When History Leaves a Mark: A New Measure of Roman Roads


This paper introduces a new measure of Roman roads that has been constructed for the Italian territory. The measure computes the length in kilometers of Roman roads at different administrative and territorial levels (NUTS, Local Labor Systems, grid cells), and contributes to the literature on historical infrastructures, providing a new precise measure to use for empirical purposes and easy to extend in all those territories where Roman roads have been constructed. From a mere econometric point of view, the index allows to capture the intensity of the treatment, providing an alternative empirical strategy for all those cases where the simple binary treatment cannot be performed.


The construction of roads was the task into which he (Gaius Gracchus) threw himself most enthusiastically, and he took great pains to ensure that these should be graceful and beautiful as well as useful. His roads were planned so as to run across the country in a straight line, part of the surface consisting of dressed stone and part of compacted gravel. Depressions were filled up, any watercourses or ravines which crossed the line of the road were bridged, and both sides of the road were levelled or embanked to the same height, so that the whole of the work presented a beautiful and symmetrical appearance Plutarch C.G.7

Historians and economists have argued how historical events influence economic development and how they have been crucial for better institutions and government attitudes. History has been questioned in terms of persistence.

Starting from the work of North (1981),Footnote 1 history has been found as having an important role in determining the current economic development, but it was during the first decade of two thousands with the contributions of Engerman and Sokoloff (1997, 2002), La Porta et al. (1997, 1998), and Acemoglu et al. (2001, 2002) that historical variables start to be identified as fundamental determinants of growth and current economic outcomes.

While literature has extensively recognized the legacy of history, during last 20 years a wide list of mechanisms of persistence has been proposed, confirming that there is not only one channel through which history acts. One of these mechanisms is represented by the historical transport infrastructure.

Recently, scholars started to show interest in old transport networks driven essentially by a twofold interest. On the one hand, understanding which is the persistence in time in the provision of the public infrastructure. On the other, unraveling the debated relationship between current transport systems and economic outcomes. In this framework, historical roads, railways or other means of transport are instrumental to avoid the endogeneity bias resulting from the potential reverse causality problem underlying this relationship.

In the last five years, the Roman road network has assumed a leading role within the branch of the ‘new economic history’ literature. The explanation has to be sought in three distinct reasons.

  1. (1)

    Importance: Roman roads represent the most considerable and tangible heritage left by the Roman Empire. The Roman domination touched three continents and almost 40 countries in the world. It lasted more than 700 years and its mark in terms of infrastructures, culture and society has been impressive. If history per se emphasizes the role played by the Roman Empire in explaining the European growth and the differences between western and eastern world development, economic research, devoted to investigating the long-term effects of the Roman Empire on modern economy, has been scarce and very recent. The positive legacy left by the Roman world nowadays has been nicely assessed by the contributions of Buringh et al. (2012), Bosker et al. (2013), Bosker and Buringh (2017) and by Michaels and Rauch (2018). However, it is only with Wahl (2017), Dalgaard et al. (2018) and De Benedictis et al. (2018) that the focus on Roman roads and the linkage between old infrastructure, new infrastructure, urbanization and current economic progress acquires importance in this literature.

  2. (2)

    Exogeneity: Roman roads have been primarily constructed for military reasons, this makes, from a mere econometric point of view, the Roman road network an ideal variable to be tested. On the one hand, it allows to directly assess the effect of history on modern economy. On the other, Roman roads represent an exogenous source of variation to identify the causal effect of other variables endogenous to historical socio-economic factors.

  3. (3)

    Data: in 2013 the Digital Atlas of Roman and Medieval Civilizations (DARMC) made freely available on the internet the geo-coded data (shape file) by McCormick et al. (2013) of the entire Roman road network measured at its peak of expansion. This allowed researchers to start investigating Roman roads and the heritage left by the Roman Empire.

In light of these reasons, this paper introduces a new measure of Roman roads for the Italian territory. The aim is, on the one hand, to discuss why the construction of a similar measure is not trivial for Italy. In Italy the Roman Empire has represented the most important domination of its history, the most durable and the most remarkable in terms of physical and intangible legacy. Starting from Rome, it entirely touched the Italian peninsula, and Roman roads were constructed across all conquered territories. In these terms the Roman Empire was a crucial domination for Italy due to its homogeneity and influential character. And the Roman transport infrastructure includes both characters: it was homogeneous across the entire Italian territory and it was influential since the military reason it was constructed for. On the other hand, this paper aims to present the measure of Roman roads expressed in kilometers and describes the methodology used to construct it. The measure has been computed for different administrative and territorial units (NUTS, Local Labor Systems and \(10 \times 10\) km grid cells) using Geographic Information Systems (GIS) methods. The detailed information provided makes the approach easy to perform for bigger or smaller territorial classifications, simple to extend in all countries crossed by the Roman road network or clear to replicate with transport infrastructure data in the form of linear shape file. Last but not least, this paper aims to highlight the importance of exploiting historical measures in the understanding of current economic outcomes.

If the literature aimed at evaluating the effect of history has mainly used a regression discontinuity design, distinguishing between a treated and a non-treated area, the need for a measure that accounts for the density rather than for the treatment allows to consider all those cases where a similar approach cannot be performed due to a spatial continuity of the historical event. The measure in kilometers of Roman roads responds to this need. The Roman domination has not been homogeneous neither across continents nor across countries. In some territories, like Great Britain or Germany, the Romans were able to establish their rule and construct their infrastructure only in one part of them; other countries, like Italy, Spain or France were, instead, totally conquered, and Roman roads and other civil infrastructure were uniformly built in the entire territory. In these areas the simple binary treatment cannot be applied and the Roman road measure allows to account for differences in history.

This paper consists of five sections. Section 2 examines the historical Roman road system, considering the main features related to the development of the network. Section 3 is the core of the paper and is divided in three parts. It first presents the data source of the Roman road network by McCormick et al. (2013). Then, it outlines the methodology used to construct the new measure of Roman roads. Lastly, it describes the novel data sets produced at the different administrative and territorial levels for the Italian territory. Section 4 analyzes the measure of Roman roads in the ‘new economic history’ literature framework, providing a correlation analysis with other variables. Concluding remarks are provided in Sect. 5.

The Roman Road Network

The Roman Empire has represented, in Italian and non-Italian history, one of the (say the) greatest empires of all time in terms of territory possessed and duration of political power. As highlighted by Laurence (1999), historians have recognized that the Roman state was involved in the development of an extensive transport network of roads starting from the fourth century bc.

The tactical purpose, the logistics for the war campaigns and the supply of the army across roads represented the spirit of the road system and of the whole Roman Empire. If the military aim was at the core of the road network, other three main aspects were strictly related to the constitution of an intricate road system: (i) the development of a great empire, (ii) the advanced engineering abilities of the Romans, (iii) the other ways of transport and trade.

Roman roads were instrumental to the war campaigns. Roth (1999) underlines that the Romans built their roads primarily for military reasons; the commercial travel was only an indirect advantage of the road network. The majority of the weight of the Roman army supplies was represented by three elements: food, fodder and firewood, and all military decisions were determined by the need to assure the provision of supplies to the army. Accordingly, Thompson (1997) explains that the construction of a road network originated from the need to ensure that a large number of horses, cattle, carts and infantry could circulate: primordial non-Roman routes were problematic during wet and rainy days, since the deep mud impeded or delayed the movement of goods and services.Footnote 2 As clearly explained by Roth (1999), the Romans were aware that moving supplies to the army by ship was cheaper and faster than by land, but transport by sea was dangerous and expensive: seafaring in the Mediterranean was limited between March and November, and it was really safe only during the summer months (from June to September).Footnote 3 Land transport, instead, had no limitations and was practicable all year round. The Romans knew that the need for a logistical infrastructure was fundamental for the movement of armies and supplies and for the enlargement of the empire, hence expansion, maintenance and repairing of the road network were continuously performed.

Historians have argued how the design of an intricate road system and the development of a great empire were strongly correlated. On this point, Thompson (1997) argues that the vast and comprehensive Roman road system changed the entire empire. Accordingly, Gleason (2013) stresses that the enlargement of the empire was possible thanks to a developed road system. The Roman army was too small to conquer the enemies of Rome, but the constitution of a vast empire was possible by investing in the construction of a complex road network rather than enlarging the infantry forces. Knapton (1996) underlines that the new conquered regions, on the one hand, enlarged the Roman Empire contributing to its power, authority and wealth; on the other hand, the payments of conquered territories were mainly used for the public infrastructures, like roads and aqueducts. This led to the development of the Romans’ engineering capability.Footnote 4

The engineering behind the construction of roads subtend incredible and high-level skills. The Romans were mainly focused on getting the road straight to make the structure and shape of the network as simple as possible. To achieve the straight configuration, they defined points that could be quickly connected by a straight line (Davies 1998). Gleason (2013) explains that to mark the road’s path with either stakes or furrows, creating linear roads, was the first purpose of the Romans. Legionaries and slaves belonging to the army were involved in the road construction process.Footnote 5

Road transport can be seen as a complementary system to river and maritime transport.Footnote 6 As described by Snedden (1998), goods were constantly transported throughout the Roman Empire and despite the risks, dangers and problems that characterized this mean of transport, the most efficient way to move goods was by sea. Ships were preferred to roads since they could transport large amounts of goods and people in a shorter time. Six hundred passengers or six thousand amphorae of wine, oil and other products were highly traded using sea transports. The Romans put much effort in improving the effectiveness of shipping, developing harbors and lighthouses (Snedden 1998).Footnote 7 For commercial goods, rivers were also used and preferred to roads, and the same principle applied for the movement of military supplies.Footnote 8 However, despite the apparently overwhelming economic advantage of trade by water, Pawson (1977) pointed to the key advantages of land transport, arguing that the land transport system could be classified in two parts: a complementary system, which was interdependent with water transport and performed a feeder and distribution role for it, and a competitive, independent system which did not rely on water transport linkages.

Data on Roman Roads

The Original Data

The original data on the Roman road network are geo-coded data constructed by McCormick et al. (2013) and released in the form of a linear shape file.Footnote 9 The layer has been obtained digitizing the information in Talbert (2000).Footnote 10 This makes the data accurate and detailed and allows to perform, by using Geographic Information Systems (GIS) approaches, reliable spatial and mapping analyses for the Roman world.

The shape file consists in 7154 segments of ancient Roman roads existing at the peak of the empire, corresponding with the death of Trajan (117 A.D.). The network covers 36 countriesFootnote 11 over three continents (Europe, Africa and Asia). The peculiarity of the shape file is represented by the way according to which roads are recorded: it does not classify single and complete roads, but it provides a list of segments which compose roads.Footnote 12 For each segment four different types of information are provided: its position in space, its length in meters, its sizeFootnote 13 and its certainty.Footnote 14

Size and certainty of the road are the two main features evaluated in the shape file. Figures 1 and 2 provide a representation of these two traits, classifying roads between major and minor and between certain and uncertain, respectively.

Fig. 1

Roman roads by size: major and minor roads. Source: Author’s elaboration from McCormick, M. et al. 2013. “Roman Road Network (version 2008),” DARMC Scholarly Data Series 2013-5

Fig. 2

Roman roads by certainty: certain and uncertain roads. Source: Author’s elaboration from McCormick, M. et al. 2013. “Roman Road Network (version 2008),” DARMC Scholarly Data Series 2013-5

Figure 2 shows that there is a reasonable balance between certain (green lines) and uncertain roads (orange lines).Footnote 15 This is less the case when considering the size of the road: Fig. 1 shows a higher presence of blue lines (minor roads) than red ones (major roads). Moreover, when drawing only major roads, it is immediately possible to distinguish present nations’ borders and to recognize the outline of the Mediterranean Sea, since a high number of roads have been constructed along the coasts.

The total Roman road network is almost two hundred thousand kilometers long.Footnote 16 The length of each segment of the Roman road network originates from the typology of the data set (geo-coded data) and provides an important information about the extent and the coverage of the whole Roman infrastructure. However, a similar information for single geographical territories, like countries, regions or cities, is not included in the shape file and ad hoc computations need to be performed in order to obtain the length of the Roman road infrastructure at different territorial levels.

Methodology: How to Compute the Roman Road Measure

Beyond being interesting per se, the information included in the shape file by McCormick et al. (2013) is appealing for the promising elaborations it allows. GIS approaches and geometric tools are the techniques researchers need to use to extract new information from the starting shape file and to generate new data devised for specific investigations.

This section analyzes all those technical aspects linked to the calculation of the measure of Roman roads in kilometers.Footnote 17

The measure is computed for Italy and relying on different administrative classifications and territorial units: NUTS, LLSs (Local Labor Systems), and grid cells.Footnote 18 Italy represents an ideal laboratory to implement these methodologies since the Italian territory was completely touched by the Roman domination and crossed by the Roman road network. The choice of exploiting the NUTS classification allows to have an index that can be used for several socio-economic analyses and combined with different administrative data. Moreover, territorial units reflect not only the administrative organization of the country, but also the institution functioning, the culture and the historical past of the territory, allowing more detailed investigations.Footnote 19 Considering the disaggregation in LLSs,Footnote 20 instead, allows to take into account territories according to firms’ activities and workers’ commuting, and to report the information on Roman roads to a dimension that is particularly used in empirical Italian studies. Conversely, the use of a gridFootnote 21 allows to get rid of the political division of the territory and to concentrate only on an equal and completely geographical dimension. This is particularly suitable since it can be combined with raster data, that are extensively used in recent economic literature, provide information about the terrain and allow spatial analyses.

All elaborations have been performed using QGIS, a free open-source GIS software.Footnote 22

Two starting layers are needed to compute the Roman road measure: the linear shape file of the Roman road network by McCormick et al. (2013) and the polygonal shape file of the administrative borders provided by Istat.Footnote 23 Both shape files need to be projected using the same spatial reference. The selection of the appropriate coordinate reference system (CRS) should consider the spatial reference system in use in the country or in the specific area the analysis looks at. Istat makes available the shape files of the Italian administrative borders for the main reference system of Italy: UTM WGS 84 Zone 32N. This reference system is employed to homogenize the small and middle scale cartography at the European level.

The first step consists in obtaining the Italian linear layer of the Roman road network. As shown on the left hand side of Fig. 3, the complete linear layer by McCormick et al. (2013) needs to be superimposed on the polygonal one with the Italian borders, and the latter is used as a mold to derive the ramification of Roman roads for the sole Italian territory.Footnote 24 In this way, segments that are in common between Italy and adjacent countries are split, and only the parts of the road included within the Italian territory are ascribed to Italy. The right hand side of Fig. 3 shows the result of this procedure: the linear layer of Roman road network for the sole Italian territory, at the NUTS 0 level.

Fig. 3

Italian Roman road layer at the NUTS 0 level. Source: Author’s elaboration from McCormick, M. et al. 2013. “Roman Road Network (version 2008),” DARMC Scholarly Data Series 2013-5 and from Istat (2011) data

Figure 4 displays the same procedure at the NUTS 3 level, but, instead of using the complete layer by McCormick et al. (2013), the new linear shape file of Roman roads at the NUTS 0 level has been superimposed on the polygonal one with the Italian NUTS 3 borders.Footnote 25 As shown in the left hand side of Fig. 4, the provincial decomposition, provided by the polygonal shape file, acts as a cutter for the segments of roads that are in common between adjacent provinces, and only the parts of roads within the boundaries of the selected province are assigned to that territorial unit. The right hand side of Fig. 4 shows the result of the ‘Intersect’ tool: the shape file for the sole Italian territory, with a distinction of segments by Italian province.

Fig. 4

Italian Roman road layer at the NUTS 3 level. Source: Author’s elaboration from McCormick, M. et al. 2013. “Roman Road Network (version 2008),” DARMC Scholarly Data Series 2013-5 and from Istat (2011) data

The same procedure should be followed to obtain the Roman road shape file by NUTS 2 and NUTS 4, and by LLSs.Footnote 26 At the grid level, instead, the Italian territory is first divided in cells of equal dimensions, as shown in Fig. 9 in the Appendix, and then using the vector grid as a cutter.

The second step consists in computing the Roman road measure in kilometers: it requires the use of a specific geometry tool that measures the length of each segment.Footnote 27 Starting from the linear shape file of Roman roads, a geometry column is added, and this new field computes the extent of each segment in meters.

Last step consists in deriving a single measure from the data by segment. The data set by segment lists all segments of Roman roads for each territorial unit, with three main information: size, certainty, length of the segment. The information about the length is divided by 1000 to express it in kilometers. To obtain a single measure data set, data by segment need to be exported to a data analysis software and the information needs to be collapsed at the territorial level. In this way, rather than having many entries (i.e. segments) for each territorial unit, the new data set has one entry (i.e. measure) for each territorial unit. The single measure can be computed overall, in order to have the index in kilometers of all Roman roads, or can be calculated according to the two traits included in the starting data set (size and certainty). This will produce different measures of Roman roads: all, major, minor, certain, uncertain, both major and certain and so on.

A further step consists in computing the measure of Roman roads in density, weighting the index by the area of the territorial unit.Footnote 28

The New Data

The new data refer to Italy and provide two pieces of information: by segment and total. Data by segment are geo-coded data with the detail about the size, the certainty and the length in meters and kilometers of each segment of Roman road. Overall data, instead, consist into six different measures: (i) kilometers of all Roman roads;Footnote 29 (ii) kilometers of major Roman roads; (iii) kilometers of minor Roman roads; (iv) kilometers of certain Roman roads; (v) kilometers of uncertain Roman roads; (vi) kilometers of both major and certain Roman roads. All measures have also been weighted by the land area, producing six further indexes that account for the density of the Roman road network.

Both total and by segment data have been created for LLSsFootnote 30, for \(10 \times 10\) km grid cells,Footnote 31 and for five NUTS levels: (i) country (NUTS 0); (ii) macro-regions (NUTS 1);Footnote 32 (iii) regions (NUTS 2);Footnote 33 (iv) provinces (NUTS 3);Footnote 34 (v) municipalities (NUTS 4).Footnote 35

Italy includes 10% of the entire network: 19,593 out of 192,861 km. 1817 out of 7154 segments when referring to sections of ancient Roman roads. Focusing this descriptive analysis on the NUTS classification, Table 1 allows to compare in terms of segments the starting data by McCormick et al. (2013) with the new five data sets. 44% of all segments in the starting data set are major roads, in Italy they represent 35% of the total Italian segments. However, at the most disaggregated level, major roads reach 48% of all roads. Almost half of the segments are evaluated as certain in the Italian layer, 44% in the complete one. These percentages are confirmed at the different NUTS levels. When considering both features (size and certainty), the majority of segments are minor and uncertain in all shape files.

Table 1 The Roman road network description: original and new data

If Table 1 takes the point of view of the original structure of the data (by segment), Table 2 allows to appreciate the existence of Roman roads by administrative unit. In terms of NUTS 1 and NUTS 2, Roman roads cross all macro-regions and regions in Italy. When focusing on Italian provinces, Table 2 shows that 108 out of 110 provinces have Roman roads;Footnote 36 less than 3000 out of 8092 municipalities when taking the NUTS 4 point of view. At the NUTS 2 and NUTS 3 level, the majority of regions and provinces have both major and minor roads and both certain and uncertain roads. At the municipality level, instead, the majority of territorial units are crossed only by major Roman roads, when taking into account the size of the road, or by uncertain roads, if considering the certainty trait.

Table 2 Roman roads by Italian NUTS

Figures 5, 6 and 7 concentrate on the main piece of information produced: the measure of Roman roads in kilometers. The measure, computed for each administrative unit in Italy, and at different NUTS levels, describes the presence of the Roman road network in the Italian territory. The double version of the measure (in absolute terms and in density) provides two different information about geography. The simple index in kilometers allows to appreciate the extent of the network and is potentially able to say something about the nature of the territory. As observed by Ramcharan (2009), rougher territories have less kilometers of roads. However, from a different point of view, more arduous surfaces need more roads to reach those lands.Footnote 37 The simple measure does not control for the size of the land and this does not allow to distinguish wider areas from smaller ones. The measure in kilometers per 100 square kilometers of land area fills this gap and gets rid of the surface of the area, by measuring the density of the Roman road network in each territory. However, it does not control for the morphology of the terrain. In this regard, defining which measure better captures the presence of Roman roads in a given territory is not trivial.

Fig. 5

All Roman roads in length by NUTS level: equal intervals. Source: Author’s elaboration from own data and from Istat (2011) data

Fig. 6

All Roman roads in length by NUTS level: quantiles. Source: Author’s elaboration from own data and from Istat (2011) data

Both Figs. 5 and 6 look at the simple measure in kilometers. Figure 5 shows the presence of all Roman roads in the Italian peninsula for four different NUTS levels, by dividing the distribution according to equal intervals of the measure. Figure 6, instead, divides the distribution in quantiles. This provides two different pictures: Fig. 5 allows to appreciate which territories are endowed with more kilometers of Roman roads, Fig. 6 ranks the territories. From both figures emerges that the distribution of Roman roads within Italy is not so informative when considering the NUTS 1 or the NUTS 2 level, since it is strictly linked to the size of the administrative units. At the NUTS 3 and NUTS 4 level, instead, the surface of the land across provinces and municipalities is more homogeneous and some more conclusions can be inferred. Figure 5 clearly shows that the province and the city of Rome with 2498 and 915 km, respectively, are, unsurprisingly, the most endowed with Roman roads; no other NUTS 3 unit or municipality reach similar extensions, highlighting a large gap between where the Roman Empire had its core and where it enlarged. However, this gap slightly reduces when considering only the consular roads.Footnote 38 As shown in Fig. 12 in the Appendix, there is a higher balance across provinces in terms of kilometers of major roads, revealing that military roads, constructed by the Romans to conquer new territories, are more homogeneously distributed in the peninsula.Footnote 39 When dividing the distribution of all Roman roads in quantiles, a weak spatial distribution across Italy can be observed. Figure 6 displays darker areas in the southern territories, informing how the Romans devoted more efforts in building their roads in the South rather than in the North. A possible explanation of this fact can be found in the way the Roman Empire enlarged: firstly towards the southern regions, then to the North. This picture is confirmed by the “numerical” information provided in Table 6 in the Appendix. Table 6 ranks Italian provinces according to the length in kilometers, considering all, major and certain Roman roads. The first place is occupied by the province of Rome, which owns, in absolute terms, the largest extent of roads; even if we consider only major or only certain roads, Rome keeps its leading position. Provinces of Latina (close to Rome) and Foggia (in Apulia, South-East) are in second and third position, respectively, in terms of total Roman roads; they switch places when looking at certain roads. Number two and three in terms of major roads are, respectively, Potenza and Cosenza in southern Italy.

Fig. 7

All Roman roads in density by NUTS level: quantiles. Source: Author’s elaboration from own data and from Istat (2011) data

Figure 7 concentrates on the measure in density. The maps of Italy at the different NUTS levels confirm the picture emerging from Fig. 6: a denser Roman road network in the southern regions and in the surroundings of Rome, that holds the character of being at the core of the main street junctions. This result is confirmed by ranking the NUTS 3 Italian provinces according to the kilometers of Roman roads per 100 square kilometers of the land area. Table 7 in the Appendix shows, again, that Rome stands out against the other provinces. Latina (in the Center) and Trieste (in the North-East) follow in second and third place. However, when considering the density of the consular roads, as in Fig. 10, there are no differences between northern and southern regions. Since they served for the conquest of new territories and the constitution of a united empire, the distribution of major roads is more homogeneous. When looking at the density of certain roads, disregarding the size, as in Fig. 11 in the Appendix, the balance between northern and southern territories is a bit less evident: certain Roman roads are denser in the areas where the main trajectories have been constructed.

The descriptive analysis highlights how the NUTS 3 level is the best territorial disaggregation to appreciate the branching of the Roman road network in Italy and how both absolute and relative measure of Roman roads in kilometers are useful in understanding the intensity of the road infrastructure.

The Roman Road Measure and the ‘New Economic History’ Research Agenda

The Roman road measure places itself in that vast strand of economic literature named ‘new economic history’. Born in 1997, this literature highlights the important role of history in the current economic development.

The first works were completely focused on one subject: colonialism of Americas. However, after some years the geographical focus changed and contributions started to look at other historical facts and other countries or continents. Also the mechanism of persistence changed. The earlier works explained the legacy of history via the persistence of institutions. More recent articles, instead, started to take into account different historical shocks, giving innovative insights on the channels through which history works.

Recently the role played by historical infrastructures and transport systems, like roads, railways, ports, attracted the curiosity of researchers. The interest by the literature has been driven by the effect that historical great transportation infrastructure projects have in reducing trade costs, playing a positive impact on productivity and increasing the level of the real income in trading regions. Indeed, public investments in transportation infrastructure and their effects on income, exports and education have been largely debated since the earliest projects in ancient times. On the one hand, the experience of historical big transportation constructions represents a helpful and practical evidence of what have been the favorable results of old transport projects. On the other, they represent an ideal tool to quantify the effect of new transportation infrastructures.

The focus on the Roman road network is recent. The first contributions on Romanness were mainly devoted to investigating the persistent effect of the Roman domination. The contributions by Buringh et al. (2012), Bosker et al. (2013), Bosker and Buringh (2017) and by Michaels and Rauch (2018) focus on the heritage of the Roman Empire that comes out in the urbanization of future periods.Footnote 40 And the important role of roads emerges in facilitating settlements and in creating urbanized centers.Footnote 41 In the contributions by Wahl (2017), Dalgaard et al. (2018) and De Benedictis et al. (2018), instead, the focus is more on Roman roads and whether they affect the modern economy by their persistent effects.

In the work by De Benedictis et al. (2018) the measure of Roman roads is tested as a strong and valid instrument for the current transport infrastructure: kilometers of Roman roads are correlated with both kilometers of railways and motorways, and, through their effect on the modern transport network, they affect Italian trade costs. In this perspective, the study can be seen as a proof of the reliability and functioning of the Roman road index. However, if the empirical validity of the measure has been tested, less is known about the interaction of Roman roads with other socio-economic factors. This section aims at discussing the index in the context of the ‘new economic history’ research.

Practically, over the past two decades, the research moved from exploiting simple economic performance measures and standard econometric methods to new proxies of development and sophisticated econometric techniques. Moreover, the cross-country perspective has been recently supported by a within-country view that benefits from new information coming from several fields, like history and anthropology (Michalopoulos and Papaioannou 2018). In order to gain more insights about how the Roman road measure works, this section provides a simple correlation analysis between the index and some variables at the Italian local level.

Focusing on the right hand side of a regression model, Roman roads are examined relating to five plain economic outcome measures: value added,Footnote 42 exports, nightlight luminosity, population and population density. The dimension considered is the NUTS 3 level;Footnote 43 however, nightlight luminosity is also analyzed at the grid cell level, the typical dimension examined in the economic contributions on persistence of history. Data on value added, exports and population are provided by Istat and refer to 2016.Footnote 44 Data on nighttime luminosity, instead, refer to 2013 and are satellite images sourced from the Defense Meteorological Satellite Program (DMSP) of the US Department of Defense and made available by the NOAA’s National Geophysical Data Center.

Table 3 Roman roads and economic outcomes: correlations
Fig. 8

Relationship between Roman roads (all, major, certain) and economic outcomes (value added, exports, population, nighttime luminosity). Source: Author’s elaboration from own data, from Istat (2011) data and from Version 4 DMSP-OLS Nighttime Lights Time Series data

Figure 13 included in the Appendix maps all potential dependent variables along the provincial dimension, and the nighttime luminosity also at the grid cell level. All six maps show higher values mostly concentrated in the Po Valley in the North and along some coastal zones in the South-Center, where the largest urban agglomerations are located. Even if interesting per se, Fig. 13 provides a further piece of information when compared with Figs. 6 and 7, both showing the distribution of all Roman roads in absolute kilometers and in kilometers relative to the land area, respectively. As already discussed, both Figs. 6 and 7 reveal a relatively higher presence of Roman roads in the southern provinces. And compared with the picture emerging from Fig. 13, it emerges a sort of reversal of fortune condition: the Roman road network is denser in relatively less developed areas. Along with these simple descriptive maps, Table 3 and Fig. 8 provide a more detailed analysis on the relationship between Roman roads and economic outcomes. Correlation indexes in Table 3 show that there is a strong significant positive relationship between the value added and the different measures of Roman roads (all, major, certain). These strong correlations are confirmed when considering the total population. Total exports and nighttime luminosity at the NUTS 3 level, instead, do not show any correlation with the Roman road index. However, when looking at the nightlight luminosity at the grid cell dimension, the positive relationship between the nighttime light intensity and the different measures of Roman roads is confirmed, although the size is more moderate. Similarly, this reasonable positive correlation is verified when considering measures of Roman roads and of total population in density. All these positive correlations can observed in Fig. 8.Footnote 45

Moving forward, and discussing about the mechanism behind the long-run effects of history, the measure of Roman roads is analyzed in the light of the potential channels of persistence. Early research has highlighted the role played by culture in the Italian territory. Guiso et al. (2016) find that municipalities in northern Italy, that were free city-states during the Middle Ages, have higher levels of civic capital today, and that the mechanism of persistence should be found in education and socialization. This validates the thesis by Putnam et al. (1993) according to whom the persistence of the Italian economic development and its differences across regions should be attributed to culture.

By relating the measure of Roman roads with what Guiso et al. (2011) name ‘civic capital’,Footnote 46 the resultsFootnote 47 show non-significant correlations at the NUTS 2 level, suggesting that the cultural channel is a mechanism that is less related with the Roman infrastructure. Indeed, as clearly underlined by all contributions on Roman roads, the Roman road network finds in the transport infrastructure the connection between past and present economic development. Table 4 exactly proves this reasoning, showing positively significant correlations between the measure of Roman roads and kilometers of motorways and railways.Footnote 48 The high values highlight, from a mere econometric point of view, the suitability of the Roman road index as an instrument for the current infrastructure since it fulfills the relevance condition. On the other hand, it proves that the existence of physical constructions was fundamental for predicting new infrastructures: new roads were more likely constructed on the top of or along ancient roads. Moreover, as clearly explained by Dalgaard et al. (2018), the persistence of the Roman road network on prosperity today is unlikely driven by cultural factors.

Table 4 Roman roads and current infrastructure: correlations
Table 5 Roman roads and geography: correlations

Finally, referring to the debate in the ‘new economic history’ literature between ‘only history matters’ vs ‘only geography matters’, the Roman road measure is analyzed relating to two geographic indexes: percentage of mountainous territory and elevation.Footnote 49 Scholars supporting the ‘only history matters’ view argue that geography affects development and growth only indirectly through its effect on institutions. The ‘pro-geography’ economists, instead, find in geographic factors a key and direct explanation for having long-run effects on economy. Table 5 reveals that the construction of Roman roads has been moderately driven by the geography of the territory. Correlation indexes are significant and with the expected negative sign only when referring to the density of the network. Moreover, elevation assumes a role only for major Roman roads. Major or consular Roman roads were the largest and most important roads constructed by the Romans for the military campaigns. For their role and relevance they can be compared to nowadays motorways. Therefore the higher correlations between geography and major Roman roads reflect what the literature on modern roads suggests: rougher territories have less kilometers of roads (Ramcharan 2009). Since the role played by geography in the construction and development of the Roman road network was relatively negligible, this simple correlation analysis seems to validate the ‘pro-history’ perspective. It should be recalled that the Romans were expert engineers able to overcome physical and natural obstacles with robust infrastructures like bridges, tunnels, canals and, of course, roads.

Concluding Remarks

In the late nineties-early two thousands the economic literature has grown interest in documenting the links between historical events and economic outcomes today. Recently, the historical transport infrastructure has been observed as one of the mechanisms that explains the legacy of history on modern economic results.

If economic works on historical transport systems multiplied in recent years, the release of the geo-coded data on the Roman road network by McCormick et al. (2013) allowed scholars to focus on the persistent effect of Roman roads.

The Roman road network represents, in the view of historians and common people, one of the main traits of the Roman domination and the major sign of the glorious past of the Roman Empire. Most of the economic research on the Roman world emphasizes how development and urbanization took advantage from the Roman domination, whereas less attention has been devoted to the effect of the Roman road network. In last two years three works have tried to fill this gap. Wahl (2017) considers the case of Germany, De Benedictis et al. (2018) look at Italy, and Dalgaard et al. (2018) adopt a more broad perspective. However, there is still room for research.

Presenting and describing a new measure of Roman roads in kilometers was at the core of this paper. The measure tracks the extension of the network in Italy and has been computed for different dimensions. On the one hand, the two main traits included in the starting geo-coded data have been used to compute the kilometers of Roman roads according to the size and the certainty of the road. On the other, the NUTS classification, the grid cell dimension and the LLS structures have been exploited to create measures for different territorial levels.

The last part of the paper was devoted to observing the Roman road measure in the context of the ‘new economic history’ literature. The index has been related to different measures of economic outcomes, highlighting that Roman roads are positively correlated with higher level of economic development today. The high correlation with the current transport infrastructure and the non-significant correlation with proxies of social capital reveals, instead, how the mechanism of transmission should be found in the physical structure of the Roman road network that enabled the construction of the new infrastructure on the top of ancient roads.

To conclude, the new measure of Roman roads produces three main contributions. First, it adds a novel index for the Italian territory to the existing group of old historical infrastructure measures. Second, it contributes to the literature on the persistent effect of history, allowing new empirical studies and applications. Third, it provides a method to generate additional measures of the Roman infrastructure for other countries and other territorial disaggregations.


  1. 1.

    An earliest proof of the long-lasting effect of history can be found in the work of Fogel (1964), who analyzes the impact of railways on the American economic development in the nineteenth century. Subsequent to the work of North (1981), a contribution by North (1990) and two papers by Greif (1993, 1994) can be considered the most influential contributions of the nineties that precede the birth of the ‘new economic history’ literature. They analyze the long-term relationship between international trade, growth and changing institutions.

  2. 2.

    The construction of a network of paved roads empowered not only the transport of goods and services, enabling the movement of larger quantities and people and making transfers easier, but armies were able to travel 25 miles a day, even in bad weather conditions (Thompson 1997).

  3. 3.

    During the Republican and Imperial periods, the Roman Empire conquered territories in the Mediterranean Sea (like Sicily, Sardinia, Corsica, the northern coasts of Africa), in the Atlantic Ocean and in the Black Sea.

  4. 4.

    The peak of the Roman Empire corresponds with full extent of the road network (117 A.D.—death of Trajan), as further proof that roads construction and the constitution of a vast empire were highly correlated.

  5. 5.

    This process included first of all the digging of a 1.5 m deep trench for the width of the road. In order to guarantee the stability and durability of the substrate, the trench was filled and packed with several textures and types of material from the land around it. Then they applied a layer of gravel or pavestones, ensuring that the road had a camber, or rise in the center, to prevent erosion and make the surface all-weather capable (Gleason 2013).

  6. 6.

    It has been argued that land transport was an inferior, expensive alternative to maritime transport.

  7. 7.

    The journey from Egypt to Rome, for example, took only two to three weeks by ship.

  8. 8.

    The access to the inland regions of the empire was allowed by the large navigable rivers: the Rhone, the Rhine, the Danube, the Tigris and the Euphrates and the Nile (Roth 1999).

  9. 9.

    The shape file is made freely available on the internet by the Digital Atlas of Roman and Medieval Civilizations (DARMC).

  10. 10.

    Talbert (2000).

  11. 11.

    Albania, Algeria, Austria, Belgium, Bosnia and Herzegovina, Bulgaria, Croatia, Cyprus, Egypt, France, Germany, Greece, Hungary, Israel, Italy, Jordan, Lebanon, Libya, Liechtenstein, Luxembourg, Macedonia, Montenegro, Morocco, Netherlands, Palestine, Portugal, Romania, Saudi Arabia, Serbia, Slovenia, Spain, Switzerland, Syria, the UK, Tunisia, Turkey.

  12. 12.

    A single road, such as the Via Appia, is composed by 67 segments, and each segment has a unique id number.

  13. 13.

    If it is a major or a minor road. Hereinafter the terms road and segment are used interchangeably.

  14. 14.

    Certainty refers to the path followed by the road. The road is certain in the existence and in the Roman origin, it is uncertain in the route.

  15. 15.

    For the majority of Roman roads in Greece and Turkey there is no information available about the certainty.

  16. 16.

    Exactly 192,861 km.

  17. 17.

    The instructions and technical notes of this section are the ones employed by the author. The procedure described is not the only one available. Expert users of GIS methods might come out with the same result using different tools or following a diverse method.

  18. 18.

    However, the measure can be computed for all those countries where Roman roads have been constructed and using different ‘within-country organizations’.

  19. 19.

    According to Eurostat the current NUTS classification is structured in three main levels: major macro-regions (NUTS 1), basic regions (NUTS 2) and small regions (NUTS 3). Municipalities and communes are defined by Eurostat as LAUs (Local Administrative Units). Formerly two levels of LAUs existed: LAU level 1 (NUTS 4) and LAU level 2 (NUTS 5). Since starting from 2017 only one level of LAU has been preserved, hereinafter the subdivision in municipalities will be referred as NUTS 4.

  20. 20.

    Local Labor Systems or Labor Market Areas or Sistemi Locali del Lavoro (SLL) in Italian are sub-regional geographical areas that identify where the bulk of the labor force is concentrated. They take into account where workers live, work and commute.

  21. 21.

    Grid cells are geometric units of a grid. They have identical size, that can be set by the researcher in meters or a different unit of length. However, when combined with countries’ layers, cells along the national boundaries are irregular in order to accommodate national borders.

  22. 22.

    Elaborations can be performed using ArcGIS as well.

  23. 23.

    The Italian National Statistical Institute provides three different polygonal shape files: one for the regions (NUTS 2), one for the provinces (NUTS 3) and one for the municipalities (NUTS 4). The polygonal shape file for the macro-regions (NUTS 1), not provided by Istat, can be obtained using the shape file of the NUTS 2 regions, creating a new field named ‘macro-region’, classifying regions according to the macro-region they belong to, and then dissolving the regional borders according to the macro-region field by using the ‘Dissolve’ geoprocessing tool. However this shape file could not work properly when exploiting some geoprocessing tools, like ‘Intersect’. Therefore, elaborations at the NUTS 1 level have been performed using the calculations for the NUTS 2 level, and then collapsing the information at the NUTS 1 level.

  24. 24.

    This geoprocessing tool in QGIS is named ‘Clip’.

  25. 25.

    However, it does not matter which linear layer for Roman roads is employed, since both lead to the same result.

  26. 26.

    Istat makes available the shape file by Local Labor Systems.

  27. 27.

    Although the starting linear layer by McCormick et al. (2013) provides the length of each segment, after having isolated and ascribed the segments to each territorial unit, the length in meters is not more valid since it refers to the complete section.

  28. 28.

    The information about the area in square kilometers at the NUTS level is provided by Istat. For the NUTS 1 level, data can be obtained by collapsing the information at the NUTS 2 level according to the macro-region the regions belong to.

  29. 29.

    The measure is 0 for all those territorial units where Roman roads are absent: this allows to construct a new binary variable that takes the value of 1 if Roman roads were present in the territory, and 0 otherwise.

  30. 30.

    Italy is composed by 611 local labor systems. Elaborations refer to the 2011 classification.

  31. 31.

    Italy can be decomposed exploiting different cell sizes.

  32. 32.

    Italy is structured in four macro-regions: North-West; North-East; Center; South.

  33. 33.

    Italy is composed by 20 regions.

  34. 34.

    The 2019 Italian National Statistical Institute (Istat) classification includes 107 provinces. However, all elaborations have been performed referring to the 2010–2016 Istat classification, according to which the Italian territory is divided in 110 provinces. The number of provinces has changed during last decades: they were 103 until 2005, 107 from 2006 to 2009, 110 from 2010 until 2016. The use of a broader classification allows a larger use of the measure, since it can be easily adapted to limited classifications by collapsing the index.

  35. 35.

    Italy is currently composed by 7926 municipalities. However, all elaborations have been performed referring to the 2010–2016 classification that includes 8092 municipalities.

  36. 36.

    The two Italian provinces where the Roman road network is absent are the province of Pordenone (in North-East) and the province of Verbano-Cusio-Ossola (in North-West).

  37. 37.

    The relationship between mountainous territories, elevation and Roman roads is examined in Sect. 5.

  38. 38.

    Consular roads, named after a consul, were the major and most important roads of the Roman Empire.

  39. 39.

    Figure 12 concentrates on the NUTS 3 level since this dimension seems to better represent differences within the Italian territory.

  40. 40.

    See Pirenne (1969) and McCormick (2001) for a more detailed analysis.

  41. 41.

    Lopez (1956) discusses the importance of the Roman road system in Europe during the Middle Ages.

  42. 42.

    At the NUTS 3 level the Italian National Statistical Institute does not make GDP data available, therefore the information about the total value added is exploited.

  43. 43.

    As mentioned before, the NUTS 3 classification is, for the Italian territory, the level that better allows to appreciate local differences in terms of institutions and government.

  44. 44.

    2016 is the last year for which data about value added at the NUTS 3 level are available. In order to make the analysis consistent, data about total exports and population refer to 2016 as well.

  45. 45.

    Variables in Fig. 8 are expressed in logarithm.

  46. 46.

    For the correlation analysis four main measures have been used: blood donation, organ donation approval, associationism, voluntary work.

  47. 47.

    Results are available upon request.

  48. 48.

    Data on kilometers of motorways at the NUTS 3 level are provided by Automobile Club d’Italia (ACI) and refer to 2011. Kilometers of railways are sourced from Istat and refer to 2005.

  49. 49.

    Data are provided by Istat.


  1. Acemoglu D, Johnson S, Robinson JA (2001) The colonial origins of comparative development: an empirical investigation. Am Econ Rev 91(5):1369–1401

    Article  Google Scholar 

  2. Acemoglu D, Johnson S, Robinson JA (2002) Reversal of fortune: geography and institutions in the making of the modern world income distribution. Q J Econ 117(4):1231–1294

    Article  Google Scholar 

  3. Bosker M, Buringh E (2017) City seeds: geography and the origins of the European city system. J Urban Econ 98:139–157

    Article  Google Scholar 

  4. Bosker M, Buringh E, van Zanden JL (2013) From Baghdad to London: unraveling urban development in Europe, the Middle East, and North Africa, 800–1800. Rev Econ Stat 95(4):1418–1437

    Article  Google Scholar 

  5. Buringh E, van Zanden JL, Bosker M (2012) Soldiers and Booze: the rise and decline of a Roman market economy in North-Western Europe. Utrecht University, CGEH Working Paper No. 32

  6. Dalgaard CJ, Kaarsen N, Olsson O, Selaya P (2018) Roman roads to prosperity: persistence and non-persistence of public goods provision, CEPR Discussion Paper 12745

  7. Davies HEH (1998) Designing Roman roads. Britannia 29:1–16

    Article  Google Scholar 

  8. De Benedictis L, Licio V, Pinna AM (2018) The long-term effects of the historical Roman road network: trade costs of Italian provinces, Working Paper Crenos 2018/01

  9. Engerman SL, Sokoloff KL (1997) Factor endowments, institutions, and differential paths of growth among new world economies: a view from economic historians of the United States. In: Harber S (ed) How Latin America fell behind. Stanford University Press, Stanford, pp 260–304

    Google Scholar 

  10. Engerman SL, Sokoloff KL (2002) Factor endowments, inequality, and paths of development among new world economies. Economia 3(1):41–110

    Google Scholar 

  11. Fogel RW (1964) Railroads and American economic growth: essays in econometric history. Johns Hopkins Press, Baltimore

    Google Scholar 

  12. Gleason JP (2013) Roman roads in Gaul: how lines of communication and basing support operational reach. School of Advanced Military Studies Monographs

  13. Greif A (1993) Contract enforceability and economic institutions in early trade: the Maghribi traders. Coalition. Am Econ Rev 83(3):525–548

    Google Scholar 

  14. Greif A (1994) Cultural beliefs and the organization of society: a historical and theoretical reflection on collectivist and individualist societies. J Polit Econ 102(5):912–950

    Article  Google Scholar 

  15. Guiso L, Sapienza P, Zingales L (2011) Civic capital as the missing link. In: Benhabib J, Bisin A, Jackson M (eds) Handbook of social economics. Elsevier, Amsterdam

    Google Scholar 

  16. Guiso L, Sapienza P, Zingales L (2016) Long-term persistence. J Eur Econ Assoc 14(6):1401–1436

    Article  Google Scholar 

  17. Knapton J (1996) The Romans and their roads—the original small element pavement technologists. In: 5th International concrete block paving conference, Tel-Aviv

  18. La Porta R, Lopez-de-Silanes F, Shleifer A, Vishny RW (1997) Legal determinants of external finance. J Financ 52(3):1131–1150

    Article  Google Scholar 

  19. La Porta R, Lopez-de-Silanes F, Shleifer A, Vishny RW (1998) Law and finance. J Polit Econ 106(6):1113–1155

    Article  Google Scholar 

  20. Laurence R (1999) The roads of Roman Italy: mobility and cultural change. Routledge, London

    Google Scholar 

  21. Lopez RS (1956) The evolution of land transport in the middle ages. Past Present 9(1):17–29

    Article  Google Scholar 

  22. McCormick M (2001) Origins of the European economy: communications and commerce, AD 300–900. Cambridge University Press, Cambridge

    Google Scholar 

  23. McCormick M, Huang G, Zambotti G, Lavash J (2013) Roman Road Network (version 2008), DARMC Scholarly Data Series, Data Contribution Series 2013-5

  24. Michalopoulos S, Papaioannou E (2018) Spatial patterns of development: a meso approach. Annu Rev Econ 10:383–410

    Article  Google Scholar 

  25. Michaels G, Rauch F (2018) Resetting the urban network: 117-2012. Econ J 128(608):378–412

    Article  Google Scholar 

  26. North D (1981) Structure and change in economic history. Norton, New York

    Google Scholar 

  27. North D (1990) Institutions, institutional change and economic performance. Cambridge University Press, New York

    Google Scholar 

  28. Pawson E (1977) Transport and economy: the Turnpike Roads of eighteenth century Britain. Academic Press, London

    Google Scholar 

  29. Pirenne H (1969) Medieval cities, their origins and the revival of trade. Princeton University Press, Princeton

    Google Scholar 

  30. Putnam RD, Leonardi R, Nanetti RY (1993) Making democracy work: civic traditions in modern Italy. Princeton University Press, Princeton

    Google Scholar 

  31. Ramcharan R (2009) Why an economic core: domestic transport costs? J Econ Geogr 9(4):559–581

    Article  Google Scholar 

  32. Roth JP (1999) The logistics of the Roman Army at War (264 B.C.–A.D. 235), Columbia Studies in the Classical Tradition, 23, Brill

  33. Snedden R (1998) Technology in the time of ancient Rome. Raintree Steck-Vaughn, Austin

    Google Scholar 

  34. Talbert RJ (ed) (2000) Barrington Atlas of the Greek and Roman World. Princeton University Press, Princeton

  35. Thompson L (1997) Roman roads. Hist Today 47(2):21–28

  36. Wahl F (2017) Does European development have Roman roots? Evidence from the German limes. J Econ Growth 22(3):313–349

    Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Vania Licio.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

The research underlying the paper was conducted during my Ph.D. studies at the University of Cagliari. I wish to thank the two anonymous referees and the editor Luca De Benedictis for their constructive comments and helpful suggestions. I also thank Anna Maria Pinna for her advice. I gratefully acknowledge the Sardinia Regional Government for funding my Ph.D. scholarship (P.O.R. Sardegna F.S.E. Operational Programme, Autonomous Region of Sardinia, European Social Fund 2007–2013—Axis IV Human Resources, Objective l.3, Line of Activity l.3.1.).

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary material 1 (rar 1985 KB)



See Figs. 9, 10, 11, 12, 13 and Tables 6, 7.

Fig. 9

Grid cell analysis: Roman roads by \(10 \times 10\) km grid cell. Source: Author’s elaboration from McCormick, M. et al. 2013. “Roman Road Network (version 2008),” DARMC Scholarly Data Series 2013-5

Fig. 10

Major Roman roads in density by NUTS level: quantiles. Source: Author’s elaboration from own data and from Istat (2011) data

Fig. 11

Certain Roman roads in density by NUTS level: quantiles. Source: Author’s elaboration from own data and from Istat (2011) data

Fig. 12

Major Roman roads in length, NUTS 3 level: equal intervals. Source: Author’s elaboration from own data and from Istat (2011) data

Fig. 13

Proxies of economic development in Italy: value added, total exports, nighttime luminosity, total population. Source: Author’s elaboration from Istat data and from Version 4 DMSP-OLS Nighttime Lights Time Series data

Table 6 Kilometers of Roman roads by NUTS 3 Italian province
Table 7 Density of Roman roads (kilometers of road per 100 square kilometers of land area) by Italian NUTS 3 province

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Licio, V. When History Leaves a Mark: A New Measure of Roman Roads. Ital Econ J 7, 1–35 (2021). https://doi.org/10.1007/s40797-020-00120-5

Download citation


  • Roman roads
  • History
  • Historical infrastructure
  • Persistence
  • Italy
  • NUTS

JEL Classification

  • H54
  • N13
  • O18
  • O52
  • R40