Flowers of Evil Industrialization and Long Run Development - Raphael Franck

Flowers of Evil?
Industrialization and Long Run Development
Journal Pre-proof
Flowers of Evil? Industrialization and Long Run Development
Raphaël Franck, Oded Galor
PII: S0304-3932(19)30214-4
DOI: https://doi.org/10.1016/j.jmoneco.2019.12.001
Reference: MONEC 3203
To appear in: Journal of Monetary Economics
Received date: 25 April 2019

Revised date: 29 November 2019
Accepted date: 2 December 2019
Please cite this article as: Raphaël Franck, Oded Galor, Flowers of Evil? Industri-
alization and Long Run Development, Journal of Monetary Economics (2019), doi:
https://doi.org/10.1016/j.jmoneco.2019.12.001
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
© 2019 Published by Elsevier B.V.

1
2 Flowers of Evil? Industrialization and Long Run Development✩
3 Raphaël Francka,∗, Oded Galorb

a
4 The Hebrew University of Jerusalem, Department of Economics, Mount Scopus, Jerusalem 91905,
5 Israel. Raphael.Franck@mail.huji.ac.il
b
6 Brown University, NBER, CEPR, IZA, and CESifo. Oded Galor@brown.edu.
7 Abstract
Is industrialization conducive for economic development in the 21st century? Exploit-
ing exogenous variation in the diffusion of steam engines across 19th century France, the
research suggests that early industrialization has had an adverse effect on long-run pros-
perity, stemming from the negative impact of the adoption of unskilled-labor-intensive
technologies in early stages of industrialization on contemporary levels of human capital
and thus the incentive to adopt skill-intensive technologies. The research thus suggests
that characteristics that enabled the onset of industrialization, rather than industrial
technology per se, have been the source of prosperity among developed economies that
experienced earlier industrialization.
8 Keywords: Economic Growth, Human Capital, Industrialization, Steam Engine,
9 Cultural Inertia.
10 JEL classification: N33, N34, O14, O33.
✩
The authors are grateful to the editor (Yuriy Gorodnichenko), Daron Acemoglu, Philippe Aghion,
Josh Angrist, Emily Blanchard, Francesco Caselli, Eve Colson-Sihra, Martin Fiszbein, Marc Klemp,
Tommaso Porzio, Jesse Shapiro, Uwe Sunde and David Weil for helpful discussions and participants in
seminars and conferences at Ben Gurion University, Brown, Clemson, Haifa, Hebrew University, MIT,
UC Merced, the Israel Economic Association, the NBER Meeting of Macroeconomics Across Time and
Space, May 2018, and the NBER SI Economic Growth Meeting, July 2018, for useful comments. We
thank Guillaume Daudin, Alan Fernihough and Ömer Özak for sharing their data with us. Raphaël
Franck wrote part of this paper as Marie Curie Fellow at the Department of Economics at Brown
University under funding from the People Programme (Marie Curie Actions) of the European Union’s
Seventh Framework Programme (FP 2007-2013) under REA Grant agreement PIOF-GA-2012-327760
(TCDOFT).
∗
Corresponding author
Preprint submitted to Elsevier December 10, 2019
11 Highlights
12 • Industrialization is not conducive for economic development in the 21st century
13 • Industrialization led to adoption and persistence of unskilled-labor-intensive tech-
14 nologies
15 • Industrialization is detrimental to human capital formation in the long-run
16 • Industrialization reduced the incentive to adopt skill-intensive technologies
17 • Roots of wealth: factors that enabled industrialization, not industrialization per se
2
18 1. Introduction
19 The process of development has been marked by persistence as well as reversals

20 in the relative wealth of nations. While some geographical characteristics that were
21 conducive for economic development in the agricultural stage had detrimental effects on
22 the transition to the industrial stage of development, conventional wisdom suggests that
23 prosperity has persisted among societies that experienced an earlier industrialization.1
24 Regional development within advanced economies, nevertheless, appears far from
25 being indicative of the presence of a persistent beneficial effect of early industrialization.
26 In particular, anecdotal evidence suggests that regions which were prosperous industrial
27 centers in Western Europe and in the Americas in the 19th century (e.g., the Rust Belt
28 in the USA, the Midlands in the UK, and the Ruhr valley in Germany) have experienced
29 a decline in their comparative development in recent decades.
30 These conflicting observations about the long-run effect of industrialization on the
31 prosperity of regions and nations may suggest that factors which fostered industrial de-
32 velopment in the Western world, rather than the forces of industrialization per se, are
33 associated with the persistence of fortune across these industrial societies. In particular, it
34 is not inconceivable that the process of industrialization per se, despite its earlier virtues,
35 has had detrimental effects on the transition to the post-industrial stage of development
36 and on contemporary prosperity. Nevertheless, despite the enormous importance of the
37 resolution of this question from a policy perspective, to a large extent, this issue has,
38 neither been raised nor been explored in the modern economic growth literature.
39 This research explores the long-run implications of industrialization on the process
40 of development. It addresses two fundamental questions: (i) is industrialization con-
41 ducive for long-run prosperity? and (ii) are the industrialized nations richer because of
42 industrialization or perhaps despite industrialization? Conventional wisdom views in-
43 dustrial development as a catalyst for economic growth, highlighting its persistent effect
44 on economic prosperity. In contrast, the study advances the hypothesis and establishes
45 empirically that while the adoption of industrial technology was conducive for economic
46 development in the short-run, acquired comparative advantage in the unskilled-labor-
47 intensive industrial sector had triggered a techno-cultural inertia, characterized by a lower
48 predisposition towards investment in human capital, that has hindered the transition to
49 more lucrative skilled-intensive sectors, adversely effecting human capital formation and
50 standards of living.2
1
The persistence effect of geographical, cultural, institutional and human characteristics have been
at the center of a debate regarding the origins of the differential timing of transitions from stagnation
to growth and the remarkable transformation of the world income distribution in the last two centuries
(e.g., Acemoglu et al. ((2001)), Galor ((2011)), Andersen et al. ((2016)), Ashraf and Galor ((2013)),
Cervellati and Sunde ((2005)), Dalgaard and Strulik ((2016)), Galor and Özak ((2016)), Litina ((2016)),
Mokyr ((2016))). For an illustration, see Appendix Figure A.1,
2
While the decline in manufacturing employment and the rise in the service sector in the process of
development will bring about a temporary decline of industrial-intensive regions, human capital formation
will permit the transition to skill-intensive sectors as long as the labor force is sufficiently flexible and
economic development would not have an adverse effect on these regions in the long-run. In contrast, the
techno-cultural lock-in effects highlighted in the current study would prevent the adjustment in the skill
composition of the labor force, would diminish the incentive of entrepreneurs to adopt skill-intensive
technologies, and would result in a long-lasting adverse effect of early industrialization on long-run
development.
3
51 A conclusive exploration of the impact of industrialization on long-run prosperity
52 ought to overcome significant empirical hurdles. First, the observed relationship between
53 industrialization and the development process may reflect the reverse causality from the
54 process of development to industrialization rather than the effect of industrialization on
55 the process of development. Second, the effect of institutional, cultural, geographical
56 and human characteristics on the joint evolution of industrialization and the process of
57 development may have governed the observed relationship between industrialization and
58 the development process. Third, the time since industrialization in a large number of
59 regions and countries is shorter than needed to assess the potential adverse effects of
60 industrialization on long-run prosperity.
61 In light of these empirical hurdles, the desirable empirical framework will be an
62 economy in which: (i) the territory has been divided into administrative units where in-
63 stitutional, cultural, human and geographical characteristics are unlikely to differ signif-
64 icantly, (ii) the creation of administrative units preceded the process of industrialization
65 and is orthogonal to the subsequent process of industrialization, (iii) industrialization
66 has occurred sufficiently early so as to permit the exploration of its potential adverse
67 long-run effects, (iv) exogenous source of regional variation in the intensity of industri-
68 alization could be identified, and (v) extensive data on the process of development since
69 early industrialization is available.
70 The economy of France appears ideally suited for this empirical exploration for these
71 reasons. First, as early as 1790, the French territory was divided into administrative
72 units (departments) of nearly equal size, designed to ensure that travel distance by horse
73 from any location within the department to its main administrative center would not
74 exceed one day. Hence, one can plausibly argue that the borders of each department
75 were orthogonal to the process of industrialization. Second, French departments have
76 been subjected to an intensive institutional and cultural unification that mitigated initial
77 cultural differences across these regions. Third, France was one of the first European
78 countries to industrialize and the extended period since its industrialization is sufficiently
79 long to permit the detection of the potential adverse effect of industrialization on long-
80 run prosperity. Fourth, exogenous sources of variation in the intensity of industrialization
81 across department could be detected. Finally, the availability of extensive data on income
82 per capita, human capital formation, wages, sectoral employment, unionization rates,
83 tariff protection, economic integration and the availability of natural resources across
84 departments permits the examination of the proposed channels through which the adverse
85 effect of industrialization may have operated.
86 The study utilizes French regional data from the second half of the 19th century
87 until the beginning of the 21st century to explore the impact of the adoption of industrial
88 technology on the evolution of income per capita. It establishes that regions which
89 industrialized more intensively experienced higher income per capita in the subsequent
90 decades. Nevertheless, industrialization has had an adverse effect on income per capita
91 by the turn of the 21st century.
92 In light of the association between industrialization and the intensity of the use of
93 the steam engine ((Mokyr, 1990, Bresnahan and Trajtenberg, 1995, Rosenberg and Tra-
94 jtenberg, 2004)), the study takes advantage of historical evidence regarding the regional
95 diffusion of the steam engine ((Ballot, 1923, Sée, 1925, Léon, 1976)) to identify the effect
96 of regional variations in the intensity of the use steam engine in 1860-1865 on the process
4
97 of development. In particular, it exploits the distances of each French department from
98 Fresnes-sur-Escaut, where a steam engine was first successfully operated for commer-
99 cial use from 1732 onwards, as exogenous source of variations in industrialization across
100 French regions.3
101 Indeed, in line with the historical account, the unequal distribution of steam engines
102 across French departments is indicative of a local diffusion process from Fresnes-sur-
103 Escaut. Accounting for confounding geographical and institutional characteristics, pre-
104 industrial development as well as distances from major economic centers, if the distance of
105 a department away from Fresnes-sur-Escaut was to increase from the 40th (426 km) to the
106 60th percentile (559 km) of the distance distribution, this department would experience
107 a drop of 275 horse power of steam engines (relative to a sample mean of 1839 hp) in the
108 1860-1865 industrial survey.
109 The validity of the distance from Fresnes-sur-Escaut as an instrumental variable for
110 the intensity of the adoption of steam engines across France is enhanced by three addi-
111 tional factors. First, conditional on the distance between each department and Fresnes-
112 sur-Escaut, distances from major centers of economic power in 1860-1865 (e.g., Paris,
113 Marseille, Lyon, Rouen, Bordeaux and London) are uncorrelated with the intensive use
114 of the steam engine over this period. Second, the distance from Fresnes-sur-Escaut is un-
115 correlated with the level and the growth rate of economic development across France in
116 the pre-industrial period. Third, it appears that the Nord department (where Fresnes-sur-
117 Escaut is located) had neither superior human capital characteristics nor higher standard
118 of living in comparison to the average department in France.
119 The study establishes that the horse power of steam engines in industrial production
120 in the 1860-1865 period had a positive and significant impact on income per capita in
121 1860, 1901 and 1930. In particular, a one-percent increase in the total horse power of
122 steam engines in a department in 1860-1865 increased GDP per capita by 0.12 percent
123 in 1860, 0.25 percent in 1901 and 0.10 percent in 1930. Nevertheless, industrialization
124 had an adverse effect on income per capita and human capital formation in the post-2000
125 period. In particular, a one-percent increase in the total horse power of steam engines in a
126 department in 1860-1865 led to a 0.08 percent decrease in GDP per capita in 2001-2005.4
127 These results are unlikely to be driven by selection on unobservables and are robust to
128 accounting for spatial autocorrelation.
129 It is important to note that the IV estimate reverses the OLS estimates of the
130 relationship between industrialization and the level of income per capita in 2001-2005,
131 from positive to negative. This reversal suggests that factors which fostered industrial
132 development, rather than industrialization per se, contributed to the positive association
3
In 1726, an Englishman named John May obtained a privilege to operate steam engines to pump
water throughout the French kingdom. Jointly with another Englishman named John Meeres, he installed
the first steam engine in Passy (which was then outside but is now within the administrative boundaries
of Paris) to raise water from the Seine river to supply the French capital with water. However it seems
that their commercial and industrial operation stopped quickly or even never took off. Indeed, when
Forest de Bélidor ((1737)) published his massive treatise on engineering in 1737-1739, he mentioned that
the steam engine in Fresnes-sur-Escaut was the only one operated in France (see, e.g., Lord ((1923)) and
Dickinson ((1939))).
4
To put these figures in perspective, it must be borne in mind that Crafts ((2004)) finds that the
contribution of steam technology to labor productivity growth in Great Britain was equal to 0.41 percent
per year over the 1850-1870 period and to 0.31 percent per year over the 1870-1910 period.
5
133 between industrialization and long-run development. In particular, once one accounts for
134 the effect of these omitted factors, industrialization has an adverse effect on the standard
135 of living in the long-run. These findings suggest that the characteristics that enabled
136 the early onset of industrialization, rather than the adoption of industrial technology per
137 se, have been the source of prosperity among the currently developed economies that
138 experienced an early industrialization.
139 The empirical analysis accounts for a wide range of exogenous confounding geograph-
140 ical and institutional characteristics, as well as for pre-industrial development, which may
141 have contributed to the relationship between industrialization and economic development.
142 First, it accounts for the potentially confounding impact of exogenous geographical char-
143 acteristics (i.e, latitude, land suitability, average temperature, average rainfall and share
144 of carboniferous area) of each French department on the relationship between industrial-
145 ization and economic development. In particular, it captures the potential effect of these
146 geographical factors on the profitability of the adoption of the steam engine, the pace
147 of its regional diffusion, as well as on productivity and thus the evolution of income per
148 capita in the process of development. Second, it captures the potentially confounding
149 effects of the location of departments (i.e., border departments, maritime departments,
150 departments at a greater distance from the concentration of political power in Paris, and
151 those that were temporarily under German domination) on the diffusion of the steam
152 engine and the diffusion of development. Third, the analysis accounts for the differential
153 level of development across France in the pre-industrial era that may have affected jointly
154 the process of development and the process of industrialization. Finally, the analysis is
155 shown to be robust to variables which could be viewed as confounding factors and are
156 thus not part of the baseline analysis, i.e., the initial level of income per capita, initial
157 population density, initial levels of human capital (e.g., the number of Encyclopédie sub-
158 scribers ((Squicciarini and Voigtländer, 2015))), the share of Jews and Protestants in
159 the population, the early use of raw material, economic integration before and after the
160 French Revolution ((Daudin, 2010)), the Napoleonic blockade ((Juhasz, 2018)), as well
161 as industrial concentration and firm size.
162 If one views each French department as a small open economy, one may argue that
163 the proper industrial policy ought to encourage the development of skilled-intensive sec-
164 tors rather than of traditional unskilled-labor-intensive sectors. However, one concern
165 could be that the negative impact of industrialization in the 21st century, at the depart-
166 mental level, does not reflect the overall effect of industrialization. A priori, it is possible
167 that industrialization generated technological spillovers such that the most industrialized
168 department within a region declined but the region as a whole prospered due to the
169 spillovers from the process of industrialization. Nevertheless, further empirical analysis
170 suggests that the negative impact of industrialization on long-run prosperity in one de-
171 partment did not generate sufficiently positive spillovers in neighboring departments so
172 as to avert the adverse effects of industrialization on long-run prosperity in the region as
173 a whole.
174 The research further explores the mediating channels through which earlier industrial
175 development has an adverse effect on the contemporary level of development. It suggests
176 that the adverse effect of industrialization on long-run prosperity reflects the negative
177 impact of earlier specialization in unskilled-labor-intensive industries on human capital
178 formation and the incentive to adopt skill-intensive technologies in the contemporary
179 era. Industrialization has triggered a dual techno-cultural lock-in effect characterized
6
180 by a reinforcing interaction between technological inertia, reflected by the persistence
181 predominance of low-skilled-intensive industries, and cultural inertia, in the form of a
182 lower predisposition towards investment in human capital. In particular, while the adop-
183 tion of industrial technologies was conducive for economic development in the short-run,
184 acquired comparative advantage in the unskilled-labor-intensive industrial sector had
185 triggered cultural inertia, characterized by lower educational aspirations, that has hin-
186 dered the transition to more lucrative skilled-intensive sectors, adversely effecting human
187 capital formation and the standards of living by the end of the 20th century.
188 The dual technological-cultural lock-in effect is established using individual level
189 data. Following the epidemiological approach for the study of cultural persistence, the
190 study exploits data on education achievements of over 2100 second generation migrants
191 (i.e., individuals who live in their birth department but whose parents migrated from a
192 different department within France). The analysis suggests that second generation mi-
193 grants whose parents were originated in historically industrial regions are significantly
194 more likely to have low human capital aspirations, as reflected by their acquisition of vo-
195 cational education, accounting for the department of birth fixed effects and for parental
196 occupation. The analysis of second-generation migrants accounts for time invariant un-
197 observed heterogeneity in the host department (e.g., geographical, cultural and institu-
198 tional characteristics), mitigating possible concerns about the confounding effect of host
199 department-specific characteristics. Moreover, since the historical industrial intensity in
200 the parental department of origin is distinct from that in the respondent’s department
201 of residence, the estimated effect of industrial intensity in the parental department of
202 origin on human capital formation captures the culturally-embodied, intergenerationally
203 transmitted effect of industrial intensity on human capital aspirations, rather than the
204 direct effect of industrial intensity. This result lend credence to the presence of cultural
205 inertia and technological inertia that have reinforced one another and triggered a dual
206 lock-in effect.
207 Furthermore, using individual data on the composition of employment across sec-
208 tors among over 1.1 million individuals, the study suggests that this cultural inertia, and
209 its adverse effect on human capital formation in the long-run, has further hindered the
210 incentive of competitive industries to adopt more lucrative skilled-intensive technologies,
211 reinforcing the suboptimal level of human capital formation and reducing the standards
212 of living in the long-run. In particular, individuals who are currently residing in a depart-
213 ment that was characterized in the 1860-1865 survey by a higher horse power of steam
214 engine are significantly more likely to be employed in unskilled-labor-intensive industrial
215 sectors. These results lend credence to the argument that historical industrial regions
216 have experienced a technological lock-in effect. Namely, acquired comparative advantage
217 in the unskilled-labor-intensive sector in early stages of industrialization is associated
218 with the relative domination of unskilled-labor-intensive firms and occupations.
219 The empirical analysis further establishes that various plausible channels do not ac-
220 count for the adverse effect of early industrialization on long-run prosperity: (1) the con-
221 tribution of industrialization to unionization and wage rates in historically industrialized
222 regions and the comparative decline of these regions in the long-run due to the incentive of
223 modern industries to locate in regions where labor markets are more competitive, (2) the
224 effect of trade protection, competition from China, and modern environmental regulation
225 in traditional industries, on the decline in the long-run competitiveness of historically
226 industrialized regions, (3) the potential negative effect of disproportional destruction of
7
227 industrialized regions during WWI and WWII on the subsequent development of these
228 regions, (4) the persistent adverse effect of selective migration (e.g. immigration of un-
229 skilled workers into industrialized regions, or the emigration of more educated workers
230 into less industrialized regions), on the composition of human capital and long-run in-
231 come per capita in historically industrialized regions, and (5) the disproportionate public
232 investment in human capital in non-industrial regions.
233 2. Data and Main Variables
234 France was among the first countries to industrialize in Europe in the 18th century
235 and its industrialization continued during the 19th century. Nevertheless, by 1914, the
236 living standard in France remained below that of England and of Germany, which had
237 become the leading industrial country in continental Europe. The slower path of indus-
238 trialization in France has been attributed to the consequences of the French Revolution
239 (e.g., wars, legal reforms and land redistribution), the patterns of domestic and foreign
240 investment, cultural preferences for public services, as well as the comparative advan-
241 tage of France in agriculture vis-a-vis England and Germany ((see the discussion in, e.g.,
242 Crouzet, 2003)).
243 This section examines the evolution of industrialization and income across 89 French
244 departments, based on the administrative division of France in the 1860-1865 period,
245 accounting for the geographical and the institutional characteristics of these regions. The
246 initial partition of the French territory in 1790 was designed to ensure that the travel
247 distance by horse from any location within the department to the main administrative
248 center would not exceed one day. The initial territory of each department was therefore
249 orthogonal to the process of development and the subsequent minor changes in the borders
250 of some departments did not reflect the effect of industrialization.
251 In light of the changes in the internal and external boundaries of the French territory
252 during the period of study, the number of departments that is included in different stages
253 of the analysis varies from 81 to 89. In particular, several departments that were split
254 into smaller units are aggregated into their historical territorial borders and regions that
255 were temporarily removed from the French territory are excluded from the analysis during
256 those time periods.5 Table A.1 reports the descriptive statistics for the variables in the
257 empirical analysis across these departments.
258 2.1. Past and Present Measures of Income, Workforce and Human Capital
259 2.1.1. Income
260 This study seeks to examine the effect of industrialization on the evolution of income
261 per capita in the process of development. Given that the industrial survey which is
5
The Parisian region encompassed three departments (Seine, Seine-et-Marne and Seine-et-Oise) be-
fore 1968 and it was split into eight (Essonne, Hauts-de-Seine, Paris, Seine-et-Marne, Seine-Saint-Denis,
Val-de-Marne, Val d’Oise and Yvelines) afterwards. Likewise, the Corsica department was split in 1975
into Corse-du-Sud and Haute-Corse. The three departments (i.e., Bas-Rhin, Haut-Rhin and Meurthe)
which were under German rule between 1871 and 1918 are excluded from the analysis of economic de-
velopment over that time period. In addition, in the examination of the robustness of the analysis with
data prior to 1860, the three departments (i.e., Alpes-Maritimes, Haute-Savoie and Savoie) that were
not part of France are excluded from the analysis.
8
262 the basis for our analysis was conducted between 1860 and 1865, the relevant data to
263 capture the short-run and medium-run effects of industrialization on income per capita
264 are provided at the departmental level prior to WWII for the years 1860, 1872, 1886,
265 1901, 1911 and 1930 by Combes et al. ((2011)) and Caruana-Galizia ((2013)). Thus, for
266 the sake of brevity, and equal spacing between those years, the analysis focuses on income
267 per capita in 1860, 1901 and 1930.
268 To assess the effects of industrialization on income per capita in the long-run, the
269 analysis is restricted to the 2001-2005 period (INSEE - Institut National de la Statistique
270 et des Etudes Economiques).6 Moreover, to lessen the potential impact of fluctuations
271 in income per capita, the effect of industrialization in the long-run is captured by its
272 differential impact on the average GDP per capita across departments over the 2001-
273 2005 period.
274 2.1.2. Workforce

275 The effect of industrialization on the sectoral composition of the workforce in the
276 post-1860 period is captured by the impact on the shares of employment in the agri-
277 cultural, industrial and service sectors. The surveys which capture the short-run and
278 mid-run effects of industrialization are those undertaken in 1861, 1901 and 1930 (Statis-
279 tique Générale de la France). Similarly, to assess the effects of industrialization on the
280 sectoral composition in the post-WWII period, all available surveys of the French popu-
281 lation across departments (i.e., 1968, 1975, 1982, 1990, 1999 and 2010) are used (INSEE
282 - Institut National de la Statistique et des Etudes Economiques).
283 Furthermore, the analysis of the underlying mechanism uses individual data from
284 the Déclaration Annuelle de Donnés Sociales in 2008 which provides representative infor-
285 mation on 1.1 million private sector workers, except for the self-employed. This dataset
286 enables us to determine whether individuals are more likely to work in firms where the
287 demand for human capital is high (i.e., scientific research & development) or low (i.e.,
288 coal industries or machine repair.) In addition, the analysis relies on the 2005 survey
289 Enquête Emploi conducted by the INSEE regarding the job prospects of employed and
290 unemployed individuals. This survey provides information on the respondents’ birthplace
291 as well as those of their parents. As such, it enables us to focus on second generation
292 migrants (i.e., individuals who were born and still live in their birth department, but
293 whose parents were born in a different department). In the analysis, these second gener-
294 ation migrants are matched to the horse power of steam engines of their mother’s birth
295 department, their father’s birth department, as well as to the birth department of their
296 parents if both of them were born in the same department.
297 2.1.3. Human Capital

298 The study further explores the effect of industrialization on the evolution of human
299 capital in the process of development. The effect of industrialization on human capital
300 formation in the pre-WWI period is captured by its impact on the literacy rates of French
301 army conscripts (i.e., 20-year-old men who reported for military service in the department
6
Data on income per capita at the departmental level is only available in the post-1995 period and
the corresponding data for the other indicators of the standards of living only in the post-2001 period.
Note that the qualitative results remain unchanged if one considers the average income per capita over
the entire sample period available, 1995-2010.
9
302 where their father lived - France ((1878-1939))). In particular, given the data limitations,
303 the analysis focuses on the share of the literate conscripts over the 1874-1883 and 1894-
304 1903 decades. As reported in Table A.1, 82.0% of the French conscripts were literate over
305 the 1874-1883 period and 94.1% over the 1894-1903 period.7
306 The effect of industrialization on human capital formation in the post-WWII period
307 is captured by its impact on the share of men and women (age 25 and above) who
308 completed a post-secondary degree (in a vocational school or in an university) as reported
309 in the available surveys of the French population across departments (i.e., 1968, 1975,
310 1982, 1990, 1999 and 2010). As can be seen in Table A.1, there was a continuous increase
311 in the educational achievements of the French population during this period. Indeed the
312 shares of men and women (age 25 and above) who completed high-school, respectively,
313 rose from 8.8% and 6.0% in 1968 to 36.3% and 39.1% in 2010.
314 Furthermore, to examine the role of the composition of human capital in the non-
315 monotonic evolution of income per capita, the study explores the impact of industrial-
316 ization on the evolution of high-, medium- and low-levels of human capital in France
317 after WWII . This composition is captured by the division of the workforce (age 25-54)
318 between executives and other intellectual professions, middle management professionals,
319 and employees, in the available surveys of the French population across departments
320 (1968, 1975, 1982, 1990, 1999 and 2010).
321 Moreover, to capture the effect of industrialization on human capital formation in
322 the contemporary period, in which school attendance is mandatory until the age of 16,
323 the study explores its impact on the shares of men and women in the 15-17 and 18-24 age
324 categories attending school or any other (post-secondary) learning institution as reported
325 in the 2010 census. As indicated in Table A.1, in 2010, most men and women age 15-17
326 (respectively 95.5% and 96.7%) attended school but fewer (44.3% and 48.0%) pursued
327 post-secondary studies.
328 Finally, to capture the general willingness of the local population to invest in human
329 capital, we use individual data from a 2001 survey pertaining to the importance that
330 individuals attribute to science and scientific research (Centre de recherches politiques
331 de Sciences Po, Enquête science 2001). Out of the various questions, we single out two
332 where individuals are asked whether they have an interest in science or not, and whether
333 they use science in their current work.
334 2.2. Steam Engines

335 The research explores the effect of the introduction of industrial technology on the
336 process of development. In light of the pivotal role played by the steam engine in the
337 process of industrialization, it exploits variations in the industrial use of the steam engine
338 across the French regions during its early stages of industrialization to capture the inten-
339 sity of industrialization. In particular, the analysis focuses on the horse power of steam
340 engines used in each French department as reported in the industrial survey carried out
341 by the French government between 1860 and 1865.8
7
In line with the historical evidence ((e.g., Grew and Harrigan, 1991)), as reported in Table A.1, a
sizeable share of the French population was literate even before the passing of the 1881-1882 laws which
made primary school attendance “free”and mandatory for boys and girls until age 13.
8
The 1860-1865 survey is the second industrial survey undertaken in France which was published
by the French government: it provides the horse power of steam engines but not the number of steam
10
342 [Insert Figure 1]
343 As depicted in Figure 1, and analyzed further in the discussion of the identification
344 strategy in Section 3, the unequal distribution of the steam engines across French depart-
345 ments in 1860-1865 suggests a regional pattern of diffusion from Fresnes-sur-Escaut (in
346 the Nord department, at the northern tip of continental France) where a steam engine
347 was first successfully operated for commercial and industrial purposes in France from
348 1732 onwards. The most intensive use of the steam engine over this period was in the
349 Northern part of France. The intensity diminished somewhat in the East and in the
350 South East, and declined further in the South West. Three departments had no steam
351 engine in 1860-1865 (i.e., Ariège and Lot in the South-West and Hautes-Alpes in the
352 South-East). Potential anomalies associated with these departments are accounted for
353 by the introduction of a dummy variable that represents them. In particular, potential
354 concerns about the distance of these departments from the threshold level of development
355 that permits the adoption of the steam engines is accounted for by this dummy variable.
356 Table A.6 reports descriptive statistics for the horse power of steam engines in each of
357 the 16 sectors listed in the 1860-1865 survey: ceramics, chemistry, clothing, construction,
358 food, furniture, leather, lighting, luxury goods, metal objects, metallurgy, mines, sciences
359 & arts, textile, transportation and wood. It shows that the five sectors with the largest
360 mean horse power per department are textile, metallurgy, mines, food industry and metal
361 objects. In particular, the textile sector had the largest average horse power of all the
362 sectors and 43% more horse power than metallurgy, the sector with the second largest
363 mean horse power. Moreover, using the descriptive statistics on the number of workers
364 in each of the 16 sectors reported in Table A.6 that the textile sector has a smaller ratio
365 of steam engine horse power per worker than the metallurgy, mining and food sectors,
366 most likely because not all the activities of the textile sector required steam engines.
367 2.3. Confounding Characteristics of each Department

368 The empirical analysis accounts for a wide range of exogenous confounding geograph-
369 ical and institutional characteristics, as well as for pre-industrial development, which may
370 have contributed to the relationship between industrialization and economic development.
371 Institutions may have affected jointly the process of development and the process of indus-
372 trialization. Geographical characteristics may have impacted the pace of industrialization
373 as well as agricultural productivity and thus income per capita. Moreover, geographi-
374 cal and institutional factors may have affected the process of development indirectly by
375 governing the pace of the diffusion of steam engines across departments. Finally, pre-
376 industrial development may have affected the onset of industrialization and may have
377 had an independent persistent effect on the process of development.
378 2.3.1. Geographic Characteristics

379 The empirical analysis accounts for the potentially confounding impact of exogenous
380 geographical characteristics of each of the French departments on the relationship between
381 industrialization and economic development. In particular, it captures the potential effect
engines. Conversely, the first industrial survey, which was carried out in 1839-1847, indicates the number
of steam engines but not the horse power of the steam engines. In the working paper of this article
((Franck and Galor, 2017a)), we establish the robustness of the results to using the 1839-1847 data, as
well data from 1897.
11
382 of these geographical factors on the profitability of the adoption of the steam engine, the
383 pace of its regional diffusion, as well as on productivity and thus the evolution of income
384 per capita in the process of development.
385 First, the study accounts for climatic and soil characteristics of each department
386 mapped in Figure A.2 (i.e., land suitability, average temperature, average rainfall, and
387 latitude ((Ramankutty et al., 2002, Luterbacher et al., 2004, 2006, Pauling et al., 2006))),
388 that could have affected natural land productivity and therefore the feasibility and prof-
389 itability of the transition to the industrial stage of development, as well as the evolution
390 of aggregate productivity in each department. Moreover, the diffusion of the steam en-
391 gine across French departments as well as the process of development could have been
392 affected by the presence of raw material required for industrialization. Our regressions
393 thus account for the share of carboniferous area in each department ((Fernihough and
394 O’Rourke, 2014)).
395 Second, the analysis captures the confounding effect of the location of each depart-
396 ment on the diffusion of development from nearby regions or countries, as well as its
397 effect on the regional diffusion of the steam engine. In particular, it accounts for the
398 effect of the latitude of each department, border departments (i.e., positioned along the
399 border with Belgium, Luxembourg, Germany, Switzerland, Italy and Spain), and mar-
400 itime departments (i.e., positioned along the sea shore of France) on the pace of this
401 diffusion process. It also accounts for the presence of rivers and their main tributaries
402 within the perimeter of the department by using data on the paths of the Rhine, Loire,
403 Meuse, Rhône, Seine and Garonne rivers as well as of their major tributaries (Dordogne,
404 Charente and Escaut).
405 Finally, the research accounts for the potential differential effects of international
406 trade on process of development as well as on the adoption the steam engine. In particular,
407 it captures the potential effect of maritime departments (i.e., those departments that are
408 positioned along the sea shore of France), via trade, on the diffusion of the steam engine
409 and thus on economic development as well as the effect of trade on the evolution of income
410 per capita over this time period.
411 2.3.2. Institutional Characteristics

412 The analysis deals with the effect of variations in the adoption of the steam engine
413 across French departments on their comparative development. This empirical strategy
414 ensures that institutional factors that were unique to France as a whole over this time
415 period are not the source of the differential pattern of development across these regions.
416 Nevertheless, two regions of France over this time period had a unique exposure to insti-
417 tutional characteristics that may have contributed to the observed relationship between
418 industrialization and economic development.
419 First, the emergence of state centralization in France, centuries prior to the pro-
420 cess of industrialization, and the concentration of political power in Paris, may have af-
421 fected differentially the political culture and economic prosperity in Paris and its suburbs
422 (i.e., Seine, Seine-et-Marne and Seine-et-Oise). Hence, the empirical analysis includes a
423 dummy variable for these three departments, accounting for their potential confounding
424 effects on the observed relationship between industrialization and economic development,
425 in general, and the adoption of the steam engine, in particular. Moreover, the analysis
426 captures the potential decline in the grip of the central government in regions at a greater
12
427 distance from Paris, and the diminished potential diffusion of development into these re-
428 gions, accounting for the effect of the aerial distance between the administrative center
429 of each department and Paris.
430 Second, the relationship between industrialization and development in the Alsace-
431 Lorraine region (i.e., the Bas-Rhin, Haut-Rhin and the Moselle departments) that was
432 under German domination in the 1871-1918 period may represent the persistence of
433 institutional and economic characteristics that reflected their unique experience.9 Hence,
434 the empirical analysis includes a dummy variable for these regions, accounting for the
435 confounding effects of the characteristics of the region.
436 2.3.3. Pre-Industrial Development

437 The differential level of development across France in the pre-industrial era may
438 have affected jointly the subsequent process of development and the process of industri-
439 alization. In particular, it may have affected the adoption of the steam engine and it
440 may have generated, independently, a persistent effect on the process of development.
441 Hence, the empirical analysis accounts for the potentially confounding effects of the level
442 of development in the pre-industrial period, more than 150 years prior to the 1860-1865
443 industrial survey. This early level of development is captured by the degree of urbaniza-
444 tion (i.e., population of urban centers with more than 10,000 inhabitants) in each French
445 department in 1700 as well as by the presence of a university in 1700 and 1793.10
446 3. Empirical Methodology
447 3.1. Empirical Strategy: The Diffusion of the Steam Engines from Fresnes-sur-Escaut
448 The observed relationship between industrialization and economic development is
449 not necessarily indicative of the causal effect of industrialization on economic prosperity.
450 It may reflect the impact of economic development on the process of industrialization as
451 well as the influence of institutional, geographical, cultural and human capital character-
452 istics on the joint evolution of process of development and the onset of industrialization.
453 In light of the endogeneity of industrialization, this research exploits plausibly exogenous
454 geographic sources of regional variation in the diffusion and adoption of steam engines
455 across France to establish the effect of industrialization on the process of development.11
9
Differences in the welfare laws and labor market regulations in Alsace-Lorraine and the rest of France
persisted throughout most of the 20th century. Moreover the laws on the separation of Church and State
are different, and these differences were reaffirmed by a decision of the Supreme French Constitutional
Court in 2013 (Decision 2012-297 QPC, 21 February 2013).
10
The qualitative analysis remains intact if the potential effect of past population density is accounted
for as we show in the Appendix.
11
The working paper version of this article ((Franck and Galor, 2017a)) develops a second component
to the identification strategy which exploits regional variations in temperature deviations (on the eve
of the industrial survey) from their historical trend to capture exogenous sources of variation in the
profitability of agriculture and therefore in the pace of industrialization and the demand for steam engine
technologies across departments. To capture the changes in the profitability of agriculture production
in the eve of the 1860-1865 industrial survey on the adoption of steam engines across departments,
the analysis exploits regional variations in the squared deviations of fall temperatures in the 1856-1859
period from the average fall temperature, over the earlier 25-year period (1831-1855). Indeed temporary
temperature deviations and their adverse effect on the supply and the stock of crops have a significant
13
456 This identification strategy is motivated by the historical account of the gradual re-
457 gional diffusion of the steam engine in France during the 18th and 19th century ((Ballot,
458 1923, Sée, 1925, Léon, 1976)). Considering the positive association between industrial-
459 ization and the intensity in the use of the steam engine ((Mokyr, 1990, Bresnahan and
460 Trajtenberg, 1995, Rosenberg and Trajtenberg, 2004)), the study takes advantage of the
461 regional diffusion of the steam engine to identify the effect of local variations in the
462 intensity of the use of the steam engine during the 1860-1865 period on the process of
463 development. In particular, it exploits the distances between each French department and
464 Fresnes-sur-Escaut (in the Nord department), where the first successful commercial and
465 industrial application of the steam engine in France was made in 1732, as an instrument
466 for the use of the steam engines in 1860-1865.12
467 Consistent with the diffusion hypothesis, the second steam engine in France that
468 was utilized for commercial purposes was operated in 1737 in the mines of Anzin, also in
469 the Nord department, less than 10 km away from Fresnes-sur-Escaut. Furthermore, in
470 the subsequent decades till the French Revolution the commercial use of the steam engine
471 expanded predominantly to the nearby northern and north-western regions. Nevertheless,
472 at the onset of the French revolution in 1789, steam engines were less widespread in France
473 than in England. A few additional steam engines were introduced until the fall of the
474 Napoleonic Empire in 1815, notably in Saint-Quentin in 1803 and in Mulhouse in 1812,
475 but it is only after 1815 that the diffusion of steam engines in France accelerated ((Sée,
476 1925, Léon, 1976)).
477 [Insert Table 1]
478 Indeed, in line with the historical account, the unequal distribution of steam engines
479 across French departments, as reported in the 1860-1865 industrial survey, is indicative
480 of a local diffusion process from Fresnes-sur-Escaut. As reported in Column 1 of Table 1,
481 there is a highly significant negative correlation between the aerial distance from Fresnes-
482 sur-Escaut to the administrative center of each department and the intensity of the use
483 of steam engines in the department (the full specification is reported in Appendix Table
484 B.1). Nevertheless, as discussed in Section 2.3, pre-industrial development and a wide
485 range of confounding geographical and institutional characteristics may have contributed
486 to the adoption of the steam engine. Reassuringly, the unconditional negative relationship
487 remains highly significant and is larger in absolute value when exogenous confounding
488 geographical controls (i.e., land suitability, latitude, rainfall and temperature) (Column
489 2), as well as institutional factors (Column 3) and pre-industrial development (Column
490 4), are accounted for. In particular, the findings suggest that pre-industrial develop-
491 ment, as captured by the degree of urbanization in each department in 1700 and the
492 characteristics that may have brought this early prosperity, had a persistent positive and
493 significant association with the adoption of the steam engine.13 Importantly, the diffusion
positive effect on the prices of the dominating crops in the subsequent years, but no longer on the
productivity in the agricultural sector, and they diminish therefore the incentive to reallocate resources
from the agricultural to the industrial sector. However statistical tests of significance suggest that
temperatures deviations are only a weak instrument to the adoption of steam engines. For this reason,
this article uses the Distance to Fresnes as its sole IV.
12
This steam engine was used to pump water in an ordinary mine of Fresnes-sur-Escaut. It is unclear
whether Pierre Mathieu, the owner of the mine, built the engine himself after a trip in England or
employed an Englishman for this purpose ((Ballot, 1923, p.385)).
13
Conceivably, human capital in the pre-industrial area could have affected the adoption of the steam
14
494 pattern of steam engines is not significantly correlated with the distance between Paris
495 and the administrative center of each department when the distance from Fresnes to each
496 department’s administrative center is excluded from the analysis (Column 5). Moreover,
497 Column 6 of Table 1 and Appendix Figure A.4 indicate that, when the distance to Paris
498 is accounted for, there is still a highly significant negative correlation between the dis-
499 tance from Fresnes-sur-Escaut to the administrative center of each department and the
500 intensity of the use of steam engines in the department.
501 Accounting for the confounding effects of geographical, institutional and pre-industrial
502 characteristics, Column (6) in Table 1 reports the significant negative relationship be-
503 tween the horse power of steam engines and the instrumental variable Distance from
504 Fresnes: a 100-km increase in the distance from Fresnes-sur-Escaut is associated with a
505 1.13 decrease in the log of horse power of steam engines in a department, relative to the
506 departmental average of log horse power of 3.26. Namely, if the distance of a department
507 away from Fresnes was to increase from the 25th percentile (336.6 km) to 75th percentile
508 (680.3 km), this department would experience a 4.57 point decrease in the log of the total
509 horse power of steam engines, i.e., a decrease in the amount of horse power worth 96.6
510 hp (relative to a sample mean of 1839.35 hp).
511 The validity of the aerial distance from Fresnes-sur-Escaut as an instrumental vari-
512 able for the intensity of the adoption of steam engines across France is enhanced by three
513 additional factors. First, Table B.2 establishes that, conditional on the distance from
514 Fresnes-sur-Escaut, distances between each department and major centers of economic
515 power in 1860-1865 are uncorrelated with the intensive use of the steam engine over this
516 period. In particular, conditional on the distance from Fresnes-sur-Escaut, distances be-
517 tween each department and Marseille and Lyon (the largest cities in France after Paris in
518 the Center and South-East of France), Rouen (a major harbor in the north-west where the
519 steam engine was introduced in 1796), and Bordeaux (a major harbor in the south-west)
520 are uncorrelated with the adoption of the steam engine, lending credence to the unique
521 role of Fresnes-sur-Escaut and the introduction of the first steam engine in this location
522 in the diffusion of the steam engine across France.14 Table B.2 further establishes that
523 conditional on the distance from Fresnes-sur-Escaut, distances between each department
524 and London (i.e., the capital of England which was the largest industrial economy in 19th
525 century Europe and the birthplace of the industrial revolution) are uncorrelated with the
526 use of the steam engine within France.15
527 Second, the distance from Fresnes-sur-Escaut is uncorrelated with economic devel-
528 opment across France in the pre-industrial period. Unlike the highly significant negative
529 relationship between the distance from Fresnes-sur-Escaut and the intensity of the use
engine, as well as the subsequent process of development. Nevertheless, in light of the scarcity of data
on reliable human capital for the pre-industrial period, the baseline analysis does not account for this
confounding factor. Instead, Section 4.2.3 shows the robustness of the results to the inclusion of pre-
industrial levels of human capital for a smaller set of departments.
14
As reported in Table B.4, the use of an alternative measure of distances based on the time needed
for a surface travel between any pair of locations ((Özak, 2010)) does not affect the qualitative results.
15
Since several measures of distances are accounted for (i.e., distance from Fresnes-sur-Escaut, distance
from Paris, as well as distance from the equator), considering more than one alternative distance at a
time is very likely to generate multicollinearity that would affect the significance of all coefficients. In
particular, as follows from the concept of vector decomposition, there exist two locations such that the
effect of a distance from a given location can be captured by the effect of distances from these two
locations.
15
530 of the steam engine in 1860-1865, Table B.3 and Figure A.3 establish that the distance
531 from Fresnes-sur-Escaut was uncorrelated with urban development and human capital
532 formation in the pre-industrial era. In particular, Column 1 in Table B.3 shows that
533 urbanization rates in 1700 are uncorrelated with the distance from Fresnes-sur-Escaut.
534 Column 4 establishes that literacy rates in the pre-industrial period, as captured by the
535 share of grooms who could sign their marriage license in 1686-1690, are uncorrelated with
536 the distance from Fresnes-sur-Escaut. Finally, Column 7 demonstrates that there is no
537 significant relationship between the presence of a university in 1700 and the distance from
538 Fresnes-sur-Escaut.16 Moreover, Table B.3 and Figure A.3 establish that the distance to
539 Fresnes is not a predictor of development in the 18th century, as captured by urbanization
540 rates in 1780 and changes in urbanization between 1700 and 1780 (Columns 2 and 3),
541 literacy in 1786-1790 and changes in literacy between 1686-90 and 1786-90 (Columns 5
542 and 6) as well as the presence of an university in the department in 1793 and the change
543 in the presence of an university between 1700 and 1793 (Columns 8 and 9).
544 Third, it appears that the Nord department had neither superior human capital
545 characteristics nor higher standard of living in comparison to the average department
546 in France. An imperfect measure of literacy (i.e., grooms who could sign their wedding
547 contract over the 1686-1690 period) prior to the introduction of the first steam engine
548 in 1732, suggests that if anything, Nord’s literacy rate was below the French average.
549 Specifically, only 10.45% of men in Nord could sign their wedding contract over the 1686-
550 1690 period while the average for the rest of France was 26.10% (with a standard deviation
551 of 14.86%) ((Furet and Ozouf, 1977)). Furthermore, using height as an indicator for the
552 standard of living suggests that the standard living in Flanders, the province of the French
553 kingdom prior to 1789 which contained Fresnes-sur-Escaut, was nearly identical to that
554 of the rest of France ((Komlos, 2005)).17 As depicted in Figure G.1 in the Appendix,
555 variations in the average height of French army soldiers from Flanders over the 1700-65
556 period were not different from those of the soldiers from other parts of France.
557 Finally, the highly significant negative effect of the distance from Fresnes-sur-Escaut
558 to the administrative center of each department is robust to the inclusion of an additional
559 set of confounding geographical, demographic and institutional characteristics, as well as
560 to the forces of pre-industrial development, which as discussed in the Appendix may have
561 contributed to the relationship between industrialization and economic development. As
562 established in Appendix Table B.5, these confounding factors, which could be largely
563 viewed as endogenous to the adoption of the steam engine and are thus not considered
564 as part of the baseline analysis, do not affect the qualitative results.
565 3.2. Empirical Model

566 The effect of early industrialization on the entire subsequent process of development
567 is estimated using 2SLS. The second stage provides a cross-section estimate of the rela-
568 tionship between the total horse power of steam engines in each department in 1860-1865
16
It should be noted that these pre-industrial measures of development are highly correlated with
income per-capita in the post-industrialized period. For instance, the urban population in 1700 is pos-
itively correlated with all our measures of GDP per capita in 1860 (0.570), 1901 (0.293), 1930 (0.551)
and 2001-2005 (0.517).
17
Concerns regarding selection bias suggest that the height of soldiers may not always be representative
of the height of the general population ((see, e.g., Baten, 2000)) but there is no reason to think that this
selection bias would be more or less intense in Flanders than in the rest of France.
16
569 to measures of income per capita, human capital formation and other economic outcomes
570 at different points in time;
Yit = α + βEi + Xi0 ω + εit , (1)
571 where Yit represents one measure of economic outcomes in department i in year t, E i is
572 the log of total horse power of steam engines in department i in 1860-1865, Xi0 is a vector
573 of geographical, institutional and pre-industrial economic characteristics of department i
574 and it is an i.i.d. error term for department i in year t.18
575 In the first stage, E i , the log of total horse power of steam engines in department
576 i in 1860-1865 is instrumented by D i , the aerial distance (in kilometers) between the
577 administrative center of department i and Fresnes-sur-Escaut;
Ei = δ1 Di + Xi0 δ3 + µi , (2)
578 where Xi0 is the same vector of geographical, institutional and pre-industrial economic
579 characteristics of department i used in the second stage, and µi is an error term for
580 department i.
581 4. Industrialization and the Evolution of the Standard of Living
582 4.1. Industrialization and the Dynamics of Income per Capita

583 The study examines the effect of the intensity in the use of steam engines in the
584 1860-1865 period on the evolution of income per capita in the process of development.
585 As established in Table 2 (the results with the full set of controls in shown in Appendix
586 Tables B.6 and B.7 ), and depicted in Figure 2, consistently with the proposed hypothesis,
587 industrialization was conducive for economic development in the short-run and in the
588 medium-run but had a detrimental effect on standards of living in the long-run.19 In
589 particular, the horse power of steam engines in industrial production in the 1860-1865
590 period had a positive and significant impact on income per capita in 1860, 1901 and 1930
591 but a negative and significant effect on income per capita during the 2001-2005 period.
593 The relationship between industrialization and income per capita in the short-run
594 and in the medium-run is presented in the top panel of Table 2 and in the first five
595 Columns of the bottom panel of Table 2. As shown in Columns (1) and (6) in the top
596 panel of Table 2 and in Column (1) in the bottom panel of Table 2, unconditionally, the
597 horse power of steam engines in industrial production in the 1860-1865 period had a highly
598 significant positive association with income per capita in 1860, 1901 and 1930. Moreover,
599 this relationship remains positive, although somewhat smaller and less significant, once
600 one progressively accounts for the confounding effects of exogenous geographical factors
18
The early industrial survey of 1839-1847 is not comparable to the 1860-1865 survey since it does
not account for the horse power of steam engine.
19
Given data limitation on income per capita across departments in the post-industrial survey period
(as elaborated in section 2), the immediate effect of industrialization on income per capita is captured
by its impact in 1860, its short-run effect by its impact in 1901, its medium-run effect by its impact in
1930, while its long-run effect is captured by its impact on the average level of income per capita across
departments over the 2001-2005 period.
17
601 (Columns (2) and (7) in the top panel of Table 2 and Column (2) in the bottom panel
602 of Table 2), institutional factors (Columns (3) and (8) in the top panel of Table 2 and
603 Column (3) in the bottom panel of Table 2) and pre-industrial characteristics (Columns
604 (4) and (9) in the top panel of Table 2 and Column (4) in the bottom panel of Table 2).
605 Finally, mitigating the effect of omitted variables on the observed relationship, the IV
606 estimates in Columns (5) and (10) in the top panel of Table 2 suggest that the horse power
607 of steam engines in 1860-1865 had a positive and significant impact on income per capita
608 in 1860 and 1901, accounting for the confounding effects of geographical, institutional,
609 and demographic characteristics. We also note that the effect of industrialization in
610 1860-1865 on log GDP per capita in 1930 is still positive and significant in Column (5)
611 of the bottom panel of Table 2. A one-percent increase in the total horse power of steam
612 engines in a department in 1860-1865 increased GDP per capita by 0.124 percent in 1860,
613 0.25 in 1901 and 0.103 percent in 1930. As such, if a department had increased its total
614 horse power of steam engines in 1860-1865 from the 40th percentile (380 hp) to the 60th
615 percentile (762 hp) of the distribution, it would have experienced an increase in GDP per
616 capita of 12.47 percent in 1860, 25.13 percent in 1901 and 10.35 percent in 1930.
618 The relationship between industrialization and income per capita in the long-run is
619 presented in the last six columns of the bottom panel of Table 2. As shown in Column (6),
620 unconditionally, the horse power of steam engines in industrial production in 1860-1865
621 has a significant positive association with the average level of income per capita across
622 departments over the 2001-2005 period. Moreover, this relationship remains positive,
623 although smaller and ultimately insignificant, once one progressively accounts for the
624 confounding effects of exogenous geographical factors (Column (7) in bottom panel of
625 Table 2), institutional factors (Column (8) in bottom panel of Table 2) and pre-industrial
626 characteristics (Column (9) in bottom panel of Table 2). However, once the effect of
627 omitted variables is accounted for, the IV estimate in Column (10) in the bottom panel
628 of Table 2, suggests that the horse power of steam engines in 1860-1865 had a negative
629 and significant impact on the average level of income per capita across departments over
630 the 2001-2005 period. A one-percent increase in the total horse power of steam engines
631 in 1860-1865 decreased GDP per capita in 2001-2005 by 0.082 percent. In other words,
632 if a department had experienced an increase in its horse power in 1860-1865 from the
633 40th percentile (380 hp) to the 60th percentile (762 hp) of the distribution, this increase
634 would have led to a 8.24 percent decrease in GDP per capita in 2001-2005. The validity
635 of the IV regression in Column (10) is confirmed by the reduced form regression which is
636 also reported in Table 2.20
637 It is important to note that the IV estimate reverses the OLS estimates of the
638 relationship between industrialization and the long-run level of income per capita from a
639 positive to a negative one. This reversal suggests that factors which fostered industrial
640 development, rather than industrialization per se, contributed to the positive association
641 between industrialization and long-run development. In particular, once one accounts for
642 the effect of these omitted factors, industrialization has an adverse effect on the standard
20
To relieve concerns that these results are driven by selection on observables, we report the AET
statistic developed by Altonji et al. ((2005)) and the δ statistic developed by Oster ((2019)). In all the
OLS regressions of Table 2, both statistics are different from the critical value of 1, thereby suggesting
unobservables cannot fully account for the size of the coefficients associated with the horse power of
steam engines.
18
643 of living in the long-run. In contrast, in earlier periods (i.e., 1860, 1901, and 1930) when
644 industrialization contributed to economic development, the net effect of industrialization
645 could have been obscured by these omitted characteristics (e.g., state capacity) that
646 while being instrumental for industrialization, were associated with the protection of the
647 agricultural sector, reducing income per capita and thus lowering its estimated association
648 with industrialization. Consequently, once the net effect of industrialization is accounted
649 for, the IV coefficient is significantly larger.
650 In particular, as discussed in Section 2.3, the regressions in Table 2 account for a large
651 number of confounding geographical and institutional factors. First, the climatic and soil
652 characteristics of each department (i.e., land suitability, average temperature, average
653 rainfall, and latitude) could have affected natural land productivity and therefore the
654 feasibility and profitability of the transition to the industrial stage of development, as
655 well as the evolution of aggregate productivity in each department. Indeed, as predicted,
656 land suitability had a significantly negative association with income per capita in 1901 and
657 1930 in the IV regressions (Column (10) in top panel of Table 2 and Column (5) in bottom
658 panel of Table 2), suggesting that more productive land had an adverse effect on the
659 incentive to adopt the industrial technology. Moreover, the latitude of each department
660 had a positive and significant relationship with income per capita in 1901 and 1930
661 (Columns (7) to (10) in top panel of Table 2 and Columns (2) to (5) in bottom panel
662 of Table 2), capturing characteristics of northern departments which were conducive to
663 economic prosperity. Moreover, the lack of statistical significance of the geographical
664 variables on log GDP per capita in 2000-2005 (Column (10) in bottom panel of Table 2)
665 is in line with the idea that geographic characteristics do not have much of a role in the
666 modern growth regime which is characterized by human capital accumulation ((Galor,
667 2011)).
668 Second, the location of departments (i.e., latitude, border departments, maritime
669 departments, departments at a greater distance from the concentration of political power
670 in Paris, and those that were temporarily under German domination) could have affected
671 the diffusion of the steam engine and the diffusion of development. However, most of
672 these factors appear orthogonal to the evolution of income per capita, except for the
673 dummy variable for Paris and its suburbs that is significantly associated with income per
674 capita in 1930 and in the 2001-2005 period (Columns (3) to (5) and (8) to (10) in bottom
675 panel of Table 2).
676 Third, the regressions account for the potentially confounding effects of the level of
677 development in the pre-industrial period, as captured by the degree of urbanization in
678 each department in 1700. The findings suggest that pre-industrial development (and the
679 characteristics that may have brought this early prosperity) had a persistent positive and
680 significant effect on later stages of development, as captured by the level of income per
681 capita in 1930 and 2001-2005 (Columns (4), (5), (9) & (10) in bottom panel of Table 2),
682 but no robust impact on the early phases of industrialization, as captured by income per
683 capita in 1860 and 1901 (Columns (4), (5), (9) & (10) in top panel of Table 2).
684 Moreover, as established in Tables B.8-B.13 in the Appendix, the association between
685 the intensity of the steam engines and income per capita in 1861, 1901, 1930 and 2001-
686 2005 is not affected by spatial correlation or by clustering the standard errors at the
687 regional level (using the current 12 administrative regions of mainland France).
688 Finally, the Appendix examines the robustness of the baseline analysis to the in-
19
689 clusion of an additional set of confounding geographical, demographic, political and in-
690 stitutional characteristics, as well as for the forces of pre-industrial development, which
691 may have contributed to the relationship between industrialization and economic devel-
692 opment. The analysis focuses on the potential impact of these confounding factors on
693 the IV regressions in Table 2. As will become apparent, some of these confounding fac-
694 tors could be viewed as “bad controls”, i.e., as endogenous to the adoption of the steam
695 engine, and they are thus not part of the baseline analysis. These are: the initial level of
696 income per capita, initial population density, initial levels of human capital such as the
697 number of Encyclopédie subscribers ((Squicciarini and Voigtländer, 2015)), the share of
698 Jews and Protestants in the population, the early use of raw material, economic integra-
699 tion before and after the French Revolution ((Daudin, 2010)), the Napoleonic blockade
700 ((Juhasz, 2018)), as well as industrial concentration and firm size.
701 4.2. External Validity and Policy Implications

702 This section examines the external validity of the finding that industrialization is
703 detrimental to long-run prosperity for less-developed societies. If one views each French
704 department as a small-open economy, one may argue that the proper industrial policy
705 ought to encourage the development of skilled-intensive sectors rather than of the tradi-
706 tional unskilled-labor-intensive sectors.
707 However, one concern could be the negative effect of industrialization in the long-
708 run at the departmental level does not reflect the overall effect of industrialization. A
709 priori, it is possible that industrialization generated technological spillovers such that the
710 most industrialized department within a region declined but the region prospered as a
711 whole due to the spillovers from this industrialization process. In order to explore this
712 important possibility, Table B.14 analyzes the effect of industrialization at the regional
713 level (using the current administrative divisions of France into 12 regions, each of which
714 consists approximately of seven departments). The results demonstrate that the regions,
715 as a whole, experienced an identical pattern, i.e., an increase in prosperity in the short-
716 run and a decline in in the long-run. Nevertheless, our empirical analysis suggests that
717 the negative impact of industrialization on long-run prosperity in one department did
718 not generate sufficiently positive spillovers in neighboring departments so as to avert the
719 adverse effects of industrialization on long-run prosperity of the region as a whole.21
720 Furthermore, when considering industrial policy in currently less developed societies,
721 one has to account for the fact that those societies could adopt existing technologies
722 of various degrees of skill-intensity. Namely, less developed countries do not need to
723 develop less-skilled industries in order to adopt skilled-intensive ones. In this respect,
724 the departmental level analysis of France suggests that less developed societies ought
725 to promote the development of skilled-intensive sectors rather than an unskilled-labor-
726 intensive industrial sector.
21
Distances from Fresnes-sur-Escaut to each department are powerful predictors of the intensity of
the use of steam engines in 1860-65 across French departments as a whole. However, distances from
Fresnes-sur-Escaut are not sufficiency refined to provide a significant prediction of the variations in the
use of the steam engine across nearby departments within each of 12 contemporary regions of mainland
France. Hence, one cannot use the proposed IV strategy to test for the reversal within each of these
12 regions (i.e., across 7 nearby departments on average). Nevertheless, reassuringly, as established in
Table B.14, the reversal in the effect of the industrialization on long run prosperity is highly significant
across these 12 regions as well.
20
727 5. Mechanisms and Alternative Channels
728 This section explores several potential mechanisms that could have led to the detri-
729 mental effect of industrialization on the standard of living in the long-run. The analysis
730 suggests that the adverse effect of earlier industrialization on long-run prosperity can be
731 attributed to the negative impact of the reliance upon unskilled-labor-intensive technolo-
732 gies in the early stages of industrialization on the long-run level of human capital and
733 thus on the incentive to adopt skill-intensive technologies in the contemporary era. The
734 adverse effect of industrialization on long-run prosperity reflects a dual techno-cultural
735 lock-in effect characterized by cultural inertia in the intergenerational transmission of a
736 lower predisposition towards investment in human capital along with technological inertia
737 manifested by the persistence predominance of low-skilled-intensive industries.
738 Consistent with the proposed techno-cultural dual lock-in hypothesis, the empirical
739 analysis exploits individual data on education achievements of over 2100 second gener-
740 ation migrants to establish that acquired comparative advantage in the unskilled-labor-
741 intensive industrial sector has ultimately depressed the demand for human capital and
742 have triggered a cultural process characterized by intergenerational transmission of lower
743 educational aspirations. Furthermore, using individual data on the composition of em-
744 ployment across sectors among over 1.1 million individuals, the study suggests that this
745 cultural inertia, and its adverse effect on human capital formation in the long-run, has
746 further hindered the incentive of competitive industries to adopt more lucrative skilled-
747 intensive technologies, reinforcing the suboptimal level of human capital formation and
748 reducing the standards of living in the long-run. Interestingly, the findings suggest that
749 the adverse effect of earlier industrialization on long-run prosperity does not reflect a
750 delayed transition from the industrial to the service sector, but rather the lower skill-
751 intensity of occupations in the service as well as in the industrial sector in historically
752 industrial regions.
753 The empirical analysis in Appendix F further establishes that various ex-ante plau-
754 sible channels do not account for the adverse effect of early industrialization on long-run
755 prosperity: (1) the contribution of industrialization to unionization and wage rates in
756 historically industrialized regions and the comparative decline of these regions in the
757 long-run due to the incentive of modern industries to locate in regions where labor mar-
758 kets are more competitive, (2) the effect of trade protection in traditional industries
759 on the decline in the long-run competitiveness of historically industrialized regions, (3)
760 the potential negative effect of disproportional destruction of industrialized regions dur-
761 ing WWI and WWII on the subsequent development of these regions, (4) the persistent
762 adverse effect of selective migration (e.g. immigration of unskilled workers into industrial-
763 ized regions, or the emigration of more educated workers into less industrialized regions),
764 on the composition of human capital and long-run income per capita in historically in-
765 dustrialized regions, and (5) the disproportionate public investment in human capital in
766 non-industrial regions. Moreover, the empirical analysis demonstrates that the adverse
767 effect of industrialization on long-run prosperity has not been mitigated by the positive
768 effect of early industrialization on population density in historically industrialized regions
769 and its beneficial effects via the spillover from agglomeration to long-run development.
21
770 5.1. Industrialization and the Dual Techno-Cultural Lock-In Effect
771 This section provides supporting evidence for the hypothesis that the adverse effect
772 of industrialization on long-run prosperity can be attributed to a dual techno-cultural
773 lock-in effect. In particular, it demonstrates that the effect of early industrialization
774 on acquired comparative advantage in the unskilled-labor-intensive industrial sector de-
775 pressed the demand for human capital and have triggered cultural inertia, characterized
776 by a lower predisposition towards investment in human capital, that has further hindered
777 the incentive of firms to adopt skilled-intensive technologies, adversely effecting human
778 capital formation and the standards of living in the long-run.
779 The empirical support for the hypothesized dual lock-in effect is based on three
780 findings. First, the analysis establishes that indeed industrialization have an adverse effect
781 on human capital formation in the long-run. Second, it demonstrates the existence of a
782 cultural lock-in effect, reflected by the persistent effect of historical industrialization on
783 investment in human capital among second generation migrants who are no longer residing
784 in these industrial regions. Third, it suggests that under investment in human capital
785 affected the contemporary composition of industries in historically industrial regions and
786 reinforced the specialization in unskilled-labor-intensive industries.
787 5.1.1. Industrialization and the Evolution of Human Capital

788 This subsection explores the validity of the first building block of the hypothesized
789 dual techno-cultural lock-in effect that has governed the detrimental effect of industri-
790 alization on the standard of living in the long-run. It establishes the adverse effect of
791 historically industrialized regions on the the contemporary level of human capital and
792 thus on the skill-intensity of the production process in these regions in the long-run.
793 As established by Franck and Galor ((2017b)), due to capital-skill complementarity,
794 intensive industrialization in the middle of the 19th century had a significantly positive
795 effect on human capital formation in the subsequent decades in France (as well as in
796 other industrial nations, see, e.g., Galor and Moav ((2006))). Nevertheless, Table 3
797 shows that, as the novelty of the industrial technology diminished, the positive effects
798 of early industrialization on human capital formation had dissipated in the early part of
799 the 20th century.22 Instead, inertia in the use of unskilled-labor-intensive technologies in
800 historically intensive industrial regions, have had an adverse effect on the time path of
801 human capital formation.
803 Furthermore, the horse power of steam engines in industrial production in 1860-1865
804 had a progressively larger adverse effect on the share of men age 25 and above who had
805 at least high-school degree in the past few decades as reported in the IV regressions in
806 Columns (1)-(6) of Table 4. These adverse effects are statistically significant in 1975,
807 1982, 1990, 1999 and 2010 (similar results are found for women as reported in Table E.1
808 in the Appendix).
22
As reported in Column (3) of Table 3, the horse power of steam engines in industrial production
in 1860-1865 had initially a positive and highly significant effect on the literacy of the French army
conscripts in the years 1874-1883. However, due to the legislation of the 1881-1882 education laws,
which made primarily schooling compulsory and free until the age of 13, the effect was quantitatively
smaller but still statistically significant in the years 1894-1903 (Column (6)). Eventually, this effect had
vanished in the years 1910-1912 (Column (9)).
22
810 Moreover, as reported in the IV regressions in Columns (7)-(8) of Table 4, the horse
811 power of steam engines in industrial production in 1860-1865 has a highly significant
812 negative effect on human capital formation in the long-run, as captured by the shares
813 of male age 15-17 or 18-24, who attended school in 2010 (note that school attendance is
814 mandatory in France until age 16). These regressions indicate that in departments which
815 utilized the steam engine more intensively 150 years ago, a smaller fraction of men age
816 15-17 remain in high-school and a smaller fraction of men age 18-24 study in institutions
817 of higher learning (similar results are found for school enrollment rates of women age
818 15-17 and 18-24 in 2010, as reported in Table E.2 in the Appendix). As will become
819 apparent, some of this adverse effect on human capital formation can be attributed to
820 cultural inertia and its effect on reduced educational aspiration among individuals from
821 historically intensive industrial regions.
822 5.1.2. Cultural Persistence of Predisposition Towards Investment in Human Capital

823 This subsection demonstrates the existence of a cultural lock-in effect in the forma-
824 tion of human capital, reflected by the persistent effect of historical industrialization on
825 investment in human capital and aspiration with respect to human capital among the
826 descendants of these historically intensive industrial regions. In particular, it demon-
827 strates that the effect of early industrialization on acquired comparative advantage in the
828 unskilled-labor-intensive industrial sector depressed the demand for human capital and
829 have triggered cultural inertia, characterized by a lower predisposition towards invest-
830 ment in human capital, that has further hindered the incentive of firms to adopt more
831 lucrative skilled-intensive technologies, adversely effecting human capital formation and
832 the standards of living in the long-run.
833 Following the epidemiological approach for the study of cultural persistence, the
834 study exploits data on second generation migrants (i.e., individuals who live in their birth
835 department but whose parents originated from a different department within France)
836 to establish the presence of cultural inertia in the formation of human capital among
837 the decedents of these historically intensive industrial regions. The analysis of second-
838 generation migrants accounts for time invariant unobserved heterogeneity in the host
839 department (e.g., geographical, cultural and institutional characteristics), mitigating pos-
840 sible concerns about the confounding effect of host department-specific characteristics.
841 Moreover, since the historical industrial intensity in the parental department of origin
842 is distinct from that in the individual’s department of residence, the estimated effect
843 of industrial intensity in the parental department of origin on human capital formation
844 captures the culturally embodied, intergenerationally transmitted effect of industrial in-
845 tensity on human capital aspirations, rather than the direct effect of industrial intensity.
846 Insert Table 5
847 The analysis exploits a representative labor survey carried out across France in 2005
848 (Enquête Emploi INSEE), focusing on second-generation migrants age 35-60 (i.e., indi-
849 viduals who live in their birth department whose parents are originated from a different
850 department within France) among the respondents. As reported in Columns (1)-(3) of
851 Table 5, the highest educational degree of second-generation migrants, whose parents
852 were originated in historically industrial regions, was a Certificat d’Aptitude Profession-
853 nelle (i.e., a two-year vocational degree which is taken around age 14-15). Moreover,
23
854 Columns (4)-(6) of Table 5 establishes that among second-generation migrants, whose
855 parents were originated in historically industrial regions, there is a lower propensity of
856 owning a business and to be self-employed.23
857 Thus, the analysis finds that second-generation migrants whose parents were orig-
858 inally from historically industrial departments are significantly more likely to have: (i)
859 low human capital aspirations, as reflected by their acquisition of vocational education
860 and their departure from the schooling system in the end of the middle school and (ii)
861 diminished entrepreneurial spirit. Furthermore, accounting for parental occupation, the
862 effects remains highly significant. This result lend credence to the presence of cultural
863 inertia that contributed to technological inertia and triggered the dual lock-in effect.
865 Consistent with these findings, Table 6 shows that early industrialization has a
866 negative impact on contemporary appreciation of science. In particular, individuals who
867 are currently employed in a department that had a higher intensity steam engines 150
868 years ago express no interest in science and report not using science in their current work
869 in a survey carried out in 2001 (Centre de recherches politiques de Sciences Po, Enquête
870 science 2001).
871 Furthermore Appendix Table E.4 demonstrates that low human capital formation
872 and diminished entrepreneurial spirit are indeed plausible mediating channels through
873 which early industrial development has an adverse effect on the contemporary level of
874 income per capita. Columns (1)-(4) report the IV regressions from Tables 4-5 for the effect
875 of the horse power of steam engines on the share of men above 25 with a post-secondary
876 degree, the shares of men age 15-17 and 18-24 enrolled in school as well as the probability
877 that a second-generation migrant is self-employed or a business owner. Columns (5)-(9)
878 report the mediating effects of these variables in IV regressions for the effect of the horse
879 power of steam engines on income per capita in 2001-2005. The finding suggests that
880 these variables are significantly correlated with log GDP per capita in 2001-2005 at the
881 1% or 5% level, and that accounting for their association with the level of income per
882 capita, diminishes the effect of the horse power of steam engines on log GDP per capita
883 in 2001-2005.
884 5.1.3. Industrialization and Technological Inertia

885 This subsection suggests that under-investment in human capital affected the con-
886 temporary composition of industries in historically industrial regions and reinforced the
887 specialization in unskilled-labor-intensive industries. In particular, it demonstrates that
888 since the effect of early industrialization on acquired comparative advantage in the unskilled-
889 labor-intensive industrial sector depressed the demand for human capital and have trig-
890 gered cultural inertia, characterized by a lower predisposition towards investment in hu-
891 man capital, this suboptimal level of human capital formation has further hindered the
892 incentive of firms to adopt more lucrative skilled-intensive technologies, adversely effect-
893 ing human capital formation and the standards of living in the long-run.
895 The long-run negative effect of industrialization on the workforce is reflected in the
896 types of firms which employ individuals. Using a 2008 governmental survey on 1.1 million
23
On entrepreneurship in France, see notably Aghion et al. ((2017)).
24
897 employees age 30-64 from the private sector, Table 7 establishes that the horse power of
898 steam engines in 1860-1865 has a negative and significant effect on the probability that
899 private sector employees work in service firms in research & development, arts & enter-
900 tainment, real estate, and other types of service activities, but a positive and significant
901 effect on the probability that they work in low-skilled firms in the metallurgical and
902 wood sectors as well as in electrical appliances and in machinery repair. As such, Table
903 7 confirms the basic thrust of the analysis that past industrialization has contributed
904 to technological inertia and explains the continued predominance of low-skilled-intensive
905 industries.
906 These results lend credence to the argument that historical industrial regions have
907 experienced a technological lock-in effect. Namely, acquired comparative advantage in
908 the unskilled-labor-intensive sector in early stages of industrialization is associated with
909 the relative dominance of unskilled-labor-intensive firms and occupations.
910 5.2. Industrialization & the Evolution of Sectoral Employment and Composition
911 This subsection establishes that the adverse effect of earlier industrialization on long-
912 run prosperity does not reflect a delay transition from the industrial to the service sector,
913 but rather the lower skill-intensity of occupation in the service as well as the industrial
914 sector in historically industrial regions.
916 As depicted in panel A of Figure 3 (based on the IV regressions in Tables D.1 and
917 D.2 in the Appendix), an intensive use of the steam engine in 1860-1865 had a highly
918 significant positive effect on the share of employment in the industrial sector in 1861,
919 1901, and 1930. This effect remains positive and highly significant in 1968, 1975, and
920 1982, but is insignificant in 1990 and 1999 and finally becomes significantly negative in
921 2010. Furthermore, as depicted in panel B of Figure 3 (based on the IV regressions in
922 Tables D.3 and D.4 in the Appendix), an intensive use of the steam engine in 1860-1865
923 had a positive and significant effect on the share of employment in the service sector over
924 the period. only in 1861, 1901, and 1930 but subsequently the effect is insignificant.
925 The study further explores the effect of industrialization on the long-run composition
926 of human capital as it is reflected in the workforce in each department. First, it focuses on
927 the evolution in the shares of executives, middle management professions, and employees
928 (i.e., individuals with high, medium, and low levels of human capital) in the labor force
929 between 1968 and 2010.24 It demonstrates that industrialization had a detrimental effect
930 on employment in skilled-intensive occupations, although industrialization had no effect
931 on the share of employment in the service sector in the long-run. As depicted in Panels
932 C–E of Figure 3 (based on the IV regressions in Tables D.5–D.7 in the Appendix), the
933 horse power of steam engines in industrial production in 1860-1865 had a significantly
934 negative effect on the share of executives and other intellectual professions as well as
935 middle management professions among individuals age 25-54 in 2010. In contrast, the
936 effect on the share of employees is positive and significant in 2010. Furthermore, these
937 contrasting effects have become more pronounced in the past few decades, suggesting that
938 the technological lock-in effect has be intensified in the past decades as the prevalence
939 skill-intensive technologies has intensified.
24
The control group is made of farmers, artisans and other self-employed individuals.
25
940 6. Concluding Remarks
941 This research explores the effect of industrialization on the process of development.
942 In contrast to conventional wisdom that views industrial development as a catalyst for
943 economic growth, highlighting its persistent effect on economic prosperity, the study
944 establishes that while the adoption of industrial technologies was initially conducive for
945 economic development, it has had a detrimental effect on contemporary standards of
946 living.
947 The study exploits exogenous source of regional variation in the adoption of steam
948 engines during the French industrial revolution to establish that regions which industrial-
949 ized more intensively experienced an increase in literacy rates more swiftly and generated
950 higher income per capita in the subsequent decades. Nevertheless, industrialization had
951 an adverse effect on income per capita by the turn of the 21st century.
952 The research further explores the mediating channels through which earlier industrial
953 development has an adverse effect on the contemporary level of development. It suggests
954 that the adverse effect of industrialization on long-run prosperity reflects the negative
955 impact of earlier specialization in unskilled-labor-intensive industries on human capital
956 formation and the incentive to adopt skill-intensive technologies in the contemporary
957 era. Industrialization has triggered a dual techno-cultural lock-in effect characterized
958 by a reinforcing interaction between technological inertia, reflected by the persistence
959 predominance of low-skilled-intensive industries, and cultural inertia in the form of a lower
960 predisposition towards investment in human capital. In particular, while the adoption of
961 industrial technologies was conducive for economic development in the short-run, acquired
962 comparative advantage in the unskilled-labor-intensive industrial sector had triggered
963 cultural inertia, characterized by lower educational aspirations, that has hindered the
964 transition to more lucrative skilled-intensive sectors, adversely effecting human capital
965 formation and standards of living.
966 The study suggests that the characteristics that enabled the onset of industrializa-
967 tion, rather than the adoption of industrial technology per se, may have been the source
968 of prosperity among the currently developed economies that experienced an early indus-
969 trialization. Moreover the study highlights the existence of an intertemporal trade-off
970 associated with industrialization. It suggests that developing countries may benefit from
971 the allocation of resources towards human capital formation and skill-intensive sectors
972 rather than toward the development of an unskilled-labor-intensive industrial sector.
26
973 References
974 Acemoglu, D., Johnson, S., Robinson, J. A., 2001. The colonial origins of comparative development: an
975 empirical investigation. The American Economic Review 91 (5), 1369–1401.
976 Aghion, P., Akcigit, U., Lequien, M., Stantcheva, S., 2017. Do entrepreneurship and self-employment
977 respond to simpler fiscal incentives? Evidence from France, working Paper.
978 Altonji, J. G., Elder, T. E., Taber, C. R., 2005. Selection on observed and unobserved variables: assessing
979 the effectiveness of catholic schools. Journal of Political Economy 113 (1), 151–184.
980 Andersen, T. B., Dalgaard, C.-J., Selaya, P., 2016. Climate and the emergence of global income difference.
981 Review of Economic Studies 83, 1334–1363.
982 Ashraf, Q., Galor, O., 2013. The out of Africa hypothesis, human genetic diversity, and comparative
983 economic development. American Economic Review 103 (1), 1–46.
984 Ballot, C., 1923. L’Introduction du Machinisme dans l’Industrie Française. Slatkine Reprints (1978),
985 Geneva.
986 Baten, J., 2000. Heights and real wages in the 18th and 19th centuries: an international overview.
987 Economic History Yearbook 41 (41), 61–76.
988 Bresnahan, T. F., Trajtenberg, M., 1995. General purpose technologies: engines of growth? Journal of
989 Econometrics 65 (1), 83–108.
990 Caruana-Galizia, P., 2013. Estimating French regional income: departmental per capita gross value
991 added, 1872-1911. Research in Economic History 29, 71–95.
992 Cervellati, M., Sunde, U., 2005. Human capital formation, life expectancy, and the process of develop-
993 ment. American Economic Review 95 (5), 1653–1672.
994 Combes, P.-P., Lafourcade, M., Thisse, J.-F., Toutain, J.-C., 2011. The rise and fall of spatial inequalities
995 in France: a long-run perspective. Explorations in Economic History 48, 243–271.
996 Crafts, N., 2004. Steam as a general purpose technology: a growth accounting perspective. Economic
997 Journal 114 (495), 338–351.
998 Crouzet, F., 2003. The historiography of French economic growth in the nineteenth century. Economic
999 History Review 62 (2), 215–242.
1000 Dalgaard, C.-J., Strulik, H., 2016. Physiology and development: why the West is taller than the rest.
1001 Economic Journal 126 (598), 2292–2323.
1002 Daudin, G., 2010. Domestic trade and market size in late eighteenth-century France. Journal of Economic
1003 History 70 (3), 716–743.
1004 Dickinson, H. W., 1939. A short history of the steam engine. Cambridge University Press, Cambridge,
1005 UK.
1006 Fernihough, A., O’Rourke, K. H., January 2014. Coal and the European industrial revolution. Working
1007 Paper 19802, NBER.
1008 Forest de Bélidor, B., 1737. Architecture hydraulique, ou l’art de conduire, d’élever et de ménager les
1009 eaux pour les différens besoins de la vie. Libraire de l’Artillerie et du Génie, Paris, France.
1010 France, 1878-1939. Annuaire Statistique De La France. Département Générale, Imprimerie Nationale,
1011 Paris.
1012 Franck, R., Galor, O., 2017a. Flowers of evil? industrial development and long-run prosperity. Working
1013 Paper 23701, NBER.
1014 Franck, R., Galor, O., 2017b. Technology-skill complementarity in early phases of industrialization.
1015 Working Paper 23197, NBER.
1016 Furet, F., Ozouf, J., 1977. Lire et écrire. L’alphabétisation des Français de Calvin à Jules Ferry. Editions
1017 de Minuit, Paris, France.
1018 Galor, O., 2011. Unified Growth Theory. Princeton University Press, Princeton, NJ.
1019 Galor, O., Moav, O., 2006. Das Human-Kapital: a theory of the demise of class structure. Review of
1020 Economic Studies 73, 85–117.
1021 Galor, O., Özak, O., 2016. The agricultural origins of time preference. American Economic Review
1022 106 (10), 3064–3103.
1023 Grew, R., Harrigan, P. J., 1991. School, State, and Society: The Growth of Elementary Schooling in
1024 Nineteenth-Century France–A Quantitative Analysis. University of Michigan Press, Ann Arbor, MI.
1025 Juhasz, R., 2018. Temporary protection and technology adoption: evidence from the Napoleonic block-
1026 ade. American Economic Review 108 (11), 3339–3376.
1027 Komlos, J., 2005. Height of French soldiers, 1716-1784. ICPSR - Computer File 04363-v1.
27
1028 Léon, P., 1976. L’affermissement du phénomène d’industrialisation. In: Braudel, F., Labrousse, E. (Eds.),
1029 Histoire économique et sociale de la France, vol. 3. PUF, Paris, pp. 475–616.
1030 Litina, A., 2016. Natural land productivity, cooperation and comparative development. Journal of Eco-
1031 nomic Growth 21 (4), 351–408.
1032 Lord, J., 1923. Capital and steam power, 1750-1800. P.S. King, London, UK.
1033 Luterbacher, J., Dietrich, D., Xoplaki, E., Grosjean, M., Wanner, H., 2004. European seasonal and
1034 annual temperature variability, trends, and extremes since 1500. Science 303 (5663), 1499–1503.
1035 Luterbacher, J., Dietrich, D., Xoplaki, E., Grosjean, M., Wanner, H., 2006. European seasonal tem-
1036 perature reconstructions. IGBP PAGES/World Data Center for Paleoclimatology Data Contribution
1037 Series No. 2006-060, NOAA/NCDC Paleoclimatology Program.
1038 Mokyr, J., 1990. The Lever of Riches. Oxford University Press, Oxford, U.K.
1039 Mokyr, J., 2016. A culture of growth: the origins of the modern economy. Princeton University Press,
1040 Princeton, New Jersey.
1041 Oster, E., 2019. Unobservable selection and coefficient stability: theory and evidence. Journal of Business
1042 & Economic Statistics 37 (2), 187–204.
1043 Özak, O., 2010. The voyage of homo-oeconomicus: some economic measures of distance, brown University
1044 Mimeo.
1045 Pauling, A., Luterbacher, J., Casty, C., Wanner, H., 2006. Five hundred years of gridded high-resolution
1046 precipitation reconstructions over Europe and the connection to large-scale circulation. Climate Dy-
1047 namics 26, 387–405.
1048 Ramankutty, N., Foley, J. A., Norman, J., McSweeney, K., 2002. The global distribution of cultivable
1049 lands: current patterns and sensitivity to possible climate change. Global Ecology and Biogeography
1050 11 (5), 377 392.
1051 Rosenberg, N., Trajtenberg, M., 2004. A general purpose technology at work: the Corliss steam engine
1052 in the late nineteenth-century United States. Journal of Economic History 64 (1), 1–39.
1053 Sée, H., 1925. La vie économique de la France sous la Monarchie Censitaire (1815-1848). Librairie Félix
1054 Alcan, Paris.
1055 Squicciarini, M. P., Voigtländer, N., 2015. Human capital and industrialization: evidence from the age
1056 of enlightenment. Quarterly Journal of Economics 130 (4), 1–40.
28
1057
1058 Tables & Figures

Table 1: The Geographical Diffusion of the Steam Engine
(1) (2) (3) (4) (5) (6)

OLS OLS OLS OLS OLS OLS
Log Horse Power of Steam Engines
Distance to Fresnes -0.005*** -0.007*** -0.009*** -0.008*** -0.013***

[0.0010] [0.002] [0.003] [0.002] [0.003]
Distance to Paris 0.001 0.009***
[0.003] [0.003]
Geographic characteristics No Yes Yes Yes Yes Yes

Institutional characteristics (except Distance to Paris) No No Yes Yes Yes Yes
Pre-industrial development No No No Yes Yes Yes
Adjusted R2 0.326 0.387 0.456 0.465 0.419 0.495

Observations 89 89 89 89 89 89
Note: This table presents the results of OLS regression analysis of the geographical diffusion of the steam engine across departments in France,
as captured by the negative association between the log number of horse power of steam engines used in the department in 1860-1865 and the
distance of the department (in kilometers) from the location of the first commercial use of the steam engine in France – Fresnes-sur-Escaut.
The regressions accounts for a range of geographical, institutional, and pre-industrial characteristics. Geographic characteristics include the
department’s latitude, land suitability, average rainfall and temperature, share of carboniferous area as well as dummies for the presence
of rivers and tributaries, maritime and border departments. Institutional measures include distance to Paris, dummies for Alsace-Lorraine
and for Paris and its suburbs. Pre-industrial development characteristics include a measure of the urban population in 1700. All regressions
include a dummy variable for the three departments which had no steam engine in 1860-1865. Heteroskedasticity-robust standard errors are
reported in brackets. *** denotes statistical significance at the 1%-level, ** at the 5%-level, * at the 10%-level, for two-sided hypothesis tests.
29
Table 2: Industrialization and income per capita, 1860, 1901, 1930 & 2001-2005
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
OLS OLS OLS OLS IV OLS OLS OLS OLS IV
Log GDP per capita, 1860 Log GDP per capita, 1901
Log Horse Power of Steam Engines 0.081*** 0.049** 0.047** 0.034** 0.124*** 0.063*** 0.051** 0.048* 0.041 0.250***
[0.016] [0.019] [0.018] [0.016] [0.040] [0.016] [0.025] [0.025] [0.027] [0.077]
Geographic characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Institutional characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Pre-industrial development Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
AET 0.171 0.129 0.112 0.383 0.273 0.308

δ 0.302 0.386 0.333 0.131 0.160 0.126
Adjusted R2 0.307 0.452 0.541 0.603 0.083 0.127 0.136 0.137
Observations 87 87 87 87 87 85 85 85 85 85
First stage: Instrumented variable – Log Horse Power of Steam Engines
Distance to Fresnes -0.012*** -0.011***

[0.002] [0.002]
F-stat (1st stage) 23.866 22.512
Reduced Form: Log GDP per capita, 1860 Reduced Form: Log GDP per capita, 1901

[0.0005] [0.0005]
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
OLS OLS OLS OLS IV OLS OLS OLS OLS IV
Log GDP per capita, 1930 Log GDP per capita, 2001-2005
Log Horse Power of Steam Engines 0.067*** 0.071*** 0.058*** 0.046*** 0.103*** 0.023* 0.024* 0.015 0.002 -0.082***
[0.015] [0.016] [0.012] [0.012] [0.026] [0.012] [0.013] [0.011] [0.008] [0.026]
Geographic characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Institutional characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Pre-industrial development Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
AET 0.182 0.114 0.080 0.419 0.099 0.061

δ 0.403 0.590 0.678 0.061 0.178 0.043
Adjusted R2 0.338 0.442 0.567 0.665 0.045 0.022 0.177 0.391
Observations 87 87 87 87 87 89 89 89 89 89

[0.002] [0.002]
F-stat (1st stage) 23.608 24.635
Reduced Form: Log GDP per capita, 1930 Reduced Form: Log GDP per capita, 2001-2005

[0.0005] [0.0004]
Note: This table presents OLS and IV regressions relating the horse power of steam engines in 1860-1865 to log GDP per capita in 1860,
1901, 1930 and 2001-2005. Geographic characteristics include the department’s latitude, land suitability, average rainfall and temperature,
share of carboniferous area as well as dummies for the presence of rivers and tributaries, maritime and border departments. Institutional
measures include distance to Paris, dummies for Alsace-Lorraine and for Paris and its suburbs. Pre-industrial development characteristics
include a measure of the urban population in 1700. All regressions include a dummy variable for the three departments which had no steam
engine in 1860-1865. The Alsace-Lorraine variable is omitted from the regressions since the Alsace-Lorraine region was not part of France
between 1871 and 1914. Aerial distances are measured in kilometers. Other explanatory variables, except the dummies, and the dependent
variables are in logarithm. Heteroskedasticity-robust standard errors are reported in brackets. *** indicates significance at the 1%-level, **
at the 5%-level, * at the 10%-level.
30
Table 3: Industrialization and Literacy of conscripts, 1874-1883, 1894-1903 & 1910-1912
(1) (2) (3) (4) (5) (6) (7) (8) (9)
OLS OLS IV OLS OLS IV OLS OLS IV
Share of Literate Conscripts
1874-1883 1894-1903 1910-1912
Log Horse Power of Steam Engines 0.012* 0.014* 0.060*** 0.007** 0.010*** 0.012* 0.002 0.002 -0.005
[0.006] [0.007] [0.018] [0.003] [0.003] [0.007] [0.002] [0.002] [0.005]
Geographic characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes
Institutional characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes
Pre-industrial development Yes Yes Yes Yes Yes Yes Yes Yes Yes
Adjusted R2 0.016 0.351 0.052 0.384 -0.012 0.274

Observations 87 87 87 87 87 87 87 87 87
Distance to Fresnes -0.012*** -0.012*** -0.012***

[0.002] [0.002] [0.002]
F-stat (1st stage) 23.351 23.351 23.351
Reduced Form: Share of Literate Conscripts

1874-1883 1894-1903 1910-1912
Distance to Fresnes -0.0007*** -0.0007*** -0.0007***

[0.0002] [0.0002] [0.0002]
Note: This table presents OLS and IV regressions relating the horse power of steam engines in 1860-1865 to the share of literate conscripts
over the 1874-1883, 1894-1903 and 1910-1912 periods. All regressions include a dummy variable for the three departments which had no
steam engine in 1860-1865. Aerial distances are measured in kilometers. Other explanatory variables, except the dummies, are in logarithm.
Geographic characteristics include the department’s latitude, land suitability, average rainfall and temperature, share of carboniferous area
as well as dummies for the presence of rivers and tributaries, maritime and border departments. Institutional measures include distance to
Paris, dummies for Alsace-Lorraine and for Paris and its suburbs. Pre-industrial development characteristics include a measure of the urban
population in 1700. Heteroskedasticity-robust standard errors are reported in brackets. *** indicates significance at the 1%-level, ** at the
5%-level, * at the 10%-level.
31
Table 4: Industrialization and Human Capital Formation, 1968-2010
(1) (2) (3) (4) (5) (6) (7) (8)
IV IV IV IV IV IV IV IV
Log Post-Secondary Degree (Male 25+) Log School Enrollment (Male)
1968 1975 1982 1990 1999 2010 15-17 18-24
Log Horse Power of Steam Engines -0.035 -0.060 -0.085** -0.083** -0.077** -0.085*** -0.012*** -0.073***
[0.045] [0.041] [0.038] [0.038] [0.034] [0.032] [0.003] [0.023]
Geographic characteristics Yes Yes Yes Yes Yes Yes Yes Yes
Institutional characteristics Yes Yes Yes Yes Yes Yes Yes Yes
Pre-industrial development Yes Yes Yes Yes Yes Yes Yes Yes
Observations 89 89 89 89 89 89 89 89
First stage: Instrumented Variable – Log Horse Power of Steam Engines
Distance to Fresnes -0.010*** -0.010*** -0.010*** -0.010*** -0.010*** -0.010*** -0.012*** -0.012***
[0.002] [0.002] [0.002] [0.002] [0.002] [0.002] [0.002] [0.002]
F-stat (1st stage) 28.302 28.302 28.302 28.302 28.302 28.302 24.635 24.635
Reduced Form:
Log Post-Secondary Degree (Male 25+) Log School Enrollment (Male)
1968 1975 1982 1990 1999 2010 15-17 18-24
Distance to Fresnes 0.0004 0.0004 0.0004 0.0009** 0.0006 0.0009** 0.0001*** 0.0001***
[0.0005] [0.0005] [0.0005] [0.0004] [0.0004] [0.0004] [1.89e-05] [1.89e-05]
Note: Columns (1)-(6) present IV regressions relating the horse power of steam engines in 1860-1865 to the evolution in the share of men age
25 and above with a post-secondary degree between 1968 and 2010. Columns (7)=(8) present IV regressions relating the horse power of steam
engines in 1860-1865 to the share of school enrollment for men age 15-17 and 18-24. All regressions include a dummy variable for the three
departments which had no steam engine in 1860-1865. Aerial distances are measured in kilometers. Other explanatory variables, except the
dummies, and the dependent variables are in logarithm.Geographic characteristics include the department’s latitude, land suitability, average
rainfall and temperature, share of carboniferous area as well as dummies for the presence of rivers and tributaries, maritime and border
departments. Institutional measures include distance to Paris, dummies for Alsace-Lorraine and for Paris and its suburbs. Pre-industrial
development characteristics include a measure of the urban population in 1700. Heteroskedasticity-robust standard errors are reported in
brackets. *** indicates significance at the 1%-level, ** at the 5%-level, * at the 10%-level.
32
Table 5: Industrialization and Human Capital Inertia: Second-Generation Migrants
(1) (2) (3) (4) (5) (6)

IV IV IV IV IV IV
Vocational Certificate Business owners & self-employed
Maternal Paternal Parental Maternal Paternal Parental
Department of Origin Parental Department of Origin
Log Horse Power of Steam Engines -0.0454*** -0.0348* -0.0712*** 0.0757*** 0.0972** 0.0978**
[0.0155] [0.0178] [0.0247] [0.0285] [0.0389] [0.0445]
Individual characteristics Yes Yes Yes Yes Yes Yes

Parental characteristics No Yes No Yes No Yes
Host Department Fixed Effects Yes Yes Yes Yes Yes Yes
Geographic characteristics Yes Yes Yes Yes Yes Yes
Institutional characteristics Yes Yes Yes Yes Yes Yes
Pre-industrial development Yes Yes Yes Yes Yes Yes
Observations 2,133 2,130 966 2,133 2,130 966
Distance to Fresnes -0.013*** -0.011*** -0.011*** -0.013*** -0.011*** -0.011***

[0.002] [0.002] [0.002] [0.002] [0.002] [0.002]
F-stat (1st stage) 39.131 28.164 37.355 39.131 28.164 37.355
Reduced Form: Vocational Certificate Reduced Form: Business owners & self-employed
Maternal Paternal Parental Maternal Paternal Parental
Department of Origin Parental Department of Origin
Distance to Fresnes 0.0006*** 0.0004** 0.0008*** -0.001*** -0.001*** -0.001**

[0.0002] [0.0002] [0.0003] [0.0004] [0.0004] [0.0005]
Note: This table presents IV regressions relating the horse power of steam engines in 1860-1865 in the origin department for the parents of
second generation migrants to the probability that the highest educational qualification of these second generation migrants is a vocational
certificate that is usually obtained around age 14-15 (Columns (1)-(3)) and to the probability that these second generation migrants are self-
employed and/or business owners (Columns (1)-(3)). All regressions include fixed effects for the respondent’s birth department. The control
variables refer to the origin department of the respondent’s mother (Columns (1)-(4)), father (Columns (2)-(5)) and both parents (Columns
(1)-(6)). Aerial distances are measured in kilometers. Other explanatory variables, except the dummies, and the dependent variables are in
logarithm. Individual characteristics include the respondent’s age and gender. Parental characteristics include dummies indicating whether
the respondent’s father or mother were workers. Geographic characteristics include the department’s latitude, land suitability, average rainfall
and temperature, share of carboniferous area as well as dummies for the presence of rivers and tributaries, maritime and border departments.
Institutional measures include distance to Paris, dummies for Alsace-Lorraine and for Paris and its suburbs. Pre-industrial development
characteristics include a measure of the urban population in 1700. Heteroskedasticity-robust standard errors are reported in brackets. ***
indicates significance at the 1%-level, ** at the 5%-level, * at the 10%-level.
33
Table 6: Industrialization and Contemporary Educational Aspirations: Lack of Interest in Science
(1) (2)
IV IV
Science
No Interest No Use in Work
Horse Power of Steam Engines 0.078** 0.164**

[0.036] [0.070]
Individual characteristics Yes Yes

Geographic characteristics Yes Yes
Institutional characteristics Yes Yes
Pre-industrial development Yes Yes
Observations 1337 1276

[0.002] [0.002]
F-stat (1st stage) 18.199 17.655
Reduced Form: Science

No Interest No Use in Work

[0.002] [0.002]
Note: This table presents IV regressions relating the horse power of steam engines in 1860-1865 to the probability that respondents age 25
and above in a 2005 survey report no interest in science and no use in science in their work. All regressions include a dummy variable for
the three departments which had no steam engine in 1860-1865. Aerial distances are measured in kilometers. Other explanatory variables,
except the dummies, and the dependent variables are in logarithm. Individual characteristics include the respondent’s age and gender.
Geographic characteristics include the department’s latitude, land suitability, average rainfall and temperature, share of carboniferous area
as well as dummies for the presence of rivers and tributaries, maritime and border departments. Institutional measures include distance to
Paris, dummies for Alsace-Lorraine and for Paris and its suburbs. Pre-industrial development characteristics include a measure of the urban
population in 1700. Heteroskedasticity-robust standard errors are reported in brackets. *** indicates significance at the 1%-level, ** at the
5%-level, * at the 10%-level.
34
Table 7: Industrialization and Technological Inertia
(1) (2) (3) (4) (5) (6) (7) (8)

IV IV IV IV IV IV IV IV
Sector of Employment
Wood Industries Metallurgy Machinery Repair Electrical Appliances Other Service Activities Research & Development Real Estate Arts and Entertainment
Horse Power of Steam Engines 0.004* 0.008* 0.003** 0.001* -0.002** -0.002* -0.002* -0.003**
[0.002] [0.004] [0.001] [0.0007] [0.001] [0.001] [0.001] [0.001]
Individual characteristics Yes Yes Yes Yes Yes Yes Yes Yes
Geographic characteristics Yes Yes Yes Yes Yes Yes Yes Yes
Institutional characteristics Yes Yes Yes Yes Yes Yes Yes Yes
Pre-industrial development Yes Yes Yes Yes Yes Yes Yes Yes
Observations 1,335,553 1,335,553 1,335,553 1,335,553 1,335,553 1,335,553 1,335,553 1,335,553
Distance to Fresnes -0.015*** -0.015*** -0.015*** -0.015*** -0.015*** -0.015*** -0.015*** -0.015***
[0.003] [0.003] [0.003] [0.003] [0.003] [0.003] [0.003] [0.003]
F-stat (1st stage) 22.111 22.111 22.111 22.111 22.111 22.111 22.111 22.111
Reduced Form: Sector of Employment

Wood Industries Metallurgy Machinery Repair Electrical Appliances Other Service Activities Research & Development Real Estate Arts and Entertainment
Distance to Fresnes -6.21e-05** -6.21e-05** -0.0001** -5.65e-06 -0.0001** -1.67e-05* -4.65e-05*** -0.0001**
[2.58e-05] [2.58e-05] [5.54e-05] [3.66e-06] [5.54e-05] [1.01e-05] [1.60e-05] [5.54e-05]
Notes: This table presents IV regressions relating the horse power of steam engines in 1860-1865 to the probability that private sector employees in 2008 work for firms in the following sectors: wood industries,
metallurgy, machinery repair, other service activities, research & development, real estate, and arts & entertainment. Aerial distances are measured in kilometers. Other explanatory variables, except the dummies,
and the dependent variables are in logarithm. Individual characteristics include the respondent’s age and gender. Geographic characteristics include the department’s latitude, land suitability, average rainfall
and temperature, share of carboniferous area as well as dummies for the presence of rivers and tributaries, maritime and border departments. Institutional measures include distance to Paris, dummies for
Alsace-Lorraine and for Paris and its suburbs. Pre-industrial development characteristics include a measure of the urban population in 1700. Heteroskedasticity-robust standard errors are reported in brackets. ***
indicates significance at the 1%-level, ** at the 5%-level, * at the 10%-level.
Fresnes sur Escaut
0 - 380
381 - 762
763 - 2403
2404 - 5191
5192 - 9048
9049 - 27638
Figure 1: The distribution of the total horse power of steam engines across departments in France, 1860-1865.
36
.4
.3
.2
.1
0
-.1
1860 1901 1930 2001-2005
Figure 2: The effect of the horse power of steam engines in 1860-1865 on GDP per capita
Note: This figure displays the estimated coefficients of Horse Power of Steam Engines in the IV regressions in the top and
bottom panels of Columns 5 and 10 of Tables 2. Intervals reflect 95%-confidence levels.
37
.15
.04
.02
.1
.05
0
-.02
0
-.05
-.04
1861 1901 1931 1968 1975 1982 1990 1999 2010 1861 1901 1931 1968 1975 1982 1990 1999 2010
Panel A. The industrial sector. Panel B. The service sector.
.02
.01
.02
.01
0
.01
-.01
0
-.01
-.02
-.01
-.03
-.02
1968 1975 1982 1990 1999 2010 1968 1975 1982 1990 1999 2010 1968 1975 1982 1990 1999 2010
C. Executives in workforce. D. Intermediary professionals in workforce. E. Employees in workforce.
Figure 3: The Effect of Horse Power of Steam Engines (1860-1865) on Sectoral Employment and
Occupational Choices.
Note: Panel A: the estimated coefficients of Horse Power of Steam Engines on the share of the workforce in the industrial sector in the IV
regressions in Tables D.1 and D.2. Panel B: the estimated coefficients of Horse Power of Steam Engines on the share of the workforce in the
service sector in the IV regressions Tables D.3 and D.4. Panel C-E display the estimated coefficients of Horse Power of Steam Engines in the
IV regressions in Tables D.5 - D.7 Intervals reflect 95%-confidence levels.
38

Flowers of Evil Industrialization and Long Run Development - Raphael Franck

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Flowers of Evil Industrialization and Long Run Development - Raphael Franck

Uploaded by

Copyright:

Available Formats

Flowers of Evil?

Industrialization and Long Run Development

Flowers of Evil? Industrialization and Long Run Development

Raphaël Franck, Oded Galor

To appear in: Journal of Monetary Economics

Received date: 25 April 2019

© 2019 Published by Elsevier B.V.

2 Flowers of Evil? Industrialization and Long Run Development✩

3 Raphaël Francka,∗, Oded Galorb

19 The process of development has been marked by persistence as well as reversals

233 2. Data and Main Variables

274 2.1.2. Workforce

297 2.1.3. Human Capital

334 2.2. Steam Engines

367 2.3. Confounding Characteristics of each Department

378 2.3.1. Geographic Characteristics

411 2.3.2. Institutional Characteristics

436 2.3.3. Pre-Industrial Development

446 3. Empirical Methodology

565 3.2. Empirical Model

Yit = α + βEi + Xi0 ω + εit , (1)

581 4. Industrialization and the Evolution of the Standard of Living

582 4.1. Industrialization and the Dynamics of Income per Capita

701 4.2. External Validity and Policy Implications

787 5.1.1. Industrialization and the Evolution of Human Capital

822 5.1.2. Cultural Persistence of Predisposition Towards Investment in Human Capital

884 5.1.3. Industrialization and Technological Inertia

1058 Tables & Figures

(1) (2) (3) (4) (5) (6)

Distance to Fresnes -0.005*** -0.007*** -0.009*** -0.008*** -0.013***

Geographic characteristics No Yes Yes Yes Yes Yes

Adjusted R2 0.326 0.387 0.456 0.465 0.419 0.495

AET 0.171 0.129 0.112 0.383 0.273 0.308

First stage: Instrumented variable – Log Horse Power of Steam Engines

Distance to Fresnes -0.012*** -0.011***

F-stat (1st stage) 23.866 22.512

Distance to Fresnes -0.002*** -0.002***

AET 0.182 0.114 0.080 0.419 0.099 0.061

First stage: Instrumented variable – Log Horse Power of Steam Engines

Distance to Fresnes -0.012*** -0.012***

F-stat (1st stage) 23.608 24.635

Distance to Fresnes -0.002*** -0.001***

Adjusted R2 0.016 0.351 0.052 0.384 -0.012 0.274

First stage: Instrumented variable – Log Horse Power of Steam Engines

Distance to Fresnes -0.012*** -0.012*** -0.012***

F-stat (1st stage) 23.351 23.351 23.351

Reduced Form: Share of Literate Conscripts

Distance to Fresnes -0.0007*** -0.0007*** -0.0007***

First stage: Instrumented Variable – Log Horse Power of Steam Engines

(1) (2) (3) (4) (5) (6)

Individual characteristics Yes Yes Yes Yes Yes Yes

Observations 2,133 2,130 966 2,133 2,130 966

First stage: Instrumented variable – Log Horse Power of Steam Engines

Distance to Fresnes -0.013*** -0.011*** -0.011*** -0.013*** -0.011*** -0.011***

F-stat (1st stage) 39.131 28.164 37.355 39.131 28.164 37.355

Distance to Fresnes 0.0006*** 0.0004** 0.0008*** -0.001*** -0.001*** -0.001**

Horse Power of Steam Engines 0.078** 0.164**

Individual characteristics Yes Yes

Observations 1337 1276

First stage: Instrumented variable – Log Horse Power of Steam Engines

Distance to Fresnes -0.009*** -0.009***

F-stat (1st stage) 18.199 17.655

Reduced Form: Science

Distance to Fresnes -0.005* -0.007* -0.009* -0.008* -0.013***

Distance to Fresnes -0.012* -0.011*

Distance to Fresnes -0.002* -0.002*

Distance to Fresnes -0.012* -0.012*

Distance to Fresnes -0.002* -0.001*

Distance to Fresnes -0.012* -0.012* -0.012***

Distance to Fresnes -0.0007* -0.0007* -0.0007***

Distance to Fresnes -0.013* -0.011* -0.011* -0.013* -0.011* -0.011*

Distance to Fresnes 0.0006* 0.0004 0.0008* -0.001* -0.001* -0.001

Horse Power of Steam Engines 0.078 0.164

Distance to Fresnes -0.009* -0.009*

Distance to Fresnes -0.009* -0.009*