The United Nation’s Sustainable Development Goals (SDGs) adopted in 2015 as a universal call to action to protect the planet and improve people’s living conditions placed urban traffic in the spotlight. In this context, the European Commission (EC) has set ambitious decarbonization goals to become climate-neutral by 2050. To achieve these goals, a set of measures have been taken at various levels (European directives and national, regional and local legislation) where the most effective and widely used tools are price signals like taxes, including congestion and parking charges, and subsidies. According to OECD (2019), fuel taxes (mainly on car fuels) are simple and cost-effective tools to limit climate change and health damage. Price signals affect the mode of transport used by agents, the car bought, the intensity of car use, and influence the hours of the day chosen to drive into certain urban areas, among others. Despite the effectiveness of taxes and subsidies, the redistributive effect of taxes and inequality issues regarding the generation of and exposition to traffic pollution pose interesting social issues that make these measures politically costly and increase the attractiveness of non-tax policy measures such as traffic restrictions in low emission zones (LEZ hereafter).

LEZ or environmental zones are urban areas with some kind of mobility restrictions on more polluting vehicles (Wolff and Perry 2010). These restrictions can be materialized both as a tax in the form of access charges (the case of London in the UK and Milan in Italy), and as access bans on some types of vehicles (as in Madrid or Lisbon) at certain times. More than 250 EU cities have already taken such measures and, whatever their political costs, have had a positive reduction effect on pollution (Transport and Environment 2019; Carslaw, et al. 2002; Holman, et al. 2015). The London Ultra LEZ has reduced traffic by 13,500 cars daily, cutting toxic air pollution by a third (London ULEZ report 2019Footnote 1). The Milan Area C,Footnote 2 which combines an access ban and a congestion charge, has achieved a nearly 30 percent reduction in cars entering the restricted zone and a  − 10% reduction in NOx. However, other authors as Boogaard et al. (2012) does not find significant decreases in traffic-related air pollution concentrations after LEZ implementation (forbidding ‘old’ trucks to enter the zone) in several Dutch cities. Some key aspectsFootnote 3 to guarantee the success of LEZ in reducing pollution are the extension of the territory covered, the degree and enforcement of the restrictions (clear and stringent restrictions with significant fines in the event of infringement), carefully designed exemptions and policy predictability (commitment of all political groups is needed). The spatial structure of cities also can play an important role in affecting emissions, as pointed by Burgalassi (2015).

This paper examines the effectiveness of the Madrid Central LEZ in reducing traffic in the short term. This analysis is important to the extent that, although the ultimate policy goal of LEZ is to reduce pollution, the instrumental tool used to achieve pollution reductions is traffic restrictions on the most pollutant vehicles. Previous studies on the effects of Madrid Central (Acción 2020) have estimated a 32% reduction in NO2 pollution in June 2019 vs. 2018. In the same way Lebrusán and Toutouh (2021) find that Madrid Central significantly reduces the NO2 concentration in the air and does not produce pollution displacement. However, these studies do not analyze how traffic intensity has been affected by the access restrictions imposed by Madrid Central. In order to evaluate the policy effectiveness of Madrid Central it is important to know whether traffic intensity has or has not been affected by the regulation. From a policy analysis perspective, a difference will obviously exist if the decrease in pollution is observed together with no changes in traffic intensity (more efficient and less pollutant vehicles but the same traffic congestion), or conversely, the drop in pollution is also accompanied by a decrease in the traffic intensity in the area affected by the restrictions (less traffic congestion and substitution of private cars by public transport). Changes in traffic congestion render different travel times/travel costs for both scenarios, which are key indicators in policy cost–benefit analysis (Eliasson 2009).

To measure the effect of Madrid Central restrictions on traffic intensity, special attention is paid to the origin–destination route followed for reaching the urban area affected by Madrid Central regulation. This information is combined with the location of traffic sensors to delimit areas where mean traffic intensity is measured in order to see whether any changes are observed after Madrid Central implementation. In this paper, the dissuasive effect of Madrid Central on urban traffic will be analyzed, not only in the Madrid Central zone but also in other urban zones using these zones like pseudo-control groups. One of the factors that limits the success of an LEZ is the response of agents: they can modify the usual routes they use to avoid restrictions, thereby increasing traffic congestion and travel times to those areas affected. Congestion could also increase in nearby zones due to parking search (Weinberger, et al. 2020) in areas that are not affected by access restrictions.

This response is more likely if the zone where restrictions are applied is relatively small, or if access is banned only in some hourly period. Previous works have shown that such rational responses cannot be neglected. For instance, Gibson and Carnovale (2015) concluded that drivers’ response to road pricing shows intertemporal substitution toward unpriced times and spatial substitution toward unpriced roads. A traffic analysis based on data from the traffic sensors network across the urban area of Madrid suggests that route substitution between zones after Madrid Central is plausible. In Sect. 4, we provide a detailed analysis of optimal routes to the Madrid Central zone to identify and locate the traffic sensors that measure potential dissuasive effects for accessing the LEZ. Sensor data are grouped into areas with Madrid Central and its bordering area as these are the zones with a higher number of sensors along the optimal routes. In Sect. 5 the effect of the restriction is analyzed, showing that while traffic intensity has decreased in the Madrid Central zone, it has increased in the bordering area. Previous studies on the effects of Madrid Central have not taken into account this substitution effect and the future design of mobility policy in the urban area of Madrid should address this undesirable outcome.

The Madrid-Central low emission zone

About six million people live in the region of Madrid (Deloitte—Ipd 2018), of which 49% are residents in the city of Madrid and another 44% reside in its metropolitan ring (the towns and cities surrounding Madrid). There are about three million motor vehicles in the region (Dirección General de Tráfico,, of which 92.4% are cars. As in other similar cities, the rate of motorization is higher in the periphery than in urban areas. In the municipality of Madrid there are 356 cars per thousand inhabitants, while in the metropolitan ring the motorization rate rises to 472 per thousand inhabitants.

Madrid Central is an LEZ that began operating on Friday, 30 November 2018, but was suspended in July 2020 by the Superior Court of Justice of Madrid due to "formal defects". It was finally declared null by the aforementioned Tribunal in May 2021. Aside from its short life span, this LEZ had a long transitional period where drivers who infringed the entry restrictions only began to be sanctioned after 15 March 2019. In practical terms, Madrid Central was only operational between March 2019 and July 2020, but following elections (May 2019), it was clear that the new municipal government would do its best to dismantle Madrid Central despite the opposition of ecological and left-wing parties and the EU itself. Finally, a new municipal regulation was approved in September 2021 (“Sustainable mobility Ordinance”/“Ordenanza de Movilidad Sostenible”) that implements a new LEZ denominated “Madrid 360” originally less restrictive than Madrid Central, but which implements progressively more restrictions within the original "Madrid Central" area and other areas of the Madrid urban area.

Madrid Central had an extension of 4.7 km2 and covered less than 1% of the municipality of Madrid. This area is considerably smaller than other European LEZs and was formed by grouping existing areas with some kind of restriction on non-resident car traffic and including others that already belonged to the current Madrid Central zone (Fig. 1).

Fig. 1
figure 1

Source: Own elaboration

Metropolitan area of Madrid and Madrid Central Low Emission Zone (lower graph).

In the Madrid Central LEZ, the most restrictive access applies to more polluting private cars, that is, vehicles that do not meet environmental standards (21% of all cars in the Madrid region in 2019, according to Dirección General de Tráfico, which comprises diesel-fueled vehicles registered before 2006 and gasoline-fueled vehicles registered before 2001. For other diesel and gasoline powered vehicles, the restrictions limit either the time or the reason for entering the zone, and drivers are obligated to leave their vehicles in a parking lot. In addition to vehicles belonging to residents, security and emergency services, and people with reduced mobility, the only vehicles that are not subject to entry restrictions are zero-emissions vehicles (less than 4% of all cars registered in Madrid). Taxis and similar services, as well as delivery vehicles, have been allowed transitional periods of adaptation with fewer restrictions than those applied to private passenger car traffic. Vehicles are also allowed to be parked in hotels and residents can give temporary authorization to non-residents. As a result, Central Madrid is a relatively small low emission zone implemented in areas with pre-existing entry restrictions characterized by a complex set of restrictions for private non-resident vehicles and based on sanctions (there is a 90-euro fine for entering the area when it is not allowed).

The evidence available points towards a clear reduction in pollutants after the implementation of Madrid. According to Acción (2020) Madrid Central lead to a 32% reduction in NO2 pollution in June 2019 vs. 2018. Also, in Lebrusán and Toutouh (2021) the authors conclude that Madrid Central significantly reduced the NO2 concentration in the air. In both works data are used from the only air pollution measurement station located within the Madrid Central perimeter (El Carmen station).

Measurements shows that the average NO2 recorded in 2019 was the lowest since 2000. This reduction was also observed in the average NO2 (µg/m3) pollution value in the month of November as shown in Fig. 2. The daytime noise levels recorded at this measurement station also point to a significant reduction in noise following the implementation of Madrid Central (Fig. 3), which is consistent with a reduction in traffic.

Fig. 2
figure 2

Source: Own elaboration based on Madrid City Council data.

NO2 monthly average level (µg/m3) recorded in November at the El Carmen measurement point (Madrid Central).

Fig. 3
figure 3

Source: Madrid City Council

Noise (dB) levels recorded in November at the El Carmen measurement point. Monthly average. 1: Average monthly daytime sound level ( 7:00 a.m.–7:00 p.m.). 2: Average monthly evening-nighttime sound level (7:00 p.m.–11:00 p.m.). 3: Average monthly nighttime sound level (11:00 p.m. –7:00 a.m.). 4: Average monthly daytime-nighttime sound level (24 h).

Despite these positive data, the results should be considered with caution as due to the absence of noise data prior to 2018 they are not conclusive. Moreover, the reduction in pollution was also observed prior to the Madrid Central restrictions, as is also pointed out by Lebrusan and Toutouh (2020). These authors also report an increase in SO2 levels and no significant variations in O3 concentration during the months post the Madrid Central implementation as compared to previous months, attributable, the authors explain to the influence of meteorological conditions. As observed in Fig. 2, from 2017 to 2018 there was a significant reduction in the level of NO2 pollution.

Pollution is highly influenced by meteorological factors, and although Madrid Central covers less than 1% of the total metropolitan area, emissions throughout the area can affect pollution records in distant areas which cannot be ignored. As pointed by Holman et al. (2015), to assess the impact on emissions of LEZs is needed to account of unusual meteorological conditions, or to the economic conditions that likely have affected the rate of replacement of vehicles and traffic volumes.

Another puzzling indicator is the registration of zero-emissions vehicles that are not affected by access restrictions to Madrid Central. According to figures of the Directorate General of Traffic (Dirección General de Tráfico, the number of zero-emissions vehicles in use in Madrid increased from 13,343 in 2018 to 21,762 in 2019 (+ 63.1%). This apparently huge increase is similar to the increase observed in the rest of Spain (+ 62.3%); hence no differential effect is observed in the case of Madrid.

For these reasons, instead of analyzing the evolution of pollution, we will analyze the evolution of local traffic intensity in Madrid, for which more information is available for a longer period of time. Data on traffic intensity are collected by more than 4000 traffic sensors distributed throughout the municipality of Madrid.

How traffic intensity is monitored in the city of Madrid

Traffic in Madrid is continuously monitored through a sensor network with 4,253 measurement points (2019 data) that are distributed among high-capacity roads, large roads and on main streets in the metropolitan area. The information captured by the traffic sensors allows determining the evolution of traffic in both its temporal and local dimensions. One limitation of these data is that they do not provide information about vehicle type, weight or emission category, or whether or not the vehicle has permission to circulate in Madrid Central (a camera network on the Madrid Central border continuously monitors access, but access to this data is restricted).

The traffic information recorded by the sensors is averaged over 15-min periods, and data on traffic intensity are provided (number of vehicles passing through that point per unit of time; i.e., number of vehicles per hour). The location of the measurement sensors in the Madrid metropolitan area together with details of the information provided can be found on the open data portal of the Madrid City Council ( Figure 4 shows the specific location of the traffic sensors and measurement points used in this study for November 2018. There have been very few changes from one year to another.

Fig. 4
figure 4

Source: Own elaboration based on Madrid City Council data (

Location of traffic sensors in Madrid in November 2018. (Right: Madrid Central and high capacity M-30 and M-40 ring roads).

This study only considers valid information recorded by the sensors on weekdays for the period 2013–2019. Madrid Central began operating on Friday, 30 November 2018, but drivers who infringed the entry restrictions began to be sanctioned only after 15 March 2019, that is, in the analyzed dataset Madrid Central was fully operative in the period, March16-December, 2019. It is expected that the effect, if any, of the enforcement of traffic restrictions on the most polluting vehicles will be clearer after comparing data before and after this measure is implemented. By taking several years, it is possible to check if there is a “jump” in the level of traffic in 2019 and to distinguish this jump from a traffic intensity trend (e.g., due to commercial activity).

Traffic intensity in Madrid exhibits a strongly intraday seasonal pattern with clear differences between working days and non-working days. As can be observed in Fig. 5, traffic intensity peaks on weekdays at 8:00 a.m. and 7:00 p.m. coinciding practically with the start and end of usual economic activity in the city. It should be noted that the most frequent school hours are from 9:00 a.m. to 1:00 p.m. and 3:00 p.m. to 8:00 p.m., while the most frequent working hours are from 7:00 a.m. to 8:00 p.m. but with a higher intensity from 7:00 a.m. to 3:00 p.m.

Fig. 5
figure 5

Source: Own elaboration based on Madrid City Council data (

Working day hourly mean profile of traffic intensity in Madrid (vehicles per hour). Data recorded by all sensors in Madrid urban area on working days. 2018.

The recorded local traffic intensity differs depending on the carrying capacity of the roads in the urban road network. Figure 6 shows the average local traffic intensity at different points and hours on weekdays in November 2018. The highest traffic intensity values in Madrid are especially concentrated on high-capacity roads and on the accesses to these roads and, to a much lesser extent, in the surroundings of Madrid Central. As can be seen, even before Madrid Central started operating, the area did not register especially high traffic intensity values, except at the end of the day (the area has a high concentration of leisure attractions). In the following section, travel to Madrid Central will be studied in depth in order to locate the most characteristic traffic points of these trips.

Fig. 6
figure 6

Source: Own elaboration based on Madrid City Council data (

Local hourly average traffic intensity for November 2018 (Madrid Central not operating yet). Blue line marks the Madrid Central border.

How traffic is affected by the Madrid Central LEZ

The configuration of the Madrid Central LEZ means that only non-residents who drive to the zone could be potentially affected by the prohibition to enter it, with the exception of taxi and delivery traffic since both are allowed access. As access to Madrid Central is restricted to a specific type of mobility (private vehicles entering the LEZ), it is important to have more information on the characteristics of these flows to understand the effects of this restriction. Transit traffic inside the LEZ can easily avoid the restriction by using alternative routes (remember the small extension of Madrid Central), so we will focus on car trips whose final destination is located within the Madrid Central zone. Because traffic sensors do not provide information about the origin and destination of movements, it is necessary to rely on other sources of information. Fortunately, in the case of Madrid, the recent Madrid Mobility SurveyFootnote 4 of 2018 (EDM18), in which 85,000 people living in the Madrid region were interviewed between February and May 2018 about their mobility characteristics, can be used for this purpose.

According to the EDM18 survey, on an average working day more than six million trips are made by private vehicle in the Madrid region, while nearly four million trips are made by public transport. The average duration of trips is 23 min in a private vehicle with a mean distance of 9.3 km. Trips to Madrid Central by non-residents in private vehicles (whose origin is detailed in Fig. 10) rose to 108,858 (1% of the total) with an average travel distance of 11.3 km (75% of trips are made from a distance of less than 14 km). Madrid Central is an important commercial and cultural attraction point with traffic flows from all points in the Madrid region, hence the greater average distance of travel. According to the EDM2018, only 1.95% of private cars owned by households that drive into Madrid Central are electric or hybrid. In the Madrid region this share is slightly lower at 1.89% (Fig. 7).

Fig. 7
figure 7

Origin of traffic flows to Madrid Central (blue area) by private car (blue circle = 14 km radius from Madrid Central centroid) Source: Own elaboration based on EDM2018 data

Information about the origin and destination of trips obtained from the EDM2018 survey allow determining optimal routes (smallest travel time path) for every combination of origins with Madrid Central as a destination using data from Openrouteservice API (OpenStreetMap contributors 2015). This information can be used to delimit the traffic sensors “activated” along these optimal routes, that is, the sensors that are most likely to record the deterrent effects on traffic through Madrid Central. This analysis is shown in the maps represented in Fig. 8. In addition to the obvious concentration in Madrid Central, the routes of vehicles to Madrid Central are concentrated on the main roads (M-30) and at the border of the Madrid Central zone. These results serve as the basis for the model proposed in the following section to test whether Madrid Central has had a dissuasive effect on traffic.

Fig. 8
figure 8

Source: Own elaboration with optimal routes and EDM2018 data

Optimal routes (smallest travel time) to Madrid Central and density of traffic sensors along the routes.


In order to isolate the effects of the Madrid Central zone on urban traffic, it is necessary to determine the differential effect on traffic intensity that can be attributed to the implementation of the LEZ. Rather than proposing a model to explain traffic intensity, a different approach will be followed. Traffic may be affected by variables linked to economic activity, population, availability of public transport and other factors. These variables can affect traffic intensity in the urban area of Madrid but with differentiated local effects. For example, a high population density in a given area will not only affect traffic intensity in that area but in all areas along the routes used by people commuting to work or going shopping. We assume that the overall traffic intensity (traffic recorded in the greater Madrid area) adequately captures common effects in all areas, thus allowing us to focus on differential local effects.

From the previous section we can conclude that the potential dissuasive effects of LEZ on traffic are more likely to be observed in certain areas. Taking into account the optimal routes used by non-residents who want to reach Madrid Central by private car, we are able to isolate the position of traffic sensors that measure this vehicle flow. Based on this information, we divide the metropolitan area of Madrid into four zones of interest. For these zones (see Fig. 9), aggregated traffic sensor data will be considered (hourly mean traffic intensity measured in each zone).

Fig. 9
figure 9figure 9

Source: Own elaboration

Proposed division of the greater Madrid area into zones and location of traffic sensors detail in each zone. (a) Zones (b) Traffic sensors by zone (excluding M30).

The P_Central (Madrid Central) zone collects all traffic data in the Madrid Central LEZ where the highest traffic impact is expected to be found. The P_border_central (border MC) zone summarizes the data from traffic sensors located in Madrid Central near surrounding areas. The surface area of this zone is 2.5 times larger than the surface area of Madrid Central (4.7 km2) with a maximum distance from the Madrid Central centroid of only 3.2 km. This zone is also expected to be affected by the LEZ regulation due to the high density of traffic sensors along the optimal routes. The P_in_M30 zone includes all the sensor data collected inside the area delimited by the high-capacity M-30 route, excluding the data grouped in other areas and also excluding the data recorded in the M-30 itself. All the remaining traffic information is included in the P_out_M30 zone.

The M-30 traffic sensors data are not considered in any zone due to the key role of this road in urban mobility as reflected in the high traffic intensity recorded along it (see Figs. 10 and 11). In the metropolitan area of Madrid, nearly all the routes between two points that are separated by a distance greater than 10 km pass through the M-30. However, the M-30 is also used for both regional and national transit transport, so local traffic restrictions are not expected to have a clear impact in traffic intensity along the M-30 route.

Fig. 10
figure 10

Traffic intensity by zone and month (pre-Madrid Central) and estimated linear trend (orange line). 2013–2019. Note: Hourly data for period July 2013-March,16, 2019. Linear trend for area i estimated with model \({I}_{i,t}={\beta }_{o,i}+{\beta }_{1,i}t+\sum_{j=2}^{12}{\beta }_{i,j}I\left(Month=j\right)+{\varepsilon }_{i,t}t=\mathrm{0,1},2,..\), where \({I}_{i,t}\) is the traffic intensity in area i at time t

Fig. 11
figure 11

Source: Own elaboration. Each graph has different y-scale

Mean Hourly traffic intensity by zone and period (pre-Madrid Central and post-Madrid Central). 2017–2019.

A first look at the evolution of traffic intensity shows that in the period 2013–2019 pre-Madrid Central a slight reduction in traffic intensity is observed in the urban area as a whole (Fig. 10). The same trend is also observed in the Out-M30 area, and, more clearly in the Madrid Central LEZ area. However, in the case of the Border Central area and In-M30 area, no significant trend is observed.

Figure 11 shows the evolution of traffic intensity by hours pre and post Madrid Central implementation in the period 2017–2019. The intraday pattern of road traffic intensity exhibits clear differences among zones and in their evolution pre and after MC. In Madrid Central (MC) and MC border area, the intensity of traffic does not show such large variations between 7:00 a.m. and 9:00 p.m. as occurs in the other two zones, which show a sharp decrease at 10:00 a.m. and an increasing intensity thereafter. Although the day-night activity cycle induces a high time correlation in traffic intensity between zones, the Moran’s I test for spatial correlation shows that the traffic intensity is not significant.Footnote 5

An analysis of the behavior of traffic intensity in the period 2017–19 (Fig. 11 bottom right graph) reveals a slight reduction in overall traffic intensity during normal hours of economic activity (7:00 a.m.–9:00 p.m.) after Madrid Central was implemented. The same evolution is observed for the Madrid Central LEZ but with an evident reduction in traffic during all business hours, and no evidence of intertemporal substitution of traffic intensity between hours, unlike that reported by Gibson and Carnovale (2015) for the case of Milan. The evolution in the other zones is not so clear, although a slight increase can be observed in the MC Border area.

To isolate the effect of the Madrid Central LEZ on traffic intensity, it is necessary to deal with the lack of a homogeneous evolution among zones and with the overall traffic intensity reduction. We will exploit the ability of panel data methodology to tackle the heterogeneity due to observed and unobserved factors in order to disentangle the contribution of reduction in overall traffic from the effect of the LEZ regulation. These local effects will be modeled using local dummy variables assuming that these effects are time invariant, and all time variation will be captured by year dummies (first version) or the overall traffic intensity (second version) (Table 1).

Table 1 Daily trips to the Madrid Central zone by means of transport and motive.

Let mean hourly traffic intensity recorded in zone i at time t be denoted as\({I}_{it}\). The “treatment” variable, Tit, is a dummy variable that takes the value of 1 when the traffic intensity measure corresponds to zone i and is recorded after Madrid Central is fully operational (March 16, 2019). The estimated coefficient associated to this variable will capture the differential effect of LEZ on each of the zones.

We propose the following alternative panel data specifications for traffic intensity in unit i at time t (hourly mean for working days in the period 2013–2019,

Version 1

$${I}_{i,t}=\mu +{\varvec{\tau}}{{\varvec{Y}}}_{{\varvec{t}}}+{\varvec{\pi}}{{\varvec{M}}}_{{\varvec{t}}}+{\beta }_{i}{T}_{it}+{c}_{i}+{{\varvec{\theta}}}_{{\varvec{i}}}{{\varvec{X}}}_{{\varvec{i}}{\varvec{t}}}+{\varepsilon }_{i,t}$$

Version 2

$${I}_{i,t}=\mu +{\varvec{\tau}}{{\varvec{I}}}_{{\varvec{t}}}+{\beta }_{i}{T}_{it}+{c}_{i}+{{\varvec{\theta}}}_{{\varvec{i}}}{{\varvec{X}}}_{{\varvec{i}}{\varvec{t}}}+{\varepsilon }_{i,t}$$

In version 1 of the model, the variables \({{\varvec{Y}}}_{{\varvec{t}}}\) are yearly dummy variables that take the value of 1 if the observation belongs to year y = (2014, 2019). The reference year (constant term) is 2013 (Table 2). \({M}_{t}\) are monthly dummies to capture seasonal variations in traffic intensity (December is the reference month). In this version, the yearly dummies have the same effect on the traffic intensity in each zone, and any potential break in trend due to the effect of Madrid Central will be captured by the variable Tit.

Table 2 Breakdonwn of sensors by zone

In version 2 of the model, instead of yearly dummy variables we include the variable \({{\varvec{I}}}_{{\varvec{t}}}\), global mean traffic intensity, as an explanatory variable common to all units to capture the effect of overall traffic conditions in Madrid driven by economic and other socio-demographic factors. The term \({c}_{i}\) captures unobserved time-invariant individual effects, \({{\varvec{\theta}}}_{{\varvec{i}}}{{\varvec{X}}}_{{\varvec{i}}{\varvec{t}}}\) aims to capture the differences in the hourly shape of traffic intensity by zones, where \({{\varvec{X}}}_{{\varvec{i}}{\varvec{t}}}\) denotes hourly time dummies (\({X}_{ih}=1\) if the observation belongs to zone i at hour h = 1,..,23 and zero otherwise) and \({\varepsilon }_{i,t}\) is an iid random component with an expected value equal to zero. The individual unobserved effect for each measurement unit will be estimated as a fixed term (with the Hausman test rejecting a random effect specification). Another specification assumes that each zone has its own idiosyncratic trend effect from overall traffic conditions, \({{\varvec{\tau}}}_{{\varvec{i}}}{{\varvec{I}}}_{{\varvec{t}}}\). The estimation resultsFootnote 6 of the different specifications are shown in Tables 3 and 4.

Table 3 Panel data estimation results. Model version 1
Table 4 Panel data estimation results. Model version 2

The estimation results show that all zones are affected by overall traffic conditions, \({{\varvec{I}}}_{{\varvec{t}}},\) as denoted by the positive estimated effect, with a higher intensity in the Madrid Central zone and its surrounding area. Interestingly, the results indicate that the LEZ has a negative effectFootnote 7 on traffic intensity in the Madrid Central area as expected. This effect is robust under different panel specifications, thus allowing us to conclude that the LEZ restrictions effectively reduce traffic.

According to the estimation obtained in model version 2 and the fixed effects specification, the estimated reduction is of 26.4 vehicles per hour on a working day in addition to an overall reduction in traffic intensity. Given that the mean traffic intensity in this zone at 12:00 p.m. (measured by the traffic sensors in the Madrid Central zone) is around 700 vehicles per hour, this amounts to a 3.8% reduction in traffic in a given hour on a working day. If the overall reduction in traffic is not taken into account, as in model version 1, the estimated reduction is slightly lower, amounting to 22.7 vehicles per hour (3.2% lower). The results also show an opposite effect in the surrounding area and we consistently observe an increase in traffic intensity in the area bordering Madrid Central. In the case of the In_M30 and Out_M30 areas, the estimated effects are not significant in the first specification, and in the alternative specification the results are not robust. The lack of a clear effect in these zones seems reasonable since the LEZ restrictions cannot be expected to affect the behavior of drivers who do not circulate in the restricted zone.

Figure 12 shows the estimated traffic intensity by months and zones using both versions of the model and indicating whether the Madrid Central LEZ was operational or not. As can be seen, the slight variations in post-MC traffic intensity produce a moderate acceleration in the downward trend of traffic in the LEZ area but not obvious changes in its border zone and in the other areas. To highlight these changes, in Fig. 13 average hourly traffic intensity is plotted using only 2018–2019 data (on March 16, 2019, Madrid Central was fully operative as infringers were only sanctioned after that date). If we focus on traffic variations solely during commercial hours, we clearly observe the asymmetric behavior mentioned above between Madrid Central LEZ and its surrounding area, mainly during central hours, and decreasing at the beginning and towards the end of business hours.

Fig. 12
figure 12figure 12

Source: Own elaboration A (c) FE model version 2 b

Estimated traffic intensity by zone and period. Monthly averages. 2013–2019. (a) FE model version 1. (b) FE model version 2 a.

Fig. 13
figure 13

Source: Own elaboration. Pre-MC: November 2018, post-MC: November 2019)

Estimated hourly Traffic intensity by zone during hours of commercial activity (7 h-22 h). November 2018 vs November 2019.(a) FE model version 1. (b) FE model version 2 a. (c) FE model version 2 b.

This behavior also shapes the relationship between the traffic intensity in the two areas, which is observable in a simple scatter plot of traffic intensity such as the one shown in Fig. 14. As can be seen, the slope of the relationship increases, indicating that traffic intensity in the border area for a given traffic intensity in Madrid Central was higher post MC. This behavior is coherent with a traffic transfer between zones.

Fig. 14
figure 14

Source: Own elaboration. Pre-MC: 2013–2019, March 15. Post-MC: Since 2019, March 16

Traffic intensity in the Madrid Central zone and its bordering area. 2013–2019.

If the Madrid Central LEZ truly serves to dissuade non-residents from driving into this zone, not only would the traffic sensors in that zone record less traffic but also the sensors on the pathway to Madrid Central as a final destination. As shown in Sect. 4, the traffic sensors along the optimal route to Madrid Central are located in Madrid Central as well as its bordering area. However, an increase rather than a decrease in traffic is observed in the border area. At this point, it is important to recall that Madrid Central is a small area covering just 4.7 km2. Seemingly, this should make drivers more likely to change their behavior; particularly when the mean distance of trips to Madrid Central in 2018 was 11.3 km, a longer distance than the overall mean trip distance of 9.3 km according to the EDM2018 survey.


Traffic management and mobility in cities is a matter of concern, not only in terms of their effect on citizens’ health, but also in terms of the optimal use of limited transport infrastructure and urban space. The intensive use of private vehicles generates non-negligible externalities that agents need to internalize in their behavior. Increased social awareness about pollution, noise and climate change is not enough to significantly change behaviors. Ambitious goals to combat pollution and climate warming should be supported in policies that discourage the use of private cars, notably old and most polluting vehicles. A congestion tax and LEZs are common tools used in European cities to limit car use. LEZs have also proven to be successful in reducing traffic and pollution.

Based on the positive experiences of other European cities, the city of Madrid implemented a LEZ in December 2018 which did not become fully operational (with sanctions for access violations) until March 2019. As in other Europeans cities, the Madrid LEZ has been seriously questioned, which has ultimately affected not only its size but also the type of vehicles affected by the restrictions and favored the adoption of long transitional periods.

Despite these limitations, the Madrid LEZ has been effective in reducing traffic in the affected area. The estimations presented in this paper show that hourly traffic intensity has decreased by 3.8% on a typical working day. When computed in annual terms, this figure represents an important effect and confirms the effectiveness of these measures in reducing traffic.

Unfortunately, a transfer of traffic to the surrounding areas has also been observed, probably due to the small area covered by ​​the LEZ. If the restrictions were fully effective in discouraging the use of private vehicles, a reduction in traffic should be perceivable in the entire access route to Central Madrid taken by non-residents using private cars. In the analysis of the delimitation of the traffic sensors located on the access routes to Central Madrid, it has been highlighted that both Madrid Central and its bordering area should record variations in traffic if the LEZ were effective in changing mobility behavior. However, the results show that while traffic in Madrid Central has decreased, it has increased in these bordering areas. This suggests that the measures put in place to dissuade car use have not been uniform, although this increase in traffic could be due to the transfer of traffic between the two areas given the small size of the LEZ.

This traffic transfer is a worrying externality of the LEZ implemented to reduce traffic and improve air quality in the area. When the quality of life of some residents is improved at the cost of worsening the quality of life of others, it is evident that the parameters of this policy measure must be reconsidered to ensure that the greatest possible number or residents will benefit from limiting car access.