Technical Change, Non-Tariff Barriers, and the Development of the Italian Locomotive Industry, 1850–1913

The locomotive industry was one of the relatively sophisticated “high tech” sectors in which Italy, a late industrializer, was successful before 1913. Using technical data on the performance of different vintages of locomotives, we construct a new industry-level index of technical change. We also study the impact of different policy instruments (import duties, non-tariff trade barriers, and other discretionary interventions) in shaping the development of the industry. Our reassessment reveals not only the sound technological performance of Italian locomotives and successful industry growth, but also the critical role played by non-tariff barriers for the development of the industry.

ranging from trade and industrial policies to other forms of government actions. 2 In his assessment of the policy mix adopted by Italian governments in the second half of the nineteenth century, Gerschenkron was characteristically blunt, pointing to a generalized "ineptness of government industrialization policies." In particular, he argued that Italian industrialization would have bene ted from a more "rationally conceived and executed tariff." The structure of the tariff, favoring iron and steel in a coal-less country, represented a major impediment to the development of the Italian mechanical engineering industry, which "was largely left to its own devices" (Gerschenkron 1962, pp. 80-83). Other historians have instead argued that the sluggish development of the industry lay in the overall "technical and organizational backwardness of the sector" and not in mistaken policies. 3 This article sheds new light on these issues by reassessing the case of the steam locomotive industry, one of the few relatively "high tech" sectors in which Italy was successful before 1913. The development of this industry was shaped by a complex mix of policies comprising trade policies, "non-tariff trade barriers," and other discretionary interventions. The locomotive industry, thus, constitutes a useful vantage point to reconsider the general effectiveness of the different types of interventions adopted by the Italian governments in the period 1850-1913. Our analysis relies on a new index of technical change based on technical indicators of the performance of different vintages of steam locomotives. The dataset includes some six thousand steam locomotives in operation in Italy from 1839 to 1913. The index casts doubt on the prevailing view that the market share of Italian steam locomotive manufacturers was severely limited by their technical backwardness. On the contrary, the new quantitative evidence on technical change suggests that the technical capabilities of Italian producers were fully adequate when compared to those of foreign producers. Ef cient production required then an annual production of about 50 locomotives per rm. Tariff-based protectionism, with positive net protection in force since the mid-1880s, was not suf cient to reach this goal. Rather, this production level was achieved through speci c procurement policies assuring regular orders to national producers. It was these non-tariff barriers that played a critical role for the development of the Italian locomotive industry.
of railway lines and of rolling stock) grew throughout this period so that by 1913 it was such that, "on a value added basis, the maintenance of [rolling stock] was as important as their initial fabrication and, unlike the latter, it was immune from foreign competition" (Fenoaltea 1983, p. 49).
From an engineering standpoint, the steam locomotive is a sophisticated piece of equipment, comprising hundreds of parts, most of them requiring accurate manufacturing and processing. Philip Scranton (1997, p. 99) regards steam locomotives as a quintessential example of nineteenth century "specialty production": "no specialty product was then more complicated than a railway engine and few were as heavily taxed in use." Mass production was not feasible: steam locomotives were manufactured in small batches and they often had to incorporate particular speci cations requested by individual customers. This meant that locomotive manufacturers had to marshal a signi cant breadth of engineering competences and skills. Even if the broad contours of locomotive design were established by the 1850s, throughout the second one-half of the nineteenth century, steam locomotives underwent a stream of technical improvements, which compelled a continuous revision and updating of many details of locomotives design and manufacturing. 5 For these reasons, the steam locomotive provides an interesting perspective on the absorption of an "advanced" technology by latecomer countries. In comparative perspective, it is possible to distinguish two main "waves" in the international diffusion of the steam locomotive. During the rst wave , the manufacturing of steam locomotives became established in England, Belgium, France, United States, and Germany. During the second wave (1880-1920), we witness the emergence, with varying success, of the manufacturing of steam locomotives in Italy, Russia, Spain, and Japan. 6 5 See Sinclair (1907). According to Greggio and Kalla-Bishop (1985, p. 99), "It was not until 1890 that fundamental design theory could be said to be settled, for what was decided then lasted until the end of steam." Analogously, Cardwell (1994, p. 348) regards the end of the nineteenth century as the moment in which the locomotive had reached "a point of nearperfection." 6 The domestic production of steam locomotives in Spain remained limited until WWI. See Comín et al. (1998) and Cayón García and Muñoz Rubio (2005). On the more successful Japanese case, see Ericson (1998). In the Italian case, the production of steam locomotives has been investigated by Merger (1986Merger ( , 1989, and, more recently, by Ciccarelli and Fenoaltea (2012). The works by Merger are mainly concerned with the production of national manufacturers in quantitative terms and do not consider in detail the technological performance of the different models of locomotives. Ciccarelli and Fenoaltea (2012) present annual estimates for the railguided vehicles industry both at the national and regional level. INDUSTRY IN ITALY: NEW EVIDENCE, NEW CONJECTURES

THE STEAM LOCOMOTIVE
The evidence presented in this article is largely based on a new dataset described in the Appendix. It comprises two major components. The rst one includes information on 5,700 locomotives, both domestic and foreign. For each locomotive we know the year and place of production, the producer's name, and a numerical identi er. The second component is based on the authoritative handbook by Giovanni Cornolò (1998) providing technical characteristics and performances of the locomotives for the period 1850-1913. By combining these two components we can chart in detail the technological characteristics of the different locomotives operating in each year.

The Process of Import-Substitution
The development of Italy's locomotive market is described in Figure  1 and Table 1 illustrating both the total number of locomotives and those produced domestically.
Until mid-1880s, domestic production was limited and the market was dominated by foreign manufacturers. The Italian import market was initially dominated by English producers, then by French producers, and nally by German manufacturers for whom, the Italian market was particularly signi cant. The Maschinenfabrik Esslingen, among the leading German producer of steam locomotives, exported during the period 1883-1887 about one-half of its production in Italy (Hertner 1984, p. 30). Furthermore, as noted by Albert Schram (1997, p. 61) German banks backed the Rete Mediterranea, and the Railways Conventions of 1885 marked the beginning of German investments in Italy. During the 1890s few locomotives were purchased, but after 1900, demand resumed with Italian producers substantially increasing their market share.
Using data on the number of locomotives, Michèle Merger (1986Merger ( , 1989 argues that the limited production of Italian manufacturers during the period 1861-1885 is due to three concomitant factors: the limited depth of technological capabilities of national rms, a lack of specialization leading to increasing costs, and the penalties induced by the tariff on iron. After 1885, with the new Railways Conventions, a new wave of investments took place. In this phase Merger argues that the increased domestic production was a result of the public policies favoring national manufacturers, in particular Pietrarsa, Ansaldo, and Breda. 7 The ultimate consolidation of the national industry of steam locomotives, according to Merger, takes place after 1905 when the production of private national producers is elicited by a new wave of investment following the creation of Ferrovie dello Stato (FS), the new public operating company managing the entire main network (including both new construction and maintenance of the railway lines, and the maintenance of rolling stock). However, the increase in the number of domestically-produced locomotives tells us little about their technical quality.

Patterns of Technical Change
We assess the dynamics of technical change using the traditional weight-to-power (whp) ratio, a synthetic indicator that is widely used in 7 These companies had been able to develop a certain base of engineering capabilities in locomotive building by virtue of different channels of knowledge transfer (recruitment of foreign personnel, travels abroad, and licensing of foreign patents), see Merger (1986) and Vasta (2002). The whp ratio provides a measurement of the performance of the locomotive in a metric which is independent of its size, facilitating thus comparisons across different models and designs. 8 For each locomotive we computed the whp ratio using the data from Cornolò (1998). The whp ratio represents, therefore, our proxy for the technical performance of the various "types" of locomotive, with lower values denoting better performance and thus improved technology. 9 The different "types" of locomotives (Gruppi FS) were classi ed in 1905 by the engineers of the newborn Ferrovie dello Stato. Locomotives belonging to the same group are very similar in terms of technical characteristics. As Appendix A details, our sample includes more than 100 Gruppi FS. We construct an industry-level index (WHP) Source: See text. 8 The whp ratio, as any other synthetic indicator of technological performance, has of course its own limitations. First, it is a ratio, so it is not directly telling on locomotive weight and power, each per se relevant. Second, the whp is not particularly informative on other technological characteristics such as, for instance, locomotive top-speed and fuel ef ciency. 9 An increase in locomotive power (due to the introduction of, say, a bigger boiler) leaving unaltered the whp ratio is not interpreted here as technological progress. The whp ratio is a performance indicator that seems particularly relevant in the Italian context also because "The Italian lines had numerous metal girders of moderate resistance, so the locomotives had to be of reduced weight, both per axle and per linear meter; as a consequence [Italian locomotives] were forced to have a particularly high power" (Diegoli, 1961, p. 108). From this point of view the evolution of Italy's steam locomotive sector during the nineteenth century can be understood as a struggle between the increasing weight of locomotives and rails'-of iron rst of steel thencapability to bear it. of technical change as the weighted average of the technical performance of the different locomotive types entering in service in that year: 10 where t represents time (year), and s gt = n gt / N t is the share of locomotive of type g introduced in year t, (with n gt and N t indicating respectively the number of locomotives of type g, and the total number of locomotives, introduced in year t ); G t denotes the number of the different types of locomotives introduced in year t, and whp g is the weight-to-power ratio of locomotives of type g. 11 This index comprises only the locomotives introduced in a given year. While the sources indicate systematically the year in which a locomotive entered in operation, the information about the year in which it was scrapped is not systematic. This prevents us from constructing the index for the composition of locomotives in operation in a given year. The index of technical change, therefore, should be regarded as an index representing the level of technological performance of the investment in locomotives in a speci c year. Figure 2 shows the evolution of the WHP index for Italian and foreign manufacturers. The area shaded in gray is delimited, year by year, by the maximum and minimum sample values of the WHP index. Interestingly enough, after 1870, it is not uncommon to nd Italian manufacturers among the best performers (locomotives with minimum value of the whp ratio in each year).
The WHP index is characterized by a decreasing trend, re ecting technical progress for both domestic and foreign locomotives. The rst period (1850-1868) has wide uctuations, which is an outcome of the limited sample size. The three decades from 1868 to 1898 shows substantial stability, with a rapid acceleration from 1898 to 1913. This pattern is not surprising. Railways, in Italy as elsewhere, were rst laid out across the plains. Conquering the mountains required more powerful locomotives and better railway infrastructures, including tracks, bridges, and embankments. The rapid acceleration observed in this last phase is concomitant with the introduction of two major technical innovations: compounding and super-heating (Tey 1910, pp. 28-36;Diegoli 1961, pp. 108-9), both of which reduced the weight-topower ratio. Compounding and super-heating were the key features of the "second-generation" of steam locomotives. 12 Compound locomotives were adopted in 1894 by the Rete Mediterranea and only later by its competitor, Rete Adriatica. Table 2 complements Figure 3 by reporting the average WHP by decade separately for foreign and domestic producers; gures on the main Italian producers of the time are also reported in columns 3-5.
The evidence presented so far is not consistent with the limited engineering capacity suggested by the prevailing literature (on this point see Fenoaltea 2011, p. 150). Rather it supports contemporary views arguing that, at least from the late 1870s, Italian manufacturers were indeed capable of designing and producing steam locomotives of quality fully comparable to that of foreign competitors, or, more precisely, of the models of locomotives that English, French, and German manufacturers Source: See text. 12 Both innovations can been seen as the response to the threat of the electric locomotives in their pioneering phase after 1900.
were importing in Italy. For example, at a conference for the 1881 Milan Exposition, Professor Leonardo Loria stated: The national locomotives are perfectly equal, both in terms of manufacturing and assembling of components, to the best foreign locomotives … . Our locomotives are not anymore mere imitation of foreign locomotives, manufactured assembling a number of components imported from abroad, rather they are locomotives fully adapted to the special conditions of our railway network, almost completely manufactured by us, where our engineers introduce important innovations … . And, as far as the cost of production is concerned, today we are not far away from foreign rms. (Loria 1881, pp. 76-77) Additional corroboration of this point can be found in the detailed account of Cornolò who mentions other models of locomotives (such as the "Ariosto" or the "Frescot") designed during the 1870s and early 1880s that were crowned with critical acclaim at international exhibitions, and in the engineering literature. 13

Government Policy
The domestic production of locomotives increased over the nineteenth century. In this section, we examine the impact of government policy on the growth of the industry. Interestingly, from 1870 to 1885, several parliamentary select committees (PSCs) debated alternative forms of State intervention to sustain the national industry. Of particular relevance here are the early 1870s PSC "on industry," the late 1870s PSC "on railways," and the mid-1880s PSC "on the revision of the trade tariff." The PSC "on industry" (Comitato dell'inchiesta industriale) investigated the capacity to compete with foreign producers and the ways in which custom duties increased domestic competitiveness. 14 The engineering sector was one of the most debated with many observers and practitioners (including businessmen, managers, and engineers) interviewed. The main locomotive producers were Ansaldo (near Genoa, established in 14 The board was promoted by Luigi Luzzatti, an economist and politician of the time, appointed Prime Minister in 1910. Luzzatti and Vittorio Ellena exerted a major in uence on the diffusion of protectionism in Italy through the tariff reforms of 1878 and 1887. Pareto himself used to refer to the 1887 protectionist reform as the "Ellena-Luzzatti" tariff (Pareto 1891). 1854) and Pietrarsa (near Naples, established in 1842). They were asked about the effects of import duties on the domestic production of steam locomotives. Both Ansaldo and Pietrarsa argued that the tariff structure worked against the domestic production of locomotives. 15 Tables 3 and  4 reproduce the evidence referring to the Pietrarsa (Naples) and Ansaldo (Genoa) workshops. 16 Table 3 reports the cost structure of a typical locomotive of the Pietrarsa workshop. Column 2 shows that some 70 percent of a "standard" steam locomotive was made of iron, while the remaining 30 percent was of pig-iron, brass-tubes, copper, steel, and bronze. Column 4 reports the total physical cost of production (excluding labor and transport costs). The total cost of materials was 29,610 lire, and 32,970 lire once import duties are included (corresponding to a 10 percent increase). The last line shows that the import duty on a steam locomotive considered as a nished product amounted to 4 lire per 100 kg. With a total weight of 40 tons, the import duty amounted to 1,600 lire. 17 The tariff's structure, therefore, implied an additional cost of 3,360 lire for domestically-produced locomotives relative to 1,600 lire for imported locomotives, resulting in a negative protection for Italian locomotive producers. Table 4 reports similar data provided by Ansaldo. For a standard "1858 Government type" locomotive, the import duties on raw materials were 2,811 lire, well above the 1,570 lire paid as duty to import the same locomotive from abroad. Although the calculations by Ansaldo are extremely coarse, they show that in the early 1870s at least one-half of a locomotive was made up of highly-protected-iron, amounting to a negative protection.
The PSC "on industry" concluded that: "there is no balance between the ("high") tariff on iron and the one ("low") on machines"; therefore "it is necessary to reduce the tariff on iron," or "alternatively, it is necessary to raise the tariff on machines" (Comitato della inchiesta industriale 1874, p. 6). In fact, both the duties on iron and locomotives were raised over time. In the subsequent PSC "on railways" (Commissione d'inchiesta sull'esercizio delle ferrovie italiane), Pietro Peirano, a manager of 15 As already noticed by Ciccarelli and Fenoaltea (2012), Felice Giordano-in his 1864 appraisal of the iron and metal working industry in Italy-argued convincingly that the engineering establishments of Naples and Genoa could potentially produce locomotives at prices that were similar to that of foreign ones. The argument made by Giordano refers to total cost of production, so that while iron was surely more expensive in coal-less Italy, labor was there relatively cheaper than in other countries (Giordano 1864, pp. 102, 359). 16 The sources report an import duty on steam locomotives of 4 lire per 100 kg in one case (Pietrarsa), and 4.62 lire per 100 kg in the other (Ansaldo). From that evidence one can infer that in one case (Pietrarsa) the data refer to some year between 1866 and 1870, while in the other case (Ansaldo) to either 1871 or 1872. 17 Giordano (1864, p. 349) reports an input weight loss that may reach some 25 percent for certain materials. As a result, the negative protection illustrated in Tables 3-4 is surely underestimated.
Ansaldo con rmed that the key factor forcing his company to give up the production of locomotives was the penalty induced by the tariff rather than the lack or backwardness of technological competency. 18 Ansaldo and Pietrarsa's delegates were also consulted by the PSC on "the revision of international trade tariff" (Commissione d'inchiesta per la revisione della tariffa doganale) instituted in the mid-1880s and reaf rmed that the cost of protecting the metal-making sector (with high import duties on iron and other production inputs) represented a still too heavy burden for the engineering sector. 19 18 Peirano con dently stated "We could outcompete foreign rms if raw materials were exempted from import duties." Even if self-interested, in our view the statement probably re ects a sincere assessment. In the same interview it is also discussed the special procurement policy practiced by Ferrovie Meridionali in favour of Pietrarsa. Clearly, if Peirano had not been intimately convinced of the critical role of the tariff, the best course of action would have been to advocate for the extension of a similar procurement policy rather than the revision of the tariff (Commissione d'inchiesta sull'esercizio delle ferrovie italiane 1879, pp. 371-72). 19 Commissione d'inchiesta per la revisione della tariffa doganale (1886, pp. 423-33).  Figure 3 shows that subsequent tariff reforms-above all that of 1887-gradually ensured a positive protection to the Italian producers of steam locomotives. The vertical distance between the continuous and dashed line provides a measure of the net-protection on steam locomotives. Obviously the measure is rough for several reasons. Iron was not the only input in the production of a steam locomotive, though most relevant in terms of weight. Second, Italian import duties were mainly tied to weight (not to value), and thus less effective the higher the price of imported goods. Finally, but importantly, the inputs' weight losses occurring during the production process are not considered. Including them would shift the dashed line upward perhaps by as much as 20-30 percent. 20 Despite these limitations, Figure 3 tentatively suggests that the net protection on steam locomotives became positive over time: a 20 This percentage is suggested by the early calculations reported in Giordano (1864, p. 349), and con rmed by the more detailed tables reported in the later Camera dei Deputati (1888, pp. 52-54). consequence of the increase, in different proportions, both of the import duties on iron and locomotives. Import duties on steam locomotives were raised from 4 to 4.62 lire in 1871; 4.62 to 8 lire in 1873; 8 lire to 14 from 1888. Import duties on iron were raised from 4.62 lire to 6.50 in 1888 and then lowered to 6.00 lire from 1892. 21 To sum up, the evidence we have discussed so far suggests that the structure of the tariff rather than alleged technological backwardness represented the main bottleneck sti ing the expansion of the industry up to mid-1880s (Figure 2). 22 This became less stringent as the effective rate of protection on steam locomotive rose in the late 1880s. 23 The growth in production after 1885 is generally seen as the result of two policy changes. The rst is the aforementioned 1887 reform of the tariff, raising the import duty on steam locomotives from 8 to 14 lire per 100 kg. 24 The second is the 5 percent clause introduced by the Railways Conventions of 1885, which required that locomotives procurement contracts should be assigned to national manufacturers if their prices were less than 5 percent higher than the best offer of foreign competitors. 25 This is a point of contention in the literature: several historians argued that the 5 percent 21 Annual data on import duties are from the Movimento commerciale del Regno d'Italia, the of cial historical source on Italy's commercial ows. Data for the year 1874 are, for instance, from Ministero delle nanze (1875, p. 54 (locomotives) and p. 61 (iron)). Data for year 1888 are, for instance, from Ministero delle Finanze (1889, p. 218 (iron) and p. 256 (locomotives)). Data on steam locomotives are separately identi ed starting with 1886. In the preceding years, the duties on steam locomotives were reported within the more general category of "steam machines." Data on iron refer to "ferro di prima fabbricazione" for the years 1866-1874, to "ferro laminato o battuto o in verghe di più di 5 millimetri di diametro" for the years 1870-1884, and to "ferro e acciaio laminato o battuto o in verghe, spranghe o barre sagomate -non aventi in sezione alcun diametro o lato di 7 millimetri o meno" for the years after 1885. Luckily enough, overlapping retrospective gures are regularly reported in the source, allowing the reconstruction of plausible time-homogeneous series of import duties. Import duties for the years before 1865 were ad valorem and thus, for the sake of simplicity, here ignored. 22 Thus con rming the intuition in Ciccarelli and Fenoaltea (2012, pp. 55-56). 23 Federico and Tena (1999) have produced a detailed set of estimates of "effective rates of protection" by using the technical coef cients of three input-output tables (one for the United Kingdom in 1907, one for Italy in 1911, and one for Italy in 1950). Their estimates suggest that the effective protection rate for the mechanical engineering was probably not very different from zero or marginally positive over the period 1889-1913. Their estimates are therefore broadly consistent with the views of informed contemporaries discussed here. 24 The 1887 tariff reform raised import duties on both iron and machines (including locomotives). Still the main Italian producers of steam locomotives were for the rst time able to export their products. A possible explanation is that starting with late 1880s exporting rms were exempted from the payment of duty on iron. According to Società Italiana Ernesto Breda (1908, p. 50), Breda exported in the 1892-1902 decade 137 steam locomotives mostly in Romania and Denmark (Società Italiana Ernesto Breda 1936, p. 66); Popescu (1987, p. 333). It may be worth noting that the Romanian market of the time was dominated by major producers such as Beyer & Peacock,Henschel,Maffei,and Wiener Lokomotivfabriks. 25 Ministero dei lavori pubblici (1901, pp. 203-204); see in particular footnote 1.

Net Protection on Steam Locomotives
clause was probably too weak to exert a major impact (Gerschenkron 1962, p. 371;Caizzi 1965, p. 381;and Papa 1973, pp. 50-51), but the most recent contributions such as Merger (1986, p. 84), Renato Giannetti (1999, p. 1134), and Giannetti and Michelangelo Vasta (2012, p. 226) claim rather that the clause played an important role in the "take-off" of domestic production of steam locomotives after 1885.

The Silent Procurement Policy, 1885-1899
There are, however, two other features of the Railway Conventions of 1885 that have received only limited attention in the literature, but are possibly far more important than the 5 percent clause and the revision of the tariff in accounting for the expansion of the domestic production. The rst is the special endowment of 15 million lire granted, with the reorganization of the railway system in 1885, to the three major operating companies to purchase railway equipment and material and also to renew their locomotive eet (Merger 1986, p. 81). The second is the active procurement policy that was put in place after 1885, which, although not established by law, amounted to favoritism of "national champions" well beyond the preference arising from the 5 percent clause.
The point is clearly stated in an of cial publication of the "public works" ministry explicitly acknowledging that a segment of the Italian market for locomotives was to be reserved for national manufacturers: In the case of locomotives the protection afforded by the aforementioned article [i.e., the ve per cent clause], was not suf cient to allow national rms to win procurement contracts and, even a higher protection would have not been enough. Hence, in order to encourage the Italian locomotive industry, which is still in its infancy and it is practiced by very few rms that are now constructing the plants necessary to win against foreign competition, several procurement contracts have been assigned to these rms by means of private deals, xing prices in such a way to allow these rms to manufacture locomotives without making losses and not making a too heavy burden for the State. (Ministero dei lavori pubblici 1889, pp.

501-2)
The 5 percent clause introduced by the Railway Conventions of 1885 was to be applied to contracts assigned by means of formal competitive calls ("licitazione"), but the franchise companies could also procure locomotives directly from the manufacturers outside of the competitive system as illustrated in Table 5.
Two points merit attention. Columns 4 and 5 show clearly that the 5 percent clause was insuf cient to tilt the price advantage in favor of national manufacturers. Second, a signi cant stimulus to the national industry was implemented by means of special deals. 26 In particular, from the Railway Conventions of 1885 to 1900, the domestic production of steam locomotives was assigned either by national auctions (284 units out of 595) or by private negotiations (311 units out of 595). To be sure, the price of foreign locomotives was lower than national ones, particularly so in the case of locomotive awarded through direct procurement contracts (about 74 thousands lire against about 48 thousands lire). Most likely, this price differential re ected both the possible higher ef ciency of foreign producers, but also the existence of implicit agreements guaranteeing high prices to Italian producers. c Open calls ("licitazioni internazionali") were opened to both foreign and Italian producers d National calls ("licitazioni nazionali") were reserved for Italian producers. e The gure (595) also includes 32 locomotives built by the Mediterranea (30) and Adriatica (2) in their own workshops, accounting for a total value of 2,611,631.62 lire. Source: Ministero dei lavori pubblici (1901, pp. 202-9).
26 Calzavarini (1966, p. 74) points to the key role played by "non-tariff trade barriers" in compensating the "the insuf cient protection of the trade regime." Calzavarini's appraisal has been largely neglected by the subsequent literature. For a useful discussion of non-tariff barriers, see Baldwin (1970).
The adoption of stronger protectionist measures to favor Italian producers appears to be a leitmotiv of the parliamentary discussions. 27 Despite various calls to exclude foreign locomotives, the government rejected such outright protection. However, the government agreed that, while the various forms of tariff protection were not suf cient, the most critical issue was the irregularity of demand. Informed contemporaries identify in the small and irregular number of orders the main problem affecting the cost competitiveness of national manufacturers. In 1881 Giuseppe Colombo wrote: The convenient manufacturing of steam locomotives requires a demand of at least 50 units per year, and a workshop with the most sophisticated and specialized machinery. Two such workshops would probably suf ce to supply the needs of whole Italian railway network. Given the methods of locomotives production prevailing today also in the major foreign rms, where many components are directly purchased by iron producers or other specialized manufacturers and … given the proved ability of Italian producers in both new productions and maintenances … the success of such undertaking in Italy would be, no doubt, possible. (Colombo 1881, p. 67) A similar point was also stressed, a quarter century later, in the 1908 celebratory volume for the 1,000th locomotive constructed by Breda: The technical dif culties that Breda had to face daily were exacerbated by two key factors: the erce competition from foreign producers, and the irregular and discontinuous nature of the orders. (Società Italiana Ernesto Breda 1908, p. 22) These statements suggest that more than the shield of tariff-based protection (with positive net protection effectively achieved from the late 1880s) it was the irregularity of demand that stymied the Italian manufacturers of steam locomotives during the nineteenth century. 28 In this context, the Railway Conventions of 1885 played a critical role in reassuring the two main national manufacturers (Ansaldo and Breda) that the new political environment was going to be quite favorable to domestic producers of locomotives. This resulted in an expansion of 27 An account of the passionate parliamentary debates on the locomotive industry at the turn of the century, not reported here for reasons of space, can be found in Ciccarelli and Nuvolari (2015). 28 The highly irregular nature of the demand was not only a feature of the Italian market. For a discussion of the "feast-or-famine" nature of locomotive demand in this period in different countries, see Ericson (1998, pp. 133-34). the productive facilities of both rms so that domestic production increased relatively swiftly from a few units to more than a hundred units. 29 THE BIRTH OF THE FERROVIE DELLO STATO (1905) AND THE NEW RAILWAYS POLICY The late nineteenth-century parliamentary debates on railways had important consequences in terms of actual economic policies. The government presented a plan concerning the acquisition of new rolling stock for the years 1900-1904. It was established (article 9 of law no. 57 of 27 February 1900) that the quantity, quality, and expenditure for new purchases of locomotives had to be agreed between the State and the operating companies. Furthermore, during the same ve year period, the Government provided the companies a non-repayable grant of 28 million lire, reinforcing thus its role as chief sponsor of the railway sector. Not surprisingly, it has been remarked that the railway plan of 1900 represented the greater State commitment towards railway since the uni cation (Papa 1973, p. 50). In 1905, the 20-year contract between the State and the operating companies (the Railways Conventions of 1885) was not renewed and the State assumed the direct management of the railway system through the creation of Ferrovie dello Stato (FS), a fully public company.
The rules of the game were changed considerably. The new legislation established that "the Government will assign the production of new rolling stock to national manufacturers … safeguarding, other things being equal, a fair distribution among the various producers" and "whenever necessary … the executive manager [of the newborn FS] can disregard open and national call and use private negotiations …", but importantly, "in case of collusive behavior or any kind of fraud by national manufacturers to the detriment of the railway administration [that is the FS], or when it will not be possible for national producers to guarantee a fair price of rolling stock, according to the prevailing market conditions, then … [the Government] … will order new rolling stock to foreign producers." 30 Two points deserve attention. The rst is that the tool of direct procurement to assign orders to national producers was explicitly 29 On the investment in production capacity (in particular in that related to the production of locomotives) in the pioneering years of Breda, see Licini (1994). 30 The framework of the new policy regime was established by law no. 137, April 1905 on "the public management of the railway sector," and by law no. 429, July 1907, on "the State management of railways not licensed to private rms." The citation refers to article 11 of law no. 137, April 1905. institutionalized. The second is that the State intended to use international auctions as a threat to limit "extortionary" pricing by national producers. 31 While the protection of the national industry was thus con rmed and even reinforced, a more rigorous expenditure policy was put in place.
Tables 6 and 7 document the effects of the new FS regime on locomotive acquisitions and prices. Table 6 provides quantity and price data from an of cial 1906 report. 32 A total of 291 locomotives were assigned to national producers by direct procurement contracts, and 112 to foreign producers through international auctions. Table 7 provides an assessment of the price gap between national and foreign manufacturers in 1906. It shows that by 1906 the price gap was only about 9 percent, signi cantly lower than in the 1885-1899 period. 33 31 Direct procurement contracts were rarely used to import locomotives from abroad. The Minister of public works Carmine mentions for instance the case of "unusual" private negotiations with foreign producers concerning 25 locomotives (Camera dei Deputati 1906b, p. 6652). 32 The report of March 1906 to the Parliament by the Minister of public works Pietro Carmine (the so called "Carmine-report," Camera dei Deputati 1906a). 33 The price ratio between national and foreign locomotives reported in Table 7 can be used to perform a tentative estimation of the total factor productivity (TFP) gap in steam locomotives production between Italy and Germany around 1906. We use the formula where A = TFP levels; W= real wages; R = rental cost of capital; C = raw materials and , , and are the shares of labor, capital, and raw material in total costs. The suf x * indicates the foreign country. Assuming that the rental cost of capital was the same in the two countries, using the data in Table 3 to estimate cost shares, data on real wages from Allen (2001), and on the price of iron (which approximates the raw materials) from Cianci (1933, pp. 329-30), one gets that the level of TFP in Italy was about 85 percent of the German level. Taking into account the crude nature of the data, the assumptions made, and that the extent to which the prices reported in Table 7 were subjected to political manipulation remains uncertain, these estimates provide only a very rough indication of the TFP gap between the two countries.
To summarize, with the creation of the FS in 1905, the State implemented a new policy mix capable to stimulate the national industry, by directly assigning to local producers new locomotive orders, and, at the same time, to safeguard the State budget by linking the price effectively paid to national producers to the best offers of foreign producers. 34 Foreign purchases were used to investigate the price of locomotives and to absorb demand peaks, as the one occurred in 1907-1908 when about 900 locomotives entered the FS locomotive eet.

CONCLUDING REMARKS
Notwithstanding the limited endowment of the country in terms of engineering competences, the Italian steam locomotive industry achieved a level of technological sophistication in line with that of foreign producers by the beginning of the twentieth century. The early development of the industry was slowed both by the vagaries of the demand which prevented the major national players from planning an ordered expansion of productive capacity and, secondly, by tariff protection on iron products, which resulted in a cost disadvantage. From 1885 onwards national manufacturers received considerable support through a discretionary procurement policy which in practical terms amounted to an effective "non-tariff barrier" and to a more regular demand for their products, tied to the spread of the railways network leading to greater economies of scale.
The design and construction of locomotives were probably among the most sophisticated segments of the nineteenth century mechanical engineering industry and it is instructive to compare the performance of the locomotive industry with other less sophisticated branches such as cotton textile machinery where the country was unable to develop any autonomous production (Besso 1910, pp. 142-43;A'Hearn 1998). Yet, there were also other complex engineering products in which Italy was also able to deliver a very good performance, such as the production of warships (Fenoaltea 2011, p. 150), some of which were even exported. This somewhat paradoxical outcome can be explained taking into account that some of the most sophisticated branches of the mechanical engineering could endure the negative effects of the tariff, by virtue of non-tariff trade barriers such as public procurement, subsidies, and other similar policies, whereas the less sophisticated branches had to bear its full burden.

Appendix A The Steam Locomotives Dataset
The steam locomotives dataset is based on two major components. The rst one is constructed with the same approach and sources used in Ciccarelli and Fenoaltea (2012) for their statistical reconstruction of Italy's rail-guided vehicles industry. 35 The second component is entirely new and covers indicators of technical characteristics and performances of locomotives.
The main source for the rst component of the dataset is constituted by the catalogues of the principal companies operating the railways system at different moments in time. Catalogues of this kind were published from the late 1880s by three main operating companies (Rete Adriatica (RA), the Rete Mediterranea (RM), and the Rete Sicula (RS)). 36 A similar catalogue was published in 1914 by the Ferrovie dello Stato (FS), providing an exhaustive list of locomotives in operating service at the date of 30 June 1914. 37 The catalogues set out the technical characteristics of each type of locomotive in service together with a numerical identi er, the year of construction, and the name of the building company. Our dataset is constructed by merging the relevant quantitative information stored in these four publications.
The second component of the dataset concerns technical features. For each locomotive model (Gruppo FS) we have retrieved the information on locomotives' weight and power (measured in HP) from Cornolò (1998, pp. 584-99).
The WHP calculations presented in the previous Patterns of technical change section refer to 4,432 steam locomotives with separated tenders; the remaining 1,268 tanklocomotives (locomotives carrying water and coal on board instead of pulling them behind in a separated tender), with a WHP far higher and well above 100, are excluded. 38 Locomotives in our sample are grouped according to a classi cation scheme adopted by the Ferrovie dello Stato (FS) in 1905. Table A1 illustrates the distribution of locomotives in our sample by (104) locomotive groups. The table includes two panels. Panel A includes locomotive groups from 100 to 560; these are mostly of " rst-generation" (i.e., using single expansion and saturated, or "wet" steam). Panel B refers to groups from 600 to 980, and includes mostly "second generation" locomotives (i.e., adopting super-heating and/or doubleexpansion). The table's columns are numbered as follows. Odd numbers refer to locomotive groups. Even numbers refer to the units of locomotives within a given group. Locomotive groups included in the same column are similar in terms of technical characteristics (speed, power, weight, wheels arrangement, etc.). The bottom part of Table  A1, panels A and B report the average sample weight, the average sample power, and the power-to-weight-ratio evaluated at the sample average. 35 A complete account (in Italian) on sources and methods can be found in Ciccarelli and Fenoaltea (2014 (1914). 38 We do not consider in our sample tank-locomotives because they are typically of reduced size and weight, and often operating in narrow gauge lines (Ministero dei lavori pubblici 1901, pp. 200-1, where the tank-locomotives are classi ed as a category apart, with an average price well below that of standard locomotives).