1 Introduction

Energy poverty is increasingly recognised as an important research topic (Bouzarovski and Petrova 2015) and as an area of policymaking in the European Union (EPOV 2020). Yet, the notion of energy poverty remains overwhelmingly focussed on energy consumption within the home, while similar issues related to transport energy have been overlooked. In a policy context where the energy- and low-carbon transition is a priority, the exclusion of transport from the energy poverty debate can hardly be sustained.

A number of metrics show that transport energy consumption is as much, if not more relevant than energy consumption within the home. In 2020, transport accounted for 11.6% of household expenditure in the EU-27, second only to housing and food (Eurostat 2021), and down from the even higher levels observed prior to the COVID-19 pandemic. The average share of expenditure on the ‘operation of personal transport equipment’ (mostly cars) was higher than that on ‘electricity, gas and other fuels’ within the home in the EU-27 (6.5% vs. 4.3%) as well as in 22 member states. In 2020, transport (both passenger and freight) accounted for 28.4% of final energy consumption in the EU-28, as compared to 28.0% for households (Eurostat 2022a), with the vast majority of transport energy coming from oil and petroleum products (EEA 2019a). 61.5% of EU oil consumption in 2020 was for transport, with electricity accounting for just 0.1% of fuel used in road transport. In 2019, the transport sector accounted for 25.8% of greenhouse gas emissions in the EU, the same share as fuel consumption by energy users for other purposes (Eurostat 2022c). Persistent reliance on fossil fuels and growing levels of travel activity, notably by car, explain why EU greenhouse gas emissions in the transport sector have increased by +19% between 1990 and 2018, in contrast with other sectors where they have declined (e.g. −22% in the residential and commercial sector) (EEA 2019b). In Europe and beyond, transport is thus the sector where effective climate policy is most urgently needed (Creutzig et al. 2015; Lamb et al. 2021).

Current EU policy aims at achieving decarbonisation through carbon pricing, which would encourage both energy efficiency and a shift from high- to low-carbon modes of energy consumption. In the transport sector, the European Commission has proposed the extension of the Emissions Trading System to road fuels from 2026 (EC 2021), which is expected to result in higher fuel prices at the pump for internal combustion engine vehicles. More broadly, the increase of taxes on (and removal of subsidies to) road fuel is considered a key climate policy measure globally (Ross et al. 2017).

Yet, transport systems in much of the EU are ‘car dependent’, as access to and the ability to use car-based travel is often essential for accessing services and opportunities and achieving social inclusion (Mattioli 2016, 2021). Despite a recent boom in electric vehicle (EV) sales, and EU plans to phase out combustion engine vehicles by 2035, the vast majority of cars on European roads still run on petrol or diesel (95.2% in 2019, see ACEA 2021). This will continue to be the case for many years, as vehicle fleet turnover is slow.

The affordability of road fuel thus looms large in the public and political debate. This is most apparent when fuel tax increases trigger mass protests or disruptions, as in the UK in 2000 (Lyons and Chatterjee 2002) or in France in 2018 with the ‘Yellow Vest’ movement (Mehleb et al. 2021). In the aftermath of Russia’s invasion of Ukraine and the resulting fuel price spike in 2022, most EU governments substantially cut fuel taxes to ensure the ongoing affordability of car use, despite the negative geopolitical implications of doing so (Gars et al. 2022; Transport and Environment 2022a). Concerns about the social and distributional impact of making car use more expensive (whether genuine or feigned) are also often raised by opponents of carbon pricing (Maestre-Andrés et al. 2019; Lamb et al. 2020). This helps explain why governments struggle to introduce CO2 pricing in the transport sector, despite the pressing need to reduce transport emissions.

The paradox here is that, while concerns about vulnerability to motor fuel price increases are widespread and consequential, the ‘transport equivalent of energy poverty’ is overlooked by energy research and has not coalesced into its own area of policymaking. This is now starting to change, with researchers calling for a broader understanding of energy poverty that includes transport (e.g. Martiskainen et al. 2021; Mattioli et al. 2017; OpenExp 2019), and the EU considering the adoption of an official definition of ‘transport poverty’ (Taylor 2022). To date, however (to the best of our knowledge), no EU country except France (Cochez et al. 2015; ONPE 2014) officially recognises transport as a dimension of energy poverty. This results in a dearth of definitions and indicators, so that in practice we know little about how many people are vulnerable to fuel price increases, who they are and where they live.

Spatial patterns in the distribution of vulnerability to fuel price increases are of particular interest, as the use of and reliance on cars differs dramatically between urban, periurban and rural areas, as do income levels and political leanings (Mattioli and Colleoni 2016; Walks 2015). In the public and political debate, the strong impact of fuel price spikes on the residents of car-dependent (and sometimes less affluent) peripheral areas is often contrasted with that on the generally more environmentally minded urban population. And yet, these discussions are rarely grounded in sound analysis of vulnerability to fuel price increases and its various dimensions. There is a need for a more rigorous empirical basis to inform and raise the level of these debates.

In this chapter, we present findings on spatial patterns of vulnerability to fuel price increases in Italy, a country where the problem is likely to be particularly pronounced due to high motorisation rate relative to income, as well as high fuel prices. In doing this, we have two goals. First, to provide an illustration of how a composite spatial indicator of vulnerability to fuel price increases can be built based on data from official sources. Second, we explore the spatial relationship between the different factors underlying vulnerability to fuel price increases in Italy, and contrast the findings with previous research from other countries.

The chapter is structured as follows. We start by reviewing the literature on the affordability of car use and vulnerability to fuel price increases, discussing the concepts, empirical indicators and patterns of spatial variability identified in the literature to date (Sect. 2). We then provide information on the case study country with regard to factors that might influence vulnerability. In Sect. 4, we explain our methodological approach to building the composite spatial indicator, and then present the empirical findings in Sect. 5. We conclude by contrasting our findings with previous research and discussing implications for future research and policy-making (Sect. 6).

2 Literature Review

2.1 Affordability of Car Use and Vulnerability to Fuel Price Increases: Concepts and Indicators

Transport research has long recognised that low-income households who own and use cars can end up spending disproportionate amounts on motoring. This can lead them to curtail travel to save on running expenses, and/or to cut expenditure in other areas (e.g. domestic heating), all of which can reduce social inclusion and well-being (Froud et al. 2002; Lucas 2011).In the literature, various terms are used to point to this problem, including ‘forced car ownership’ (Banister 1994; Carroll et al. 2021; Curl et al. 2018; Currie and Senbergs 2007; Jones 1987; Mattioli 2017), ‘car-related economic stress’ (Belton-Chevallier et al. 2018; Mattioli et al. 2018; Mattioli and Colleoni 2016; Rock et al. 2016), ‘transport energy precarity’ (Cochez et al. 2015; Jouffe and De Massot 2013) and ‘transport energy poverty’ (OpenExp 2019; Robinson and Mattioli 2020). What these notions have in common is that they present the problem as a ‘static state’. Other concepts such as ‘oil vulnerability’ (Dodson and Sipe 2007; Leung et al. 2018; Lovelace and Phillips 2014; Rendall et al. 2014; Runting et al. 2011), ‘transportation energy vulnerability’ (Liu and Kontou 2022) and ‘vulnerability to fuel price increases’ (Mattioli et al. 2019) emphasise the dynamic aspect of the problem, highlighting which people and places would be more likely to experience hardship if the costs of motoring were to increase.

Our contribution in this chapter fits into the second strand of the literature, i.e. we adopt a vulnerability perspective. However, all indicators of the affordability of car use can inform a discussion of vulnerability to fuel price increases and, more broadly, of energy poverty in the transport sector. In the European literature, several quantitative indicators have been proposed to assess the affordability of car use. They can be classed into three categories, as illustrated in Table 1:

Table 1 Overview of indicators of the affordability of car use in the European context
Full size table
  1. (i)

    adaptations of domestic energy poverty indicators for use in the transport sector; these are typically based on a ratio between expenditure and household income (or total household expenditure as a proxy for income), based, e.g. on household budget survey data

  2. (ii)

    ‘forced car ownership’ indicators typically identify households who own cars despite being in deprivation, which it is assumed might lead to affordability problems;

  3. (iii)

    composite indicators capture the multidimensional nature of the phenomenon; this group includes indicators of vulnerability to fuel price increases, where vulnerability is often conceptualised as the product of exposure, sensitivity and adaptive capacity (as discussed).

One can also distinguish whether the indicators are based on survey, census or modelled data (or combinations thereof); and whether the unit of analysis is households/individuals and/or spatial units. Considering all these aspects gives the complex picture illustrated in Table 1.

While many of the affordability indicators proposed in Europe (Table 1) consider the costs of car use only (e.g. Berry et al. 2016; Cochez et al. 2015; Mattioli et al. 2018), others include expenditure on public transport as wellFootnote 1 (e.g. Nicolas et al. 2012; Lovelace and Philips 2014; Verry et al. 2017). In practice, however, car use tends to account for most transport-related expenditure in high-income countries, both in the aggregate and for most households (Kauppila 2011). Studies on vulnerability to fuel price increases typically focus on the direct impact on the cost of car use. While fuel price rises have an indirect impact on the costs of some public transport modes like buses, this is more indirect and of smaller magnitude, as labour costs account for a large share of final costs.

As shown in Table 1, most household-level studies use adaptations of indicators of domestic energy poverty. Studies focussed on spatial units (e.g. municipalities or census units) often use composite indicators. The first studies in this vein were conducted in Australia and used the concept of ‘oil vulnerability’ (Dodson and Sipe 2007; Runting et al. 2011), which is essentially equivalent to vulnerability to fuel price increases. As argued by Leung et al. (2018) and Mattioli et al. (2019), most of these indicators conceptualise vulnerability as being constituted by three components: exposure, sensitivity and adaptive capacity (see Adger 2006). This is illustrated in Table 2, along with examples of indicators for the sub-dimensions from the literature.

Table 2 The three dimensions of vulnerability to fuel price increases: definitions and examples of indicators
Full size table

In a nutshell, spatial indicators of vulnerability to fuel price increases identify as most vulnerable areas characterised by high levels of car use (high exposure), low economic resources (high sensitivity) and with reduced opportunities to shift from car use to other modes (low adaptive capacity). In practice, however, the adaptive capacity dimension is not always taken into consideration, whether due to data availability limitations or to the assumption that it correlates strongly with exposure—i.e. those areas with high levels of car use are also characterised by high levels of car dependence (low availability and viability of alternative transport modes). For a review of indicators of exposure, sensitivity and adaptive capacity used in the literature see Leung et al. (2018) and Mattioli et al. (2019).

2.2 Spatial Patterns of Vulnerability to Fuel Price Increases

Most studies on the affordability of car use find that the problem is worse in peripheral, periurban and rural areas as compared to city cores (e.g. Cochez et al. 2015; Liu and Kontou 2022; Lovelace and Philips 2014; Mattioli and Colleoni 2016; Nicolas et al. 2012; Simcock et al. 2021; Verry et al. 2017). This is mainly due to higher levels of car use and car dependence. From a vulnerability perspective, this means greater exposure and less adaptive capacity to fuel price increases. However, some studies show that low-income households can be reliant on and reluctant to do without cars even in large cities, which results in economic stress (e.g. Curl et al. 2018).

This pattern is compounded when suburban and periurban areas are less affluent than city cores, which makes residents less able to afford the costs of car use. From a vulnerability perspective, this means that greater sensitivity to fuel price increases adds on to greater exposure and lower adaptive capacity, in a ‘triple whammy’ of sorts. Research has found this to be the case in several constituencies including Australian cities (Dodson and Sipe 2007; Runting et al. 2011) and the Munich city region in Germany (Büttner et al. 2013).

However, the opposite pattern has been observed as well, i.e. when the periphery is more affluent than the core of the metropolitan area. That is the case, e.g. in New Zealand (Rendall et al. 2014), in Lyon, France (Büttner et al. 2013) and in most of England (Mattioli et al. 2019). From the perspective of vulnerability to fuel price increases, this means that city cores tend to compensate higher sensitivity with less exposure and better adaptive capacity. Conversely, in periurban areas higher income (i.e. low sensitivity) can protect residents from the worst consequences of fuel price increases even if car use is high (i.e. high exposure) and there are little modal alternatives to the car (i.e. low adaptive capacity). These counteracting effects explain for example why household-level studies in the UK have found an even incidence of ‘car-related economic stress’ and ‘forced car ownership’ across the urban–rural spectrum (Mattioli 2017; Mattioli et al. 2018).

From this perspective, the trend towards the gentrification of city cores and the ‘suburbanisation of poverty’ is to be regarded critically, as it leads less affluent groups to find affordable housing in the car-dependent areas. This exacerbates problems of transport affordability and vulnerability to fuel price increases (Allen and Farber 2021; Coulombel 2018; Currie and Delbosc 2011; Dodson and Sipe 2007; Mullen et al. 2020; Polacchini and Orfeuil 1999; Sterzer 2017).

Most spatial studies on vulnerability to fuel price increases look at differences within metropolitan areas, and as such are unable to compare different regions to each other. Interregional differences can be important though, especially in countries with regional divides in terms of economic development. A recent study in England (Mattioli et al. 2019), e.g. found that metropolitan areas in the North are more vulnerable to fuel price increases than London and the South-East, on account of both lower levels of income and worse public transport provision. The study also highlights the complex interplay of vulnerability dimensions at multiple spatial scales: adaptive capacity and sensitivity tend to compensate each other within English metropolitan areas (as car-dependent periurban areas tend to be more affluent) but they compound each other at the interregional scale (as poorer regions also have worse public transport provision).

3 Case Study

Our study focusses on Italy. This is a country for which little evidence of vulnerability to fuel price increases exists, even though it is likely to be particularly vulnerable. While there is a moderate correlation between income and motorisation rate for EU regions (Pearson’s R =  +0.44), most Italian regions have higher motorisation rates than one would expect based on income aloneFootnote 2 (Fig. 1). This suggests that many households in Italy own and operate vehicles despite low income, which may lead to affordability problems. In other words, Italian regions are characterised by a combination of relatively high exposure and relatively high sensitivity to fuel price increases in European comparison.

Fig. 1
A scatter plot of passenger cars per 1000 inhabitants versus purchasing power standard per inhabitant. The line of fit for both Italian N U T S 2 and other N U T S 2 is increasing.

Relationship between motorisation rate and income of households for EU regions (NUTS2) in 2019, with line of fit. Source Authors on Eurostat data (Eurostat 2022d, 2022e)

Full size image

Until the 2022 war in Ukraine, Italy had some of the highest petrol and diesel prices in the EU, partly due to high taxes (Fuels Europe 2021). This reduces the affordability of motor fuel, even though the high share of taxation in the end consumer price might also cushion the impact of global oil price fluctuations and of additional environmental taxes (as the final price is less sensitive to these changes in relative terms).

Kokoufikis and Uihlein (2022) find that Italy had the second-highest average share of household expenditure on transport fuels of all EU countries in 2015 (4% in densely populated areas, and over 6% in sparsely populated areas). The average share of household expenditure on personal transport was particularly high for working households and for couples (over 6%). In 2018, Italy was third in the EU-28 for the share of transport energy expenditures out of total expenditures of the first income quintile of the population (5.2%) and was ranked seventh-worst performer for its overall performance in alleviating transport energy poverty (OpenExp 2019). Recent modelling work finds that, in a scenario where oil prices double, Italy would be one of the most impacted EU countries in terms of average additional household expenditure (nearly +10%, and over +20% for the 5% of households that are most affected) (Steckel et al. 2022). The Italian parliament estimated that between June 2021 and September 2022, the spike in energy and other prices increased average household expenditure by +3.7%, even when mitigation measures are taken into account (Ufficio Parlamentare di Bilancio 2022). Even though there is no official definition or indicator of transport energy poverty in Italy at present, the second report of the Italian Energy Poverty Observatory recognises transport as an important dimension (Faiella et al. 2020).

Italy is characterised by profound spatial inequalities, which are likely to have a bearing on the geography of vulnerability to fuel price increases. The country is well-known for the strong and long-standing North–South divide, with the latter being worse off in economic, infrastructural, and socio-institutional terms (Felice 2018). The typical ‘urban socio-spatial configuration’ of Italian cities is a concentric one, with the core being richer than the periphery, although this pattern is less clear in the South (Kesteloot 2005).

4 Methods

We propose a composite indicator of vulnerability to fuel price increases for Italian municipalities that includes four variables covering two dimensions: exposure and sensitivity. Ideally, the composite indicator should cover adaptive capacity as well. However, information on the availability of transport modes alternative to the car is available for only a few Italian municipalities.

We measure exposure with indicators of car ownership and use. We draw this information from the 2011 Census of population and housing (ISTAT 2011), which asks respondents about the number of household cars, as well as the destination and main travel mode (in terms of distance) of the journey to work or education for those who habitually make such trip. This information is made available for Italian municipalities, but not for more disaggregate spatial units. We derive three variables: (i) the percentage of households owning at least one car; (ii) the share of workers or students who regularly travel to work or education who use a private car (either as driver or passenger) for that trip; (iii) an estimate of the average distance of commuting from home to the place of work or study by car.

We derive the last variable from an origin/destination matrix reporting commuting flows between Italian municipalities by travel mode, derived from the Census. We distinguish between two types of commuting trips: external journeys, where people move between two different municipalities, and internal journeys, where people commute within the same municipality. For the first type, the distances by car between each pair of municipalities were calculated in kilometres considering the two centroids and a road graph.Footnote 3 For the second type, two random points (origin/destination) for each internal movement were randomly defined within each municipality, which is an approach adopted by previous research (e.g. Lovelace et al. 2022). Then, the distance between these pairs of points was used for determining an estimate of the length of internal journeys. The sum of travel distance for internal and external trips for each municipality was divided by the number of commuters to calculate the average commuting distance by car.

Both journey-to-work variables were adjusted to address the skewness of their distribution and mitigate the effect of extreme outliers. First, a logarithmic transformation was applied. Second, the minimum values of each variable plus a constant equal to 0.001 were added to the obtained values in order to avoid the presence of negative and null values without altering the distribution of the transformed variables. All three variables were then further normalised using a z-score transformation to allow their comparability (calculated by subtracting the mean from the original value and then dividing by the standard deviation) and aggregated using their arithmetic mean. The resulting indicator is considered a measure of exposure.

The indicator of exposure covers 7,876 Italian municipalities (about 98% of the municipalities in 2011). The remaining municipalities were excluded for two main reasons. Firstly, it was impossible to reasonably estimate commuting distance in two types of municipalities: border areas and islands. In some municipalities bordering other countries (namely, Switzerland, France and San Marino) a large share of the population works abroad and crosses the border daily. However, the Census does not collect precise information on the place of work when this is abroad, thus making the estimation of travel distance impossible. Therefore, we excluded those municipalities with a share of cross-border commuters on the total number workers or students that is higher than the national average plus one standard deviation. Municipalities located on small islands were excluded as a large share of the population commutes to the mainland, and the journey by sea does not allow for a precise estimate of car distance. Secondly, we had to merge some small municipalities in order to match the Census data with the IRPEF income data, which takes into account the fusion of some contiguous municipalities happened after 2011 that established new administrative divisions.

To measure sensitivity to fuel price increases we use estimates of average per capita income at the municipal level drawn from the Italian Ministry of Economy and Finance (2012) data on the personal income tax (IRPEF) of individuals for the year 2012 (referring to 2011 income). The variable is normalised using a z-score and then multiplied by −1 so that higher values of this indicator (i.e. lower average incomes) correspond to higher degrees of sensitivity. Figure 2 (in Sect. 5) shows the distribution of the four transformed variables in the Italian municipalities.

Fig. 2
4 heatmaps of Italian municipalities, a to d, indicate the car ownership, car mode share for the community, average community distance by car, and the average per capita income, respectively.

Variables included in the composite indicator (in red the exposure dimension and in blue the sensitivity dimension; darker shading indicates values associated with higher vulnerability to fuel price increases). Note that the values in panel D were multiplied by −1, so that higher values indicate lower income (which is associated with higher vulnerability)

Full size image

The final score of the composite indicator of vulnerability is given by the combination of the obtained exposure and sensitivity indicators. Two alternative versions of the composite indicator were calculated. In the first one, exposure and sensitivity indicators were weighted equally assuming that both have the same relevance for vulnerability to fuel price increases. In the second one, a double weight was attributed to the exposure component. This version of the indicator assumes that actual levels of car use for commuting can also be used as a proxy of how car dependent the municipality is, capturing to some extent the adaptive capacity dimension that we are unable to assess directly. Previous research provides two justifications for this. First, there is evidence that indicators of exposure and adaptive capacity are typically more correlated to each other than to sensitivity indicators (Mattioli et al. 2019). Also, it is not uncommon for studies on vulnerability to fuel price increases to use measures of actual travel behaviour as indicators of adaptive capacity (e.g. Lovelace and Philips, 2014; Leung et al. 2018) or to merge the exposure and adaptive capacity dimensions into one (e.g., Akbari and Habib 2014; Dodson and Sipe 2007). In practice, most composite indicators of vulnerability assign a weight ranging from 50% (e.g. Dodson and Sipe 2007; Akbari and Habib 2014; Runting et al. 2011) to 66% (Büttner et al. 2013; Leung et al. 2018; Mattioli et al. 2019) to variables measuring car ownership, car use and car dependence, with the remaining 33% to 50% weight assigned to socio-economic variables such as income. Our approach in this study is consistent with this practice, while also allowing us to explore the robustness of the findings to different weighting schemes.

To ease interpretation, both versions of the composite indicator were indexed to their maximum values. The municipality with the highest value received a score of 1,000 and all the other values were rescaled accordingly. In simple terms, our composite indicator identifies as vulnerable to fuel price increases municipalities characterised by low income, high car ownership, high car mode share for the journey to work or education, and high average distance of commuting trips by car.

The methodology used for normalising and aggregating the single variables can affect the final composite indicator (Greco et al. 2019). Liu & Kontou (2022) demonstrate how this can affect metrics of vulnerability to fuel price increases. Therefore, this study adopts a strategy to evaluate the impacts of the chosen approach against alternative ones, i.e. the robustness of our findings to alternative methodological choices. Specifically, we compared the results of the composite indicator obtained by using z-scores as a normalisation approach and arithmetic mean as aggregation methods to the results based on other normalisation (i.e. indexing, ranking or min–max transformation) or aggregation (i.e. geometric mean) techniques (Dugato et al. 2014; Saisana et al. 2005). These alternatives are applied to either the calculation of the exposure sub-indicator and of the final composite indicator.Footnote 4

Overall, our indicator has some limitations that must be considered when interpreting the results. First, we lack a direct indicator of adaptive capacity. Second, the ideal measure of exposure would be an estimate of household expenditure on fuel for all travel purposes (see e.g. Liu and Kontou 2022; Mattioli et al. 2019). Since such information is not available for Italian municipalities, we use indicators that refer to car ownership and car use for commuting. While this is a common approach (e.g. Leung et al. 2018; Dodson and Sipe 2007; Lovelace and Philips 2014; Runting et al. 2011) one must keep in mind that travel to work and education accounts for only a low share of trips—i.e. 32–36% in Italy (ISFORT 2021). Third, we estimate car travel distance for commuting based on data on the municipality of origin and destination, using centroids and random points. Although common, this method could lead to biased values, particularly for municipalities with a large area. Fourth, our spatial unit of analysis is the municipality, which is relatively coarse and varies widely in terms of population size in Italy (from a few dozen inhabitants to more than 2,5 million in Rome). A study using smaller and more consistent spatial units such as census tracts would better capture patterns of vulnerability within metropolitan areas, and especially within large municipalities. These limitations notwithstanding, our study illustrates how a composite indicator of vulnerability can be derived even in countries where data availability is less than ideal. Finally, while we use the most recent data available, these are already more than ten years old. While it is possible that the situation has changed since then, we believe that our indicator captures structural features of the geography of vulnerability to fuel price increases in Italy. These are relevant for current debates, and our methodological approach can be used to update the analysis as soon as more recent data is made available.

In the next section, we present the findings using univariate and bivariate choropleth maps as well as hotspot maps and crosstabulations. While most of our analysis focusses on the entire Italian territory, we also show spatial patterns of vulnerability within two of the largest Italian metropolitan areas (Milan and Naples). A focus on metropolitan areas is in line with previous research on this topic, and can potentially inform policy-making at the local and regional level, which is where many transport policy decisions are made.

5 Results

To properly interpret the composite indicator of vulnerability, it is essential to first consider the spatial distribution of the constituent variables (Fig. 2). These show that exposure and sensitivity to fuel price increases have rather different spatial patterns in Italy.

Regarding car ownership (Fig. 2, panel A), we find a clear North–South divide, with higher shares of households with cars in the North, likely due to greater affluence. Major cities stand out from their surrounding areas as they have lower levels of car ownership, particularly in the North (e.g. Milan, Turin, Genoa, Bologna, Venice, Florence and Bolzano). This pattern is less clear in the Central and Southern part of the country, although it can be observed in the metropolitan area of Naples. The highest shares of households with cars are observed in periurban areas in the Po Valley, as well in Tuscany and in some (but not all) alpine regions. The lowest levels of car ownership are mostly in inner areas in the South, particularly along the Apennine Mountain range.

The second indicator of exposure, i.e. car mode share for commuting, shows a rather similar spatial pattern, with some differences (Fig. 2, panel B). South Tyrol has some of the lowest car mode shares in the North, despite high levels of car ownership. Besides periurban areas in the Po Valley, some of the highest levels of car use are observed in the central regions of Marche, Umbria and Lazio. There are also some areas with very high car mode share in the Southern region of Apulia and in the island of Sicily. Overall, the North–South divide is slightly less pronounced for car use than for car ownership.

Average commuting distance by car shows a more complex geography (Fig. 2, panel C). The longest estimated commuting distances by car are found in inner, mountainous and sparsely populated areas, as well as in the Lazio region around Rome. Distances are relatively low in most of the densely populated Po Valley in the North, despite the high levels of car ownership and use there.

The income indicator (Fig. 2, panel D) shows the well-known divide between North and South of the country. Note also how large cities and their immediate surroundings stand out as richer than periurban areas in the North, in Tuscany and in and around Rome, while this pattern is much less clear in the South. This is consistent with what is known from urban research (Kesteloot 2005).

Figure 3 shows the two versions of the composite indicator of vulnerability according to the alternative weighting procedures. The results are consistent and highly correlated (Pearson’s R =  + 0.87). The Central and Southern regions of the country concentrate most of the highly vulnerable areas. This is mostly due to the influence of the sensitivity dimension, as these regions are traditionally poorer and less economically developed (as shown in Fig. 2, panel D). Further, the indicators show how large cities are usually less exposed than other minor municipalities. In this case, the reason is likely to be due to the lower exposure of urban populations versus rural or suburban ones who rely more on car use. These considerations are confirmed by observing the variations between the two versions of the indicator. Giving more importance in the final indicator to car ownership and use (exposure dimension) reduces the macro-regional differences and highlights the urban–rural divide.

Fig. 3
2 heatmaps of Italy indicate the 2 versions of the composite indicator of vulnerability to fuel prices. The color bar ranges from less than negative 2.5 to greater than 2.5 standard deviation.

Two versions of the composite indicator of vulnerability to fuel prices. Version one (left) with equal weights between the two dimensions and version two (right) with a double weight for the exposure dimension

Full size image

In both versions of the composite indicator, large cities in the North and parts of the Centre stand out as less vulnerable than their surrounding areas. This is because exposure and sensitivity tend to compound each other: large city residents are both less exposed (i.e. they own and use cars less) and less sensitive (i.e. they have higher incomes) than the residents of the surrounding region. This pattern is particularly pronounced in Rome and is still visible in Naples, but much less so in the rest of the South.

Overall, the first version of the indicator suggests that most of the South of Italy is rather vulnerable to fuel price increases, with some scatters of very high vulnerability in, e.g. the Eastern part of Sicily and the inner part of Sardinia. Version two shows a similar pattern but highlights more vulnerable areas in the Centre (particularly around Rome and in Umbria), as well as in some periurban and low-density areas in the Po Valley. Interestingly, in both versions the Trentino-South-Tyrol region stands out as the least vulnerable, due to a combination of high income and low car use for commuting, despite very high car ownership.

To better understand the drivers of vulnerability and what can be done about them, it is important to consider the influence of the two dimensions on the final indicator. Figure 4 shows a scatterplot and a bivariate map that categorise Italian municipalities into four clusters according to the joint distribution of the exposure and sensitivity indicators. The map shows a clear opposition between most of the South, where higher-than-average sensitivity is compensated to some extent by lower-than-average exposure (green areas), and much of the Centre-North, where high exposure is compensated by low sensitivity (red areas). This is consistent with the scatterplot, showing a moderate inverse relationship between exposure and sensitivity at the national level (Pearson’s R =  − 0.42).

Fig. 4
A scatterplot of exposure versus sensitivity and a bivariate map that categorizes Italy into 4 clusters such as high exposure and sensitivity, high exposure and low sensitivity, low exposure and high sensitivity, and low exposure and sensitivity. The plots are mostly concentrated at the center.

Classification of Italian municipalities according to exposure and sensitivity to fuel price increases

Full size image

The most critical municipalities (depicted in blue) record values higher than the national mean for both dimensions. They are a minority and largely located in rural areas, mainly in the Centre of the country and in the two main islands (Sardinia and Sicily). There are however small clusters of this type of municipalities scattered throughout the whole country (e.g. in the Southern part of Veneto and in Apulia). The least vulnerable areas combine low values for both exposure and sensitivity and are depicted in orange. They are concentrated in cities or large metropolitan areas in the Centre and North of the country. Most of the Trentino-South-Tyrol region falls into this category.

The results of the bi-dimensional analysis are further confirmed by observing that 46% of the municipalities in Central regions have high levels of both exposure and sensitivity. This percentage is significantly higher than for the other Italian macro-areas, except Islands (30%), and much higher than the national average (17%) (Table 3). Just 16% of municipalities in the South combine high exposure and high sensitivity, demonstrating that high values for the composite indicator of vulnerability in these regions are driven largely by economic deprivation.

Table 3 Distribution of Italian municipalities by cluster in each macro-area. Note: the regions are allocated as follows: North-West (Aosta Valley, Liguria, Lombardy and Piedmont), North-East (Emilia-Romagna, Friuli-Venezia Giulia, Trentino-South Tyrol and Veneto); Centre (Lazio, Marche, Tuscany and Umbria); South (Abruzzo, Apulia, Basilicata, Calabria, Campania and Molise); Islands (Sicily and Sardinia)
Full size table

When looking at municipality size (Table 4), the results confirm that large cities are less likely to experience serious vulnerability to fuel price increases. Municipalities larger than 10,000 inhabitants (accounting for 68.9% of the population in 2011) are disproportionately more represented in the lower quartile of the distribution of the vulnerability indicator, as well as in the cluster combining low exposure and low sensitivity. Smaller municipalities, on the other hand, are more vulnerable to fuel price increases, particularly in terms of exposure. Municipalities with less than 5,000 inhabitants (accounting for 16.9% of the population) are notably overrepresented in the cluster that combines high exposure and high sensitivity. For context, only about one out of seven Italians (14.9%) live in large cities of more than 250,000 inhabitants (the least vulnerable), with most of the population (54.0%) residing in small- and medium-sized municipalities between 10,000 and 250,000 inhabitants.

Table 4 Distribution of Italian municipalities by size, vulnerability value and cluster. Note: version one of the composite indicator of vulnerability is used for this analysis
Full size table

Analysing the results of the vulnerability indicator at the national level helps to identify the areas that would be more impacted by a rise in fuel prices. In Italy, this largely mirrors some structural characteristics of the country and regional socio-economic divides. However, specific vulnerable situations may also be identified at the micro- or meso-level. For example, municipalities with an average value of the composite indicator relative to the national average may be still considered as disadvantaged if located in overall low-vulnerability region, and vice versa. These patterns can be overlooked when looking at national maps. Indeed, most published spatial analyses of vulnerability to fuel price increases (reviewed in Sect. 2) focus on specific metropolitan areas or city regions.

For these reasons, we conducted a more specific analysis focussed on two Italian metropolitan areas, highlighting relative hot and cold spots of the vulnerability indicator (version one). The selected case studies are the metropolitan areas around the cities of Milan and Naples as defined by Boffi and Palvarini (2011). The selection of these two areas follows two main criteria. First, they are respectively the first and the third Italian metropolitan areas in terms of the overall population (OECD 2022). This allows for considering complex and extensive urban systems. Second, they are located in different socio-economic contexts in the North-West (Milan) and in the South of the country (Naples). Figures 5 and 6 depict the presence of statistically significant clusters of high (hot spots in red) or low (cold spots in blue) values of vulnerability within the two urban areas. These clusters are defined using the Getis-Ord G* statistic (Getis and Ord 1992) and consider only the municipalities belonging to each of the two metropolitan areas. The maps also report the railway and motorway networks, as these tend to influence patterns of car use.

Fig. 5
A map of Milan indicates cold and hot spots with 99, 95, and 90% confidence, not significant, motorways, and railroads.

Relative hots spots and cold spots of vulnerability to fuel price increases in the metropolitan area of Milan (based on version one of the indicator)

Full size image
Fig. 6
A map of the metropolitan area of Naples indicates cold and hot spots with 99, 95, and 90% confidence, not significant, motorways, and railroads.

Relative hot spots and cold spots of vulnerability to fuel price increases in the metropolitan area of Naples (based on version one of the indicator)

Full size image

In the Milan metropolitan area, there is a large cold spot for vulnerability in the core of the urban area, comprising the municipality of Milan plus a large number of bordering municipalities to the North, stretching toward the North-East and around some primary rail lines. Some secondary cold spots emerge around other medium-large cities that represent secondary poles of the metropolitan area, such as Bergamo and Lecco. The most vulnerable situations are concentrated in an extended area across the provinces of Bergamo and Brescia. The maps for the constituent variables (not reported here for the sake of brevity) suggest that this is due to a combination of low income, high share of car use for commuting and long average length of car commutes.

The metropolitan area of Naples shows a similar cold spot of vulnerability in the municipality of Naples, although it does not extend much into the surrounding municipalities. Other cold spots are located along the Sorrentine Peninsula and around the city of Salerno (a sub-pole in the metropolitan area). Both are characterised by relatively better socio-economic conditions and lower levels of car ownership and use than the rest of the metropolitan area, and are well-served by railway. The main hot spot of vulnerability is located in the Northern part of the metropolitan area, at the border with the province of Caserta, namely in a cluster of municipalities around Casal di Principe. This is characterised by a concentration of low income, as well as by high levels of car ownership and use.

6 Discussion and Conclusions

Questions of vulnerability to motor fuel price increases are rather present in public and political debates, but are yet to catch the attention of energy poverty researchers and policymakers. This chapter has provided an overview of quantitative empirical approaches to assessing the affordability of car use with indicators, both from a static and dynamic perspective. We encourage researchers to build on this body of knowledge and conduct further empirical investigations on this important topic. This would improve our understanding of who is vulnerable to fuel price increases in Europe, raising the level of debates about this issue.

Our study illustrates how a composite spatial indicator of vulnerability can be built even for countries with limited data availability, like Italy. As such, the lack of ‘perfect’ data should be no excuse not to conduct similar studies in other countries. From all we know, Italy is likely to be one of the EU countries that are most vulnerable to fuel price increases, which highlights the relevance of our study.

From a methodological perspective, to the best of our knowledge, this is only the second study (after Liu and Kontou 2022) to conduct a robustness test of how variable aggregation, normalisation and weighting affect the patterns highlighted by the composite indicator of vulnerability to fuel price increases. Like them, we find that these methodological choices (e.g., in our case weighting) can lead to slightly different results, which must be considered when interpreting the findings. At the same time, the test results reassure us that our composite indicator is relatively robust to methodological choices concerning variable aggregation and normalisation.

In terms of substantive findings, a key message from our study is that the South of Italy is particularly vulnerable to fuel price increases. The fact that regions in Southern Italy have higher levels of motorisation than most EU regions with similar income levels suggests that this macro-region may well be one of the most vulnerable in the entire union.

At a closer look, however, our findings are more nuanced. We find that exposure and sensitivity to fuel price rises have a different geography within Italy, with the former being most severe in the Centre-North, and the latter in the South. The main factor behind higher vulnerability in the South (relative to the North) is lower income, which in many (but by no means all) municipalities is compensated to some extent by lower car ownership and use. The opposite happens in much of the North, where higher exposure through car ownership and use is mitigated in most places by higher income. In other words, high vulnerability in the South is largely driven by economic deprivation and as such, it is to some extent a manifestation of a long-standing economic divide within Italy.

Our analysis of the joint spatial distribution of exposure and sensitivity suggests that perhaps the areas of most concern are in the Centre of the country. In this in-between area, nearly half of municipalities combine high exposure and high sensitivity to fuel price increases. This is also the case in around a third of municipalities in the islands of Sicily and Sardinia. This is a novel finding, and one that warrants policy attention and further investigation.

When focussing on patterns of vulnerability within metropolitan areas, we find that exposure and sensitivity to fuel price increases tend to compound each other. The typical ‘urban socio-spatial configuration’ of Italian metropolitan areas (at least in the Centre-North) means that peripheral areas are both poorer and more car-dependent than the core. This pattern is consistent with Dodson and Sipe’s pioneering research on ‘oil vulnerability’ in Australia (2007). In Australian metropolitan areas, this is the product of a relatively recent and dramatic reversal of the earlier geographical patterns of income distribution, whereby poverty was shifted to suburban areas as a result of neoliberal reforms (Randolph and Tice 2014, 2017). For cities in the Centre-North of Italy, this configuration is rather a long-standing characteristic, which continued to this day (Kesteloot 2005). Still, trends towards a further ‘gentrification’ of inner cities and ‘suburbanisation of poverty’, as observed in many countries (Allen and Farber 2021; Baley and Minton 2018; Kneebone and Berube 2013; Lunke 2022), would make vulnerability to fuel price increases worse, even in Italian metropolitan areas.

It is interesting to compare our findings to Mattioli et al.’s (2019) study of England. In England, the different dimensions of vulnerability to fuel price increases tend to compensate each other within metropolitan areas (as periurban areas are more car-dependent but also more affluent than cities), but to compound each other at the interregional scale (as Northern city regions are both poorer and more car-dependent than Greater London). What we find for Italy is precisely the opposite pattern: within metropolitan areas, the factors tend to compound each other (at least in the Centre-North), while at the interregional scale, they compensate each other—as the North has more car-oriented travel patterns, but is more affluent and thus less sensitive to price increases.Footnote 5

We draw three implications from these comparisons. First, research on vulnerability to fuel price increases must carefully consider the interplay between the different dimensions of vulnerability, rather than just composite indicator scores. This helps to better understand what is causing the problem or what solutions might work in different places. Second, researchers must be attentive to both urban socio-spatial configurations and interregional inequalities, and how these vary between countries. Third, international comparison can provide useful insights into the causes of spatial patterns of vulnerability, which might not be apparent if one focusses on one country only. Research in this area would thus benefit from broadening the range of places where this type of analysis is conducted, as this could lead to a broader understanding of possible causes and policy responses.

With regard to policy implications, our analysis shows that fuel price increases, whether policy-induced or not, have a differential impact across the Italian territory. This must be taken into account when designing policy measures, be they aimed at increasing the cost of fuel for environmental reasons or at mitigating price rises due to market volatility or geopolitical events. Policymakers are also advised to consider the different factors behind vulnerability and their configuration, which can vary from place to place. Places where vulnerability is mainly the result of economic deprivation, such as the South of Italy, call for different interventions than places where vulnerability is mainly the result of excessive levels of car use. Measures aimed at improving adaptive capacity and reducing car dependence can be helpful in both types of area though.

An interesting question in this context is how the electrification of the vehicle fleet will affect vulnerability to fuel price increases. Liu and Kontou’s (2022) modelling study finds that the diffusion of electric vehicles would reduce both absolute levels of vulnerability and spatial disparities in vulnerability in Illinois (US). This happens because EVs reduce exposure to fuel price increases, i.e. the average share of income spent on running motor vehicles. Current trends in EV adoption, however, also have the potential to widen inequalities, particularly in the Italian case. First, high-income households, who are already less sensitive to fuel price increases, are the most likely to buy EVs (Wicki et al. 2022). Second, as of 2022, Italy has the second-lowest electric vehicle market share in Western Europe (Transport and Environment 2023). This is likely to widen the gap between Italy and other EU countries in terms of vulnerability to fuel price increases, by reducing other countries’ exposure more rapidly than Italy’s. Third, to date, the number of EVs per capita is much higher in the Centre-North of Italy as compared to the South (InsideEVs 2022). A continuation of this trend would reduce spatial inequalities between Italian regions in terms of exposure, but might increase them in terms of vulnerability, by widening the gap between the (already disadvantaged) South and the rest of the country.

With regard to future research, our analysis of the Italian case could be improved or built upon in three ways. First, an indicator of adaptive capacity to fuel price increases could be developed by leveraging publicly available spatial data on public transport departures or similar to generate accessibility metrics. This would be key to refine the composite indicator of vulnerability and might well bring to light different spatial patterns. We expect however that the inclusion of adaptive capacity would widen the gap between the North and the South of the country, as well as between urban and periurban and rural areas. Second, a more disaggregated analysis could be possible if Italian Census and income data were made available at the census tract level, or if it were possible to model them, as it is the case for England (Mattioli et al. 2019) and the US (Liu and Kontou, 2022). Finally, a replication of this study with more recent Census data (if and when they will become available) would help keeping track of how patterns of vulnerability to fuel price increases have evolved in a decade characterised by economic stagnation, fuel price fluctuations and further growth in motorisation.