The impact of access to railroads on economic growth
in the Ottoman Empire, 1893-1914
Avni Önder Hanedar1
Dokuz Eylül University
Abstract
This paper tests whether access to railroads in a judicial district changed the size of
the population in the respective judicial district for the years 1893 and 1914. The
empirical results indicate some evidence for a positive effect of access to railroads on
population in the respective judicial districts. This paper concludes that railroads
induced economic growth, leading to higher population in the respective places.
I. Introduction
In the late nineteenth and early twentieth centuries, there was an important transition
in the Ottoman Empire regarding its transportation networks, which could have
caused economic growth. This consisted of the construction of railroad lines between
1856 and 1914. The building of railroads began in 1856 and railroad networks grew
rapidly until 1914. As shown in Figure 1, railroad lines connected Rumelia2 with
other parts of the Ottoman Empire, such as Hejaz, Syria, Iraq, and Anatolia. The
railroad networks were not as dense as one might expect because of financial
difficulties, wars, and political obstacles, limiting construction even up until 1914
(von Pressel 1966, 91; Schoenberg 1977; Ortaylɪ 2010, 163-65; Geyikdağɪ 2011, 908).
Figure 1. Railroad Lines in the Ottoman Empire (1856-1914)
Source: Figure 1 is taken from McCarthy (2001, 29).
In the Ottoman Empire, there were only a few railroad lines that were built by
the state itself. Many railroad lines were constructed and operated by the foreign
railroad companies of major European powers.3 This could be due to the fact that the
Ottoman state did not have the funds to finance the construction of railroads. Its
financial situation became worse during the 1850s and it went bankrupt in the 1870s
1
e-mail: [email protected]
Land of the Ottoman Empire in Europe.
3
They were the UK, France, Germany, and Austria-Hungary.
2
1 because of its debts. The Ottoman Empire was also behind the contemporary
technological knowledge for building railways (Schoenberg 1977; Gülsoy 2010, 27;
Geyikdağɪ 2011, 119-126).
The Ottoman state primarily wanted the railroads to be built so it could
exercise political control over its territories. The major European powers constructed
several railroads, such as the line connecting İzmir with Aydɪn, to extend their
economic control and import raw materials from the Ottoman Empire. The decreasing
transportation costs made it easy to obtain imports from the Ottoman Empire.
However, only a few railroad lines were allocated to places of economic importance
(Kolars & Malin, 1970; Karkar 1972, 65; Schoenberg 1977; Quataert 1977; Quataert
1996, 806; Illich 2007, 91-2; Gülsoy 2010, 27-8; Hülagü 2010, 23-25).
In the historical literature, there are controversial arguments for the impact of
railroads on economic growth in the Ottoman Empire. Several researchers argue that
there had been a substantial increase in trade and agricultural production as places
had gained access to railroads.4 On the other hand, while railroads are generally
viewed as having a positive effect on growth, there is a strand of literature where it is
argued that railroads could lead to lower production in connected places, as there was
an increase in imports due to railroads, which resulted in the domestic industry being
hampered in places where railroad lines were located (Quataert 1996, 814). In another
strand of the literature it is argued that railroads did not lead to economic growth in
the Ottoman Empire. This strand is based on the argument that railroad lines were not
adequately connected to one another (Ortaylɪ 2010, 166),5 and as a result passenger
and goods traffic was not intense.6
To shed light on this issue, this paper examines whether access to railroads
induced economic growth in the Ottoman Empire, using data for 455 judicial districts
(known as kazas) from 32 provinces (known as vilayets) in the Ottoman Empire for
the years 1893 and 1914. Hornung (2012) uses population growth as a proxy for
economic growth in Prussia. He reports the positive impact of railroads on population
figures in connected cities in Prussia for the period 1840-1871 and argues that
railroad access induced economic growth and increased employment opportunities in
connected cities. Higher employment opportunities attracted people into the
respective cities to look for jobs. There is no data source with information about
economic outcomes, trade, transportation costs, and internal migration at the judicial
district level. This paper conducts a similar analysis to that of Hornung (2012).
Population data come from the 1881/82-937 and 1914 censuses, which have been
4
See British Parliamentary Paper (1896, 10), Karkar (1972, 65, 82), Quataert (1977), Eldem (1994,
94), Gülsoy (1994, 245-6), Quataert (1996, 810-4), Quataert (2005, 126), Gülsoy (2010, 181-2,
270), and Hülagü (2010, 14-45).
5
This is because all major European powers wanted to build railroads to control places politically and
economically (Illich 2007, 91-3; Geyikdağɪ 2011, 54-5). Furthermore, the construction of the
railroads by one power would create a threat to its rivals` economic and political dominance in
the respective location. Competition among the major European powers to control the respective
location via railroad investments would lead to conflicts (Ortaylɪ 2010, 165-6). For instance, in
1836 Colonel Chesney, who was a British engineer, had proposed the construction of several
railroads in the Ottoman Empire to the UK government. The railroads would have connected the
Syrian coast with Baghdad (von Pressel 1966, 92). However, since railroads would create a threat
to the economic dominance of France in these areas, France blocked the building of these
railroads using a French diplomat, that is, Ferdinand de Lesseps (Gülsoy 2010, 43-4).
6
Another reason for this is that railroads were not connected with highways (Eldem 1994, 105).
7
The 1881/82-93 census was conducted after 1881 and submitted to the Sultan in 1893 (Karpat 1985,
30-6).
2 made available by Karpat (1985, 122-50, 167-90). This paper constructs a detailed
map that shows the location of the railroad lines by both place and year.
II. Empirical approach
This paper runs the following regression:
log(Picpt ) =β 0+β1 log(Ricpt ) + β2 Dicp × γ t + β3Micpt + ρ p + uicpt
where i, c, p, and t indicate the judicial district, county, province, and year,
respectively. log(Picpt) is the logarithm of the judicial district level population for
judicial district i, located in county c of province p, in year t. log(Ricpt) is the logarithm
of the distance between the nearest railroad line and each judicial district, in year t. β1
is the coefficient of interest.
To control for initial conditions in a judicial district, this paper uses Dicp that equals 1
if natural disasters were seen in judicial district i of county c, located in province p,
between the 1800s and 1850s, and zero otherwise. Micpt is a dummy variable which
equals 1 if there was at least one operating mine, such as a coal, gold, or copper mine,
in judicial district i of county c, located in province p, in year t, and zero otherwise.
There was an increase in the population and production in several places, such as
Balya judicial district in Balɪkesir, after coal mines had been operating in the 1900s in
the respective locations (Su 1939, 6-12). In addition, the regression includes the
interaction of Dicp with year dummies (ƴt) that control for the time varying effects of
different initial conditions in judicial districts on the population. ƿp are province fixed
effects that account for time-invariant characteristics at the province level (for
example, the presence of mountains, rivers, and lakes, and geographic size). Lastly,
uicpt is the error term.
If being closer to a railroad line was positively associated with economic growth
through increasing production and trade which attracted an inflow of workers and
their families, then β1<0. However, a positive coefficient estimate of β1 implies that
being closer to a railroad line could lead to lower production and trade through
increasing imports which resulted in hampering of domestic industry, and
consequently population went down due to increasing migration.
III. Results
The OLS (Ordinary least squares) estimate for the effect of railroads in column (1) of
Table 1 is statistically significant at 10 per cent. The point estimate reveals that a one
per cent decrease in the distance from a judicial district to the nearest railroad line
leads to an increase in the population of the respective judicial district by 0.03 per
cent, on average.
Table 1. Impact of railroads in the Ottoman Empire (1893 and 1914)
Dependent variables: log(Population) and log(Distance to railroad lines)
1
2
3
4
log(Distance to railroad lines)
-0.033*
-0.056
--0.033*
(0.019)
(0.103)
(0.019)
Presence of natural disasters
0.065
0.052
-0.466
0.065
(0.143)
(0.158)
(0.316)
(0.143)
Presence of mines
0.223***
0.226***
0.092
0.223***
(0.072)
(0.069)
(0.147)
(0.072)
log(Distance to trade and
--0.302***
caravan routes)
(0.110)
log(Distance to lines proposed
----0.118
by Colonel Chesney)
(0.073)
log(Distance to lines proposed
----0.004
by Levant Herald)
(0.069)
3 Province fixed effects
Presence of natural
disasters×year fixed effects
Constant
N. of obs.
R2
F
YES
YES
YES
YES
YES
YES
YES
YES
9.628***
(0.031)
686
0.41
--
9.670***
(0.188)
686
0.41
--
1.008***
(0.296)
686
0.58
7.48***
9.628***
(0.031)
686
0.41
--
Notes: Column (1) reports the OLS estimates for equation (1). In column (2), regression is estimated by 2SLS (Twostage Least Squares). Column (3) shows the first-stage regression of 2SLS. The dependent variable in column (3) is
the logarithm of the population in a judicial district in year t. Column (4) reports the OLS estimates for the
regression that examines the impact of lines that were planned but that were not built. Heteroskedasticity-robust
standard errors corrected for clustering at the province level are reported in parentheses. *** and * denote
statistically significantly differences from zero at the 1% and 10% levels, respectively. N. of obs. is the number of
observations. In the last row, the F statistic on the excluded instrument in the first-stage regressions is reported.
Lastly, the distance to the trade and caravan routes is the instrumental variable.
In line with the findings of Hornung (2012), the empirical results imply that
access to railroads had induced economic growth, which led to a higher population
figure in the respective judicial district. The results are also parallel to those in the
historical literature, indicating that there was an increase in international and
interregional trade because of lower transportation costs, which led to higher
agricultural production in places that had gained access to railroads. Higher trade and
agricultural production resulted in increasing employment opportunities, which
attracted an inflow of people looking for employment and jobs (Schoenberg 1977;
Eldem 1994, 94; Gülsoy 1994, 245-8; Quataert 1996, 810-4; Gülsoy 2010, 181-5,
270; Hülagü 2010, 14-45).
This paper also addresses the possibility that the placement of railroad lines
could be endogenous to the population size (that is, reverse causality), which leads to
a bias in the OLS estimates for the impact of railroads. This is because in the
historical literature it is indicated that one of the railroad lines was allocated to places
that were commercially important (that is, İzmir and Aydɪn) (Kolars & Malin 1970;
Karkar 1972, 65, 79). The Ottoman state wanted several railroad lines to be built in
unpopulated places, such as Bandɪrma and Balɪkesir, due to famines. The railroad
lines would decrease food shortages and prevent future famines in the respective
places (Quataert 1996, 789; Erler 2010, 304-12).
To address the reverse causality issue, this paper pursues two different
strategies: (I) an instrumental variable for railroads based on trade and caravan routes
is constructed; and (II) this paper examines the effect of un-built lines on the
population in the proposals of Colonel Chesney and Levant Herald.8
The second stage result, using the instrument (that is, the logarithm of the
shortest distance between each judicial district and the trade and caravan routes) to
predict the distance to the railroads, is presented in column (2) of Table 1. The IV
estimate for the effect of the railroads is not statistically different from zero.
Furthermore, the IV estimate is larger than the OLS coefficient in magnitude, which
suggests that the OLS estimate is biased downward. The coefficient estimate for the
impact of the distance to the trade routes in column (3) is statistically different from
zero at 1 per cent. However, the IV estimate is not precise enough to provide
8
Before 1870, Colonel Chesney and Levant Herald proposed the construction of several lines that
would have been located in Syria, Iraq, and Anatolia. The positioning of the proposed lines was
related to the population in a similar way to how the railroad lines were actually built.
Furthermore, events that were less likely to be related to economic conditions in the judicial
districts (such as financial difficulties faced by the Ottoman state, which could have arisen due to
wars and its debt) were mainly responsible for the fact that none of the proposed lines attracted
considerable interest. 4 information on railroad line placement decisions, as the F-statistic in column (3) is
not larger than 10.
To further address reverse causality concerns, in column (4), this paper
examines the effect of lines that were planned but that were not constructed. Column
(4) reports that the coefficient estimate for the impact of the distance to the railroads
is negative and statistically different from zero, as before. In addition, column (4)
reports that the coefficient estimates for the effects of Chesney`s proposal and Levant
Herald`s proposal are not statistically significant. These results imply that reverse
causality is not an important concern. This could be due to the fact that the railroads
in the Ottoman Empire were primarily built for political, religious, or military rather
than economic reasons. This is also in line with Jedwab & Moradi (2012), who
examine the impact of railroads on cocoa production and population growth in
Ghana.
IV. Conclusion
The findings show that railroad access induced economic growth, which led to
expansion in the affected areas. Railroads can be an important driver of economic
growth in places where road networks are poor before railroads are built, as those of
the Ottoman Empire were, even if economic factors do not play an important role in
the construction of railroads.
This paper contributes to previous literature in several respects. It is the first
study that uses a unique historical dataset of railroads and judicial district population
figures to analyse the impact of railroads in the Ottoman Empire. It provides
empirical evidence to help inform the historical literature as to whether access to
railroads induced economic growth in the Ottoman Empire. It is the first study that
uses a unique historical dataset on trade and caravan routes in the Ottoman Empire as
well as proposals by Colonel Chesney and Levant Herald in order to deal with the
endogeneity issue of where railroad lines in the Ottoman Empire were built; that is, it
represents the first attempt to study decisions about the placement of railroad lines
empirically. Finally, it extends the empirical literature on the impact of railroads by
presenting empirical evidence from history, using a unique historical dataset on the
Ottoman Empire, whose borders included many present-day countries.9
Bibliography
British Parliamentary Paper (1896). Report by Major Law on Railways in Asiatic
Turkey. 19th Century House of Commons Sessional Paper no. 8019: 1-21.
Eldem, V. (1994). Osmanlɪ İmparatorluğu’nun İktisadi Şartlarɪ Hakkɪnda Bir Tetkik.
İstanbul: Türk Tarih Kurumu Yayınlarɪ.
Erler, M. Y. (2010). Osmanlɪ Devleti`nde Kuraklɪk ve Kɪtlɪk Olaylarɪ (1800-1880).
İstanbul: Libra Yayinevi.
Geyikdağɪ, V.N. (2011). Foreign Investment in the Ottoman Empire: International
Trade and Relations 1854-1914. London: Tauris Academic Studies.
Gülsoy, U. (1994). Hicaz Demiryolu. Istanbul: ErenYayinlarɪ.
Gülsoy, U. (2010). Kutsal Proje – Ortadoğu` da Osmanlɪ Demiryollarɪ. İstanbul:
Timaş Yayinlarɪ.
Hornung, E. (2012). ‘Railroads and Micro-regional Growth in Prussia’. Competitive
Advantage in the Global Economy (CAGE) Working Paper no. 79.
Hülagü, M. M. (2010). The Hejaz Railway: the Construction of a New Hope. New
York: Bluedome Press.
9
These countries are Turkey, Greece, Syria, Iraq, Lebanon, Israel, Palestine, Jordan, and Yemen, and
some parts of Albania, Macedonia, Kosovo, and Saudi Arabia.
5 Illich, N. S. (2007). ‘German Imperialism in the Ottoman Empire: a Comparative
Study’. Unpublished PhD Thesis. Texas: Texas A&M University.
Jedwab, R. & A. Moradi (2012). ‘Revolutionizing Transport: Modern Infrastructure,
Agriculture and Development in Ghana’. LSE Open Access Publications.
Karkar, Y. N. (1972). Railway Development in the Ottoman Empire, 1856-1914. New
York: Vantage Press.
Karpat, K. H. (1985). Ottoman Population 1830-1914: Demographic and Social
Characteristics. Wisconsin: The University of Wisconsin Press.
Kolars, J. & H. J. Malin (1970). ‘Population and Accessibility: An Analysis of
Turkish Railroads’. The Geographical Review. 60: 229-46.
McCarthy, J. (2001). The Ottoman Peoples and the End of Empire. New York:
Oxford University Press.
Ortaylɪ, I. (2010). Osmanlɪ İmparatorluğu’nda Alman Nüfuzu. Istanbul: Timaş
Yayinlarɪ.
Quataert, D. (1977). ‘Limited Revolution: The Impact of the Anatolian Railway on
Turkish Transportation and the Provisioning of Istanbul, 1890-1908’. The Business
History Review. 51(2):139-60.
Quataert, D. (1996). ‘The age of reforms, 1812–1914’. in İnalcık, H. and Quataert, D.
eds., An Economic and Social History of the Ottoman Empire, 1300-1914.
Cambridge, Cambridge University Press: 749-943.
Schoenberg, P.E. (1977). ‘The Evolution of Transport in Turkey (Eastern Thrace and
Asia Minor) under Ottoman Rule, 1856-1918’. Middle Eastern Studies. 13: 359-72.
Su, K. (1939). Balıkesir Madenleri. Istanbul: Resimli Ay Matbaasɪ.
Von Pressel, W. (1966). ‘Railway Projects in Turkey, 1872-1900’. in Issawi, C. ed.,
The Economic History of the Middle East 1800-1914. Chicago, University of Chicago
Press: 90-4.
6