income/low fuels costs. The box and whisker plot indicates that the former tend to be in the second to fourth income deciles rather than the lowest and they also appear to correspond to a similar effect observed for VED in Fig. 6.

Comparison with expenditure on domestic gas and electricity
Given the increasing push to electrify transport, as well as space/ water heating and cooking, there is a need to begin to understand how energy use from cars relates to domestic energy consumption (HM Government, 2011. Fig. 8 shows data from the Living Cost and Food Survey (ONS, 2012) for relative expenditure on domestic energy. These range from £723 for the lowest income decile to £1149 for the highest. This compares with the greater range for the fuel component of motoring costs in Fig. 1 running from £260 to £2574.
For the work presented in this paper, average prices for gas and electricity were calculated from the UK Department of Energy and Climate Change 2012 Quarterly Energy Report (DECC, 2012) for a kWh of gas and electricity based on a 'typical' annual household consumption of 18,000 kWh and 3300 kWh respectively. The calculated prices based on the standard credit payment (not direct debit or prepayment, and taking no account of 'Economy 7 0 (dual tariff) differentials) across all suppliers was £0.042 per kWh for gas and £0.143 per kWh for electricity. These were then applied to LSOA level data from DECC on average household gas and electricity consumption (DECC, 2014). Use of other fuels (oil, bottled gas, solid fuels etc.) has not been incorporated into the analysis, but as Fig. 8 shows, this is a small fraction of expenditure overall. However, it is also very unevenly distributed, particularly with regard to where use is due to properties not being connected to the mains gas grid (see Chatterton et al., 2016a). Fig. 9 provides a comparison of the fuel costs of car use alongside expenditure on domestic gas and electricity consumption. Average household expenditure on gas and electricity (for households with a gas or electricity meter) tends to increase together, although the distribution indicates expenditure on gas compared to electricity varying by up to a factor of two. In terms of expenditure on road fuel (by households with cars), again expenditure increases together, with those households spending more on one, tending to spend more on the other. However, there is a divergent tendency in the areas of higher expenditure, with one cluster having very high expenditure on road fuel but not on domestic energy, as well as a group that have lower expenditure on car fuel but high domestic energy consumption.

Proportion of income spent on VED, road fuel and domestic energy
In order to better evaluate the financial impact of expenditure on VED, road fuel and domestic energy in different areas, the average household expenditure has been calculated as a percentage of median income for each LSOA. The plots in Fig. 10 show on the x-axis, the mean of the median income values for each income decile, and on the y-axis, Fig. 5. Average household annual road fuel costs (for households with cars) by ONS Urban/Rural classification (Y-axis cropped at £3000 to exclude extreme outliers). Fig. 6. Average household annual VED costs by Experian median income and income decile (Y-axes cropped at £600 to exclude extreme outliers).
T. Chatterton et al. Transport Policy 65 (2018) 30-39 the mean expenditure as a percentage of income for these (with (very small) 95% confidence intervals around the mean, and noting that these are the means of the area aggregatesnot of individual households). Following Santos and Catchesides (2005), costs for road fuel and VED are presented for all households and only those households with cars. Then, for domestic energy costs and total costs (VED þ road fuel þ domestic energy), results are only provided across all households as it is not possible to attribute differentials in domestic energy use separately to households with and without cars. Overall, the percentage of income spent on motoring costs decreases as income increases, with the lowest income deciles spending around twice as much of their income on the car and domestic energy components as the highest income deciles. When the motoring costs are examined across all households, and not just ones with cars, this effect is still present but less strong and with a flattening out of the curve for the second to fifth percentiles. Fig. 11 presents the data on spending as a percentage of income spatially (white areas on the map are areas where for technical reasons income data wasn't available). As with Fig. 4 these are scaled in deciles. These same deciles (based on expenditure for households with cars) have been used for both maps to highlight the differences more clearly. The maps show a strong tendency for the proportion of income spent on fuel and VED to increase towards the peripheries of the country as wages and accessibility reduce, and to decrease along the spine of the country and particularly around London where income and connectivity are highest.

Discussion and conclusions
This analysis has taken a novel approach to the calculation of    iii) There are other household costs relevant to mobility that have not been considered, such as expenditure on public transport. Here, we have compared motoring costs in relation to expenditure on domestic energy consumption, due both to the availability of readily compatible data, but also because of the increasing inter-relation between these due to the current and predicted trends towards the electrification of vehicles. However, there are other spatial data that might merit consideration in future work, such as housing costs. Further work in this area would be beneficial, as although theory suggests that households trade-off increased housing costs with transport costs, evidence often suggests that things are much more complicated than this (Mattioli et al., 2016). It is also important that average house price data is considered in conjunction with information on tenure. However, given the universal nature of the data sources used here (arising from annual readings from~26 million electricity meters,~24 million gas meters and~29 million vehicle odometers and vehicle profiles), this analysis should provide a valuable insight into patterns of  T. Chatterton et al. Transport Policy 65 (2018) 30-39 expenditure both in its own right, and in comparison to studies based on data from different sources. There is a significant debate about whether existing taxes on car use are socially regressive which relate to the extent to which car use can be considered a luxury or a necessity (see, for example, Sterner, 2012; Santos and Catchesides, 2005). The relatively simple analysis provided here does not provide great insight into how 'essential' cars are for different people, or in different locations. However, it does indicate the strong tendency for expenditure on VED and fuel costs together with other household energy costs to be regressive, in that expenditure on these items represents a higher proportion of household income at lower income bands, particularly if only households that own cars are considered. The actual effects of this are likely to be greater in actuality than represented here due to the inability of poorer households to pay by the cheapest means which will exacerbate these costs (e.g. if unable to pay a 12-month lump sum for car tax, they must pay 5-10% extra, or they may have pre-pay electricity meters which cost more). Moreover, although expenditure on fuel often has a discretionary element to it, for many people, some car use will be regarded as a basic need and so, however low income is, expenditure will not reach zero. At the same time, it needs to be remembered that a significant proportion of households (26%) don't have access to a car and are reliant on other forms of transport (particularly public transport), which, in turn, may be dependent on tax revenue to operate. Consequently, the case for reducing motoring taxation as a socially progressive policy is highly complex.
It can be argued that the grading of VED by age and CO 2 band of vehicle enables it to be less regressive as a mode of taxation than a fixed rate, as it allows people to effectively choose what rate of tax they are happy to pay and to choose a vehicle accordingly. However, in reality, whilst vehicle size is often a choice, it is also the case that newer vehicles (which tend to be more efficient and therefore attract lower VED) also tend to be more expensive, whilst older, more inefficient cars which attract higher rates of VED may be more affordable at the point of purchase, locking poorer households into higher running costs in the longterm (Lucas and Pangbourne, 2012). Future work will enable investigation of the interplay between vehicle age, size and price, and the extent to which VED appears to have influenced purchasing patterns by different income groups.
The future changes to VED that are due to apply from 2017 will set a standard rate of VED at £140 after the first year for all except electric vehicles and thus remove (except at purchase) any VED incentive towards purchasing cleaner non-electric vehicles. It may be the case that we are moving to a time in the uptake of electric vehicles where this absolute tax distinction between 'zero-emission' and 'polluting' is appropriate. However, VED is not the only way in which those able to afford to purchase EVs will enjoy significant financial benefits, as not only are EVs more efficient to run in terms of energy, but, in the UK, the fuel is taxed significantly less. In 2015, domestic electricity invoked a total tax of 5% VAT, compared to a mean total tax of over 68% on petrol. 2 Given that the initial purchase price of electric vehicles is relatively high, the greater ability of the wealthy to purchase access to cheaper mobility through EVs is going to have significant implications both for social justice and the Government's tax revenue. However, increasing tax on electricity would potentially only exacerbate the already regressive nature of energy prices illustrated above.