WPS6814


Policy Research Working Paper                       6814




          Engineers, Innovative Capacity
         and Development in the Americas
                           William F. Maloney
                          Felipe Valencia Caicedo




The World Bank
Development Research Group
Macroeconomics and Growth Team
March 2014
Policy Research Working Paper 6814


  Abstract
  This paper offers the first evidence on the prevalence of                         side elements that might be confounded with engineering
  a central actor in modern growth theory—the engineer.                             or patenting. Instrumenting engineering using the
  Using newly collected sub-national, and international                             Land Grant Colleges program further limits remaining
  data as well as historical case studies, it then argues                           endogeneity. A 1 SD increase in engineers at the turn
  that differences in innovative capacity, captured by the                          of the 20th century accounts for a 16 percent increase
  density of engineers and patents at the dawn of the                               in US county income today, and patenting capacity
  Second Industrial Revolution, in fact, are important to                           contributes another 10 percent. This can partly explain
  explaining present income differences across US counties,                         why countries with similar levels of income in 1900, but
  states within countries, and between the US and Latin                             ten fold differences in engineering density diverged in
  America. This remains the case after controlling for                              their growth trajectories over the next century.
  literacy, other higher order human capital, and demand




  This paper is a product of the Macroeconomics and Growth Team, Development Research Group. It is part of a larger
  effort by the World Bank to provide open access to its research and make a contribution to development policy discussions
  around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors
  may be contacted at wmaloney@worldbank.org.




         The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
         issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
         names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
         of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
         its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.


                                                       Produced by the Research Support Team
Engineers, Innovative Capacity and Development in the
                       Americas∗
                  William F. Maloney†                  Felipe Valencia Caicedo‡




       Keywords: Innovative Capacity, Human Capital, Engineers, Technology Diffusion,
       Patents, Growth, Development, History.

       JEL: O1, O31, O33, O4, N1


   ∗
     Corresponding author Email: wmaloney@worldbank.org. We thank Ufuk Akcigit, Leo Feler, Claudio
Ferraz, Christian Fons-Rosen, Oded Galor, Steve Haber, Lakshmi Iyer, David Mayer-Foulkes, Stelios
Michalopoulos, Guy Michaels, Giacomo Ponzetto, Luis Serven, Andrei Shleifer, Enrico Spolaore, Uwe
                  ander, Hans-Joachim Voth, Fabian Waldinger, David Weil and Gavin Wright for helpful
Sunde, Nico Voigtl¨
discussions. We are grateful to William Kerr for making available the county level patent data. Our thanks
to Carlos Andres Molina, Melissa Rubio, and Mauricio Sarrias for meticulous research assistance. This paper
was partly supported by the regional studies budget of the Office of the Chief Economist for Latin America
and Knowledge for Change (KCP) trustfund.
   †
     Development Economics Research Group, The World Bank.
   ‡
     Department of Economics and Institute of Macroeconomics and Econometrics, University of Bonn.
                       “You have all the elements, but you cannot make steel”

                                                                                      —Andrew Carnegie1




1        Introduction
    Carnegie’s taunt to the owners of the Birmingham Steel Company in Alabama high-
lights the complexity of technological transfer to lagging regions.                     As Wright (1986)
documents, shortfalls in management, learning by doing, and the R&D capacity necessary
to adapt technologies to local conditions all impeded the establishment of a dynamic
steel (as well as textile and lumber) industry in the American South, and hence slowed
catch-up to the North.           The issue of differential technological transfer remains central
to the more general convergence debate.2 Most recently Comin et al. (2008); Comin &
Ferrer (2013) argue that the diverging measured intensity of use of new technologies (as
a share of economic activity) plausibly explains observed TFP differentials and can drive
simulations that closely track the magnitudes of the Great Divergence of the last two
centuries (e.g. Pritchett, 1997; Galor & Weil, 2000; Galor, 2011).3 That human capital
is a critical ingredient in technology adoption also enjoys a substantial supporting literature.4


    The nature of the relevant human capital has been less clear. Literacy, and accumulated
years of schooling or enrollment have received the most attention, although other dimensions
figure importantly in the literature as well: Lucas (1993); Young (1993) and Foster &
Rosenzweig (1995), among others, stress the importance of accumulated “learning by doing”;
    1
        Cited in Wright (1986) p.171.
    2
      See also, Parente & Prescott (1994); Eaton & Kortum (1999, 2001); Caselli (2001); Comin & Hobijn
(2004); Keller (2004); Klenow & Rodriguez-Clare (2005); Comin et al. (2008, 2010a); Comin & Hobijn (2010);
Comin et al. (2012).
    3
     Though not clear on the underlying determinants, recently Acemoglu et al. (2013) argue for the
importance of a the shift in industrial composition from those with low to high innovative capacity as critical
to growth.
    4
    See, for example, Nelson & Phelps (1966); Foster & Rosenzweig (1996); Cohen & Levinthal (1989);
Benhabib & Spiegel (1994); Basu & Weil (1998); Temple & Voth (1998); Howitt (2000); Acemoglu & Zilibotti
(2001); Caselli (2001); Comin & Hobijn (2004); Benhabib & Spiegel (2005); Aghion et al. (2005); Howitt &
Mayer-Foulkes (2005); Ciccone & Papaioannou (2009); Goldin & Katz (2009).
                                                      1
Baumol (1990) and Murphy et al. (1991), entrepreneurial skills and orientation; in his
classic article on endogenous growth; Romer (1990) highlights the research engineer; Mokyr
(2005), the minority of “trained engineers, capable mechanics and dexterous craftsmen on
whose shoulders the inventors could stand” (pg.16); Rosenberg (2000) and Nelson (2005),
the accumulated ability and scientific institutions to manage new ideas for innovation and
invention; Cohen & Levinthal (1989); Griffith et al. (2004), the capacity for research and
development needed for technological transfer. Yuchtman (2014) shows that in early 20th
century China, engineers enjoyed massive wage premia over all other classes of higher
education and that modern Western education and engineers in particular were critical to
China’s ability to adopt Western technologies.


    Several authors including Mokyr (1998), and Howitt & Mayer-Foulkes (2005) stress that
higher-order human capital and the institutions that generated and housed it may have
had an even more determinant role at the dawn of the Second Industrial Revolution (circa
1870-1914), which saw an increased emphasis on more structured scientific inquiry such as
laboratory-based R&D.5 This scientifically oriented human capital, and a technologically
savvy entrepreneurial class6 were necessary to tap into the expanding and increasingly
sophisticated global stock of knowledge and convert it into local growth. The technological
leap forward also meant an erosion in the efficacy of existing levels of human capital and
innovative capacity relative to that needed to continue to adopt (see Howitt (2000); Aghion
et al. (2005)). Building on this insight, Howitt & Mayer-Foulkes (2005) argue for multiple
equilibria in innovation where countries whose human capital evolved with the frontier at
the time of the technological leap forward could innovate or adopt, but those whose frontier
adjusted human capital did not keep up slipped to an equilibrium where even the adoption
   5
    Rosenberg (2000) and Nelson (2005) stress the incremental and cumulative nature of technological
progress and related institutions more generally as a central dynamic of industrialization.
   6
     As numerous authors have stressed from Schumpeter (1934) to the present, technological progress
without entrepreneurs to take it to market does not lead to growth (see Aghion & Howitt, 1992; Baumol,
1990, 2010; Glaeser, 2007; Glaeser et al., 2009, 2010; Braunerhjelm et al., 2010; Iyigun & Owen, 1998, 1999).
Over the longer term this reflects the accumulation of a specific kind of human capital, at the very least,
suited to the evaluation and management of risk, but extending to skills for managing people, credit, and
technologies which need to be learned.
                                                     2
of technologies was difficult, and stagnation followed.7 Iyigun & Owen (1998, 1999) relatedly
argue that if the the initial stock of both professional (scientific) and entrepreneurial human
capital is too low, the return to accumulating human capital will be low and economies can
find themselves in a development trap.


   Empirically, documenting the impact of even very basic measures of human capital on
growth or relative incomes has proved surprisingly complex since the seminal studies of
Barro (1991) and Mankiw et al. (1992).8 Most recently, and in a similar spirit to the present
work, Hanushek et al. (2015) find that differences in human capital account for 20-35 percent
of the current variation in per-capita GDP among US states. Further, there have been
relatively few efforts to systematically capture what kind of human capital matters, or even
to document the stocks of different types of capital. Goldin & Katz (2011); Goldin (1999);
Goldin & Katz (1999) have documented the evolution of secondary and tertiary education
in the context of US growth. Judson (1998), Wolff & Gittleman (1993), Self & Grabowski
               o-Climent & Mukhopadhyay (2013) attempt to document whether tertiary
(2004), Castell´
education matters more or less than primary education. Murphy et al. (1991) document
that countries with a higher proportion of engineers grow faster relative to those with a
higher proportion of law concentrators.


   The problem is exacerbated when a longer historical perspective is taken. Campante
& Glaeser (2009) argue that the “lion’s share” of the differences in long run income level
between the cities of Buenos Aires and Chicago is human capital, but they lament that
literacy remains the primary, albeit coarse, measure (see, for example, Mariscal & Sokoloff
               ander & Squicciarini (2014) show that mid-18th century French cities with
(2000)). Voigtl¨

   7
     See also Gancia et al. (2008) for related discussion of convergence clubs resulting from education-
technology complementarities.
   8
      For overviews see Krueger & Lindahl (2001); Sianesi & VanReenen (2003); Stevens & Weale (2004)
and Acemoglu & Zilibotti (2001); Pritchett (2001); Easterly (2002); Benhabib & Spiegel (1994). A recent
literature has sought to explain the often small measured impact by incorporating measures of the quality
of human capital (see, for example Hanushek & Woessmann, 2007; Behrman & Birdsall, 1983; Hanushek &
Woessmann, 2012).
                                                   3
                                  edie, capturing higher order human capital, grew faster
more subscriptions to the Encyclop´
after the dawn of the industrial revolution. Waldinger (2012) shows that the dismissal of
scientists in Nazi Germany led to a permanent decrease in growth.


   To date, however, neither the historical prevalence of Romer’s research engineers or
Mokyr’s engineers and mechanics, nor Rosenberg’s and Nelson’s systems of innovation have
been quantified in a globally comparable form. That is, the profession has no systematic
historical evidence on allegedly the central actor in modern economic growth.9


   In this paper we, first, make this very basic contribution. We establish the stylized facts
surrounding the relative density of engineers at the end of the 19th century at the national
level for the Western Hemisphere and representative benchmark countries of Europe, and the
same at the sub-national level for six countries in the Americas. This is done in a systematic
way drawing on graduation records, membership in professional societies, and census data.
We see these estimates, in the first instance, as a measure of what they are: the stock
of higher level scientifically oriented human capital directly linked to productive activity
available at the beginning of the Second Industrial Revolution. Second, because our national
measures are based on graduates of domestic engineering schools, they also proxy for the
universities and institutions that support them, and for which data are more elusive (see
Nelson, 2005). Further, we are sympathetic to the use in Murphy et al. (1991) of engineering
density as a proxy for “good entrepreneurship,” perhaps more specifically the advancement
of the long-run cumulative process of developing entrepreneurial technological depth and
accompanying cultural factors discussed by Phelps (2013). In sum, we see it as a broad
proxy for innovative capacity. We document an extraordinary variance in engineering density
across countries of very similar income levels in 1900-for instance Argentina, Chile, Den-
mark, Sweden and the Southern US- that correlates to their divergent income positions today.


   9
     Collected articles in Fox & Guagnini (1993) have examined the evolution of engineering capacity in
several advanced countries although the comparability of the measures across countries is not always clear,
and there is no attempt to establish a link to economic performance.
                                                    4
    To more confidently establish the importance of innovative capacity as a determinant of
income, we then develop a rich data set for the United States at the county level. This vastly
expands the degrees of freedom and permits controlling for an extensive set of geographical,
growth related, human capital variables and instrumenting for possible endogeneity of
engineering using the Land Grant Colleges program. Further, we incorporate patenting
density as collected by Acemoglu et al. (2013). Patenting appears only weakly explained
by county engineering density and hence, we treat it as driven by human capital more
dedicated to invention (Griliches (1990); Sokoloff (1988)) while engineering may arguably
be capturing more adoptive activities. Both variables show an important and robust effect:
a one standard deviation in engineering density in 1900 accounts for a 16% rise in income
today, and patenting capacity another 10%.        For a broader sample of countries in the
hemisphere, we then demonstrate the continued relevance of engineering at the “state” level.
Applying the well-estimated US coefficient to the national level data, a 1 standard deviation
in engineering density in 1900 accounts for a 36% difference in income today. We then use
historical evidence to validate our ranking of innovative capacity and through case studies
document the divergent development outcomes arising within industries and even identical
goods in countries with the differing innovative capacities we quantify. Finally, we present
some suggestive findings on the determinants of innovative capacity.



2     Data

2.1    Measuring engineering/innovative Capacity

Our primary measure of innovative capacity at the national level across all countries is the
number of engineers with domestically emitted university degrees per 100,000 male workers.
Unlike literacy data, which countries sometimes collect as part of the census, countries do
not tabulate such information in a uniform fashion. Hence, we construct these series using
three sources of data.

                                              5
    Engineering Graduates : To the degree possible, calculations are done with actual
graduates of engineering colleges and universities within the country.                    Clearly, many
engineers even in the US acquired valuable training on the ground, or may have had partial
degrees from some type of technical program. However, such skills are difficult to capture
with any degree of commonality across geographical units. As a consistent metric across
countries, we take the number of degrees awarded.10 Though most countries also employed
foreign engineers, we are interested in indigenous technical capacity and the institutional
structure to generate it, so we focus on domestically trained engineers. Some nationals
studied abroad, however the historical evidence from Argentina, Chile, Colombia and
Mexico suggests the numbers are small and difficult to document. Since the working life
of an engineer is roughly 40 years, we begin accumulating the stock in 1860, discounting
the stock in each period by .983 as the rate of death/attrition in each year. In some cases,
we have a long series of graduation records which make this procedure straightforward. In
                                       u, Spain, Sweden, Venezuela, or Mexico, the flow
Bolivia, Denmark, Mexico, New York, Per´
                                                                                om (1982)’s
of graduated engineers is available for the 1860-1900 period. We refer to Ahlstr¨
estimates for Germany and France, the frontier countries, which are allegedly generated
in a virtually identical manner, yet we do not have access to his data and these are not
our calculations. In general, the larger and more advanced the country, the greater the
number and types of institutions and hence the more difficult guaranteeing comparability
and this also influenced our decision to limit our geographical focus.11 In other cases, for
instance, Argentina, Brazil, Chile, Colombia, Ecuador, and the US as a whole, the informa-
tion is less complete and we bring other sources of data to bear to fill in the gaps in the series.


    Membership in Engineering Societies : Data on membership in Engineering societies

  10
     Fox & Guagnini (1993) tabulate for several advanced countries the number of students enrolled. However,
we find often that the difference between students enrolled and eventual graduates can differ greatly so
enrollment rates are not as reliable.
  11
    In fact, he calculates the stock for France, Germany as well as Sweden for which we do have the raw
underlying data and hence can verify that we are doing the calculations in a virtually identical fashion.
                                                     6
or official registries validate broad orders of magnitude of our generated stocks. In some
cases, such as Brazil, registry with the government was required to be a practicing engineer.
What is considered an engineer, however, is less clear and hence these measures are less
definitionally tight. In other countries, such as Colombia or Argentina, membership in
Engineering associations was not required so registration likely underestimates.


    Census Data : Census data are also available in several countries. Census data have the
advantage of being collected over time by several countries and across sub-national units.
However, here, also, it is the individual respondent who is deciding whether he is an engineer
or not with limited institutional confirmation, or detail on the actual level of education.
This is fine for within country analysis but less reliable for cross country comparisons.12
Sub-national data derive from the Argentine census of 1895, Mexican National Census of
1895, Chilean Census of 1907, Colombian census of 1912, Venezuelan Census of 1926, and
US Census of 1900. For the US county level data, we use the census of 1880 to generate
measures of literacy, and density of engineers, lawyers and medical doctors.


    Annex I discusses how these three sources of data were employed for each country in detail.


    Patenting For the US, we also construct a measure of average patenting activity from
1890-1910 drawing on patents granted by the United States Patent and Trademark Office
(USPTO) again, at the county level, provided by Acemoglu et al. (2013). This is likely to
capture Mokyr’s more structured scientific inquiry at the frontier that began to assume a
central role in the Second Industrial revolution in addition to its direct interpretation as new
ideas. It does not appear driven by engineers or the other human capital variables and hence
interpret it as captures a distinct type of human capital (see Sokoloff (1988). Comparable
patent data is not available for the Latin American countries and hence the variable only


  12
     As numerous local historians have noted (Serrano (1993) in Chile, Bazant de Salda˜na (1993) in Mexico)
and in the case for which we have the best information, the US, censuses are often substantially higher than
the actual graduates of engineering programs.
                                                     7
enters in the county-level US analysis.



2.2       Sub-national Income per Capita

Income in 2005 PPP US Dollars is drawn from a highly disaggregated spatial data set on
population, income and poverty constructed on the basis of national census data by the
World Bank (2009) for the World Development Report on Reshaping Economic Geography.13
Modern county level US income in dollars is mean household income taken from the 2000
US Census.



2.3       Controls

As controls in our regressions, we also employ data on:


                                                                                un
   Literacy : Aggregate literacy rates we take from (Mariscal & Sokoloff, 2000; N´ ˜ez,
2005). Sub-national literacy rates for Argentina, Chile, Mexico, Venezuela, and the US
taken from same census data as above. County level from 1880 census.


   Secondary Schooling : Similar to Goldin & Katz (2011); Goldin (1999) for the US we
construct a proxy for secondary schooling by the share of 14-17 years olds who report that
they are attending school (1880 US census).


   Higher Level Non-Engineering Human Capital : For the US, this is measured as number
of lawyers and medical doctors per 100,000 inhabitants (US Census 1880). We also aggregate
across occupational categories the EDSCOR50 measure that indicates the percentage of
people in the respondent’s occupational category who had completed one or more years of
college. Together these allow us to ensure that our engineering measure is not proxying for
the availability of higher order human capital in general.

  13
       For further detail see Maloney & Valencia (2015)
                                                      8
   Railroads : At the national level, we employ the density of railroads measured as
                                                            on & Ram´
kilometers of track per 1000 square kilometers in 1900 (Pach´       ırez, 2006; Thorp,
1998).      At the sub-national level, we employ the Interstate Commerce Commission’s
data on miles of track per 100 square miles converted to the same units above for con-
sistency in 1899 (ICC, 1899). Individual country level data is not available for Latin America.


   Manufacturing output per capita : The value of manufactured products and labor in
manufacturing in 1870 taken from NHGIS to compute the per capita and per unit of labor
labor manufacturing product.14


   Population Density in 1900 : These are collected from census data from the individual
countries. Argentina (1895), Brazil (1900), Chile (1907), Colombia (1912), Mexico (1895),
Peru (1876), Venezuela (1926), and the US (1900).


   Pre-colonial Population Density : This measures the estimated number of indigenous
people per square kilometer just before colonization. See Maloney & Valencia (2015) for
more detail.


   Slavery : As a measure of institutions that is available for the United States sample, we
used the 1860 Census as well as the data compiled in Nunn (2008).


   Geographical Controls : in addition to the set of sub-national geographical variables
collected by Bruhn & Gallego (2011) including temperature, altitude, and annual rainfall,
we add a measure of agricultural suitability and river density as developed in Maloney &
Valencia (2015). Clearly, populations could also be sustained by marine-based economies
where farmland and rivers were of less importance.        Hence proximity to the coast for


  14
       https://www.nhgis.org/documentation/gis-data
                                                      9
saltwater trade, transport, fishing potential and amenities potentially persists in importance,
much as it was subsequently for European settlement, and to capture this we employ a
measure of distance to the coast as calculated by Gennaioli et al. (2013). We further include
a measured of ruggedness of terrain from Nunn & Puga (2012).


     Table 1 presents the summary statistics at the state level and table 2 at the US county
level.



3        Historical Innovative Capacity as a Potential Driver
         of Present Income Differences: National Data
     Figure 1 plots our measure of national innovative capacity against GDP per capita,
both in 1900.15 The availability of data means that our effective sample going forward is
restricted to the relatively larger countries depicted here. Several facts merit note.


     First, there is substantial variance in the stock of engineers that is weakly related to
income in 1900. The Northern United States with a density of 160 is the highest in our
sample, roughly double the average for the country as a whole, 84, while the US South
shows over a third of the engineering density of the North at 60. Lagging as it is, the
American South is miles ahead of the Latin American countries who average under 20.
What is most striking is that countries that we tend to associate with declining relative
position across the previous century, especially Argentina and to a lesser degree Chile and
Mexico, show densities below countries of very similar levels of income: Argentina and Chile
had roughly the same level of income as the American South, Sweden and Denmark yet
had roughly a third of its engineering capacity of the South, and a fifth of the Scandinavian

    15
      Maddison does not tabulate a separate series for the American South, but Mitchener & McLean (2003)
estimates place the US South roughly 50% below the national average and New England 50% above. Imposing
these differentials on Maddison’s data, places the South roughly 15% higher than Spain and the North roughly
triple. Clearly, issues can be taken with even Maddison’s Herculean effort, however, the available alternatives
do not suggest that the picture would change much. Prados de la Escosura (2000) PPP based estimates with
the OECD correlate .89 with those of Maddison, and do not significantly change the level of Spain relative
to the US, although they move Portugal up perhaps 40%, now above Mexico but still 20% below Spain.
                                                     10
countries (100). Even if the number of engineers were underestimated by a factor of two,
the lag with the US and Scandinavia would still be dramatic. We argue that that natural
resource rents, while elevating income, were not being deployed as they were in the US or
Scandinavia to the development of innovative capacity that would prepare them for the
next phase of industrialization. In Howitt & Mayer-Foulkes (2005)’s framework, we have
countries with similar levels of Schumpeterian backwardness, but with radically different
levels of absorptive capacity.


   Second, the dominance of the US in the Western Hemisphere is clearly not being driven
by some idiosyncratic US data issue that would exaggerate its density. The US average
is broadly in the same league as Denmark and Sweden, and even the North is below the
                      om (1982) for France (200) and Germany (250),the frontier countries
calculations by Ahlstr¨
of the era. Nor is some sort of idiosyncratic data issue driving the consistently low scores of
Latin America. All cluster very near each other and the colonial mother countries.



3.1    Consistency with historical evidence

Our engineering estimates are consistent with historical evidence. France and Germany
were acknowledged leaders in the sciences and engineering. The relative positions of the two
peripheral areas- Scandinavia vs. the Iberian peninsula correspond closely to Landes (1998)’
characterization of their attitudes towards science and the enlightenment. Both Sweden
and Denmark’s institutions of higher technical learning date from the 1700s. Sweden’s high
density is consistent with the characterization by Sandberg (1979) of the country as the
“Impoverished Sophisticate.” The overproduction of engineers led many to emigrate to
the US and 19th century Swedish engineers are credited with inventing the blowtorch, ball
bearings, ship propellers, the safety match, the revolver, the machine gun, dynamite, and
contributing to the development of bicycles, steam turbines, early calculators, telephony
(Ericsson) among others.



                                              11
   The US started relatively early and energetically in the training of engineers.       The
first institution of engineering education emerged from the revolutionary war at West
Point, established in 1802, which trained engineers for both military and civilian purposes.
Subsequently the American Literary, Scientific and Military Academy at Norwich, Vermont
awarded its first Civil Engineering degrees in 1837, and Rensselaer School in New York,
in 1835. By 1862 there were roughly a dozen engineering schools in the East, but also
as far west as Michigan and south as Maryland. The Polytechnic College of the State
of Pennsylvania, founded in 1853 granted degrees in mechanical engineering in 1854, and
mining engineering in 1857.


   The passage of the Morrill Land Grant Act in 1862 led to an acceleration in the
establishment of engineering programs, roughly sextupling the number in the decade after
passage.   The Act led to the establishment of the Columbia School of Mines in 1864,
Worcester Polytechnic in 1868, Thayer School of Civil Engineering at Dartmouth College in
1867, Cornell University as well as new Universities in Iowa, Nebraska, Ohio, and Indiana.
It also gave impetus to the foundation and consolidation of engineering schools in the South.
As early as 1838 the University of Tennessee was teaching courses in Civil Engineering,
but in 1879 it began awarding doctorate degrees in Civil and Mining Engineering. Texas
AM awarded its first degree in Civil Engineering in 1880, Virginia Tech in 1885 in Mining
Engineering, and the University of Kentucky, although having an engineering program
dating from 1869, graduated their first civil engineer in 1890. Auburn University in Alabama
began its engineering program in 1872, and North Carolina State in 1887. In sum, the
post-Civil War period saw the expansion of engineering education throughout the country.
We further explore the use of the Morrill Land Grant as an instrument in subsequent sections.


   It also saw a deepening, with the profession in the U.S. diversified further into sub-
branches. For example, the University of Missouri established both Civil and Military
Engineering departments in 1868, and the first department of Electrical Engineering in

                                             12
1886. The establishment of professional societies in Civil Engineering (1852), Mining (1871),
Mechanical (1880) and Electrical (1884), testifies the the consolidation of a process of
specialization and diversification. By 1890, a modern and world class engineering profession
was firmly established in the US.


   Canada’s degree of sophistication is probably higher than that suggested by the density
numbers. Although the first graduates were in the 1870s, substantial engineering courses
were in place by the 1850s. Further, the articulation of the different fields of engineering
occurred later than in the US but not much.16 It is also the case that the four principal
Canadian Universities emitting graduates- McGill, University of Toronto, Ecole Polytech-
nique in Montreal, and Queen’s University in Kingston Ontario- lay within a circle of 350
mile radius with Cornell University at its center, and that includes many of the principal
US departments of the time. Hence, Canada was likely part of the greater New England
scientific community.


   The data for Latin America are consistent with the observation by Safford (1976) in
his classic Ideal of the Practical that “Latin American societies in general, and the upper
classes in particular, have been considered weak in those pursuits that North Americans
consider practical, such as the assimilation, creation, and manipulation of technology and
business enterprise in general”(p 3). The national scientific establishments and professional
training of civil engineers appeared much later and on a smaller scale. As an example,
perhaps the richest country in Latin America at the time, Argentina, began graduating
engineers only in 1870, and Peru, one of the premier mining centers of the hemisphere, in
1880, roughly on the same time line as Alabama. Further, in countries like Colombia and
Mexico, political instability undermined programs begun relatively early leading to very low
levels of graduation. In addition, the process of diversification and specialization was not as
advanced as was the case, for example, in Missouri. General engineering associations were

  16
     The development of, for instance, mechanical engineering as a separate course occurred about 20 years
after in the US, and Electrical Engineering 10-15 years after.
                                                   13
set up in many countries around the same time that, in the US, associations in individual
sub-fields were established.


   Graham (1981) argues that Brazil, consistent with our estimates, lagged far behind
the American South in every aspect of industrialization, transportation and agricultural
technology (p. 634). In agriculture, there was little use of plows, scrapers, cultivators or
mechanical seeders until the 20th century, partly because the low level of literacy rendered
pointless the agricultural journals found commonly in the American South. A long literature
has focused on Argentina’s weakness in innovation effort in comparison to countries such as
Australia and Canada seen as similarly endowed. (see, for example Diaz Alejandro, 1985;
Duncan & Fogarty, 1986; Campante & Glaeser, 2009). The determined efforts in both these
countries to achieve widespread literacy in the prairies had no analogue in Latin America,
nor did the extensive expansion in the form of experiment stations, seed testing services,
and technical assistance.



3.2    Aggregate correlations

In sum, the historical evidence confirms that lagged as the American South was relative to the
North, as Carnegie suggests above, Latin America lagged even more. As a quick check, basic
correlations suggest that these differences in engineering densities for 11 countries are indeed
correlated with present income. Since we have population density and income variables at
the subnational level, we run a simple regression as a panel and bootstrap the clustered SEs
at the country level (the results also hold with simple OLS):




       Y2005,ij = α + γE Eng1900i + γpop P op1900,ij + γR Rail1900i + +βL Lit1900,i +   ij (1)




where the variables are defined as above for country i and sub-national unit j: the dependent
variable is income per capita today; the explanatory variables are Literacy, Engineering

                                               14
density in 1900, Railroad density in 1900, Population density in 1900, and a set of geograph-
ical controls. Table 3 shows that engineering in 1900 appears significantly in explaining
today’s income per capita despite controlling sequentially and then together for population
density and railroad density. Were we to include Denmark and Sweden, whose density we
have confidence in, not to mention France and Germany, the results would be even stronger.
Columns 4 suggests that with few observations, it is impossible to separate the impact of
engineering and literacy although in subsequent exercises we can.




4         Establishing the Importance of Engineering and
          Patenting Capacity: US County Level Data
     We construct engineering and patenting density data for 2380 US counties.17 Table 4
presents the OLS estimates including the core controls for geography, economic activity and
education which reduce the number of observations to 1905. All specifications have robust
errors and are clustered at the state level and include both OLS and fixed effect estimators.
We estimate:




Y2000,i = α+γE Eng1900,i + γP P at1900,i + Geoi γGe + Y1900,i γGr + HC1900,i γH + µstate + i (2)


To be consistent with the other samples, we begin with an OLS specification of engineering
density along with a vector of geographical controls. Errors are robust and clustered by
state.       In column 1, engineering emerges strongly significantly and of expected sign in
the OLS specification. Temperature and distance from the coast both enter negatively
and significantly. Column 2 replicates the specification but with the smaller sample lim-


    17
         A small number of counties change their borders in the last century and we drop them.
                                                       15
ited by slavery and education and results in only minor change in significance and magnitude.


   Column 3 then includes the set of income related controls that together work to reduce
the correlation that may occur simply because denser, richer areas with more infrastructure
and in particular manufacturing and railroads, may be correlated both with engineers in 1900
and also with income today through non-innovation related channels. We further include
a measure of slavery as an institutional variable that may be correlated with low human
capital but also provides other institutional channels to present income. Population density,
our measure of lagged economic activity, and railroads enter positively and significantly and
will remain so throughout the analysis. Slavery, which enters significantly and negatively
with just geographical controls (not shown) loses significance.


   The next 3 columns seek to ensure that the engineering measure is not simply a proxy
for human capital more generally. Column 4 includes literacy which enters positively and
significantly and reduces the engineering coefficient by just over 10%.             Column 5 then
includes a proxy for secondary education whose importance across the next century Goldin
& Katz (2011) and Goldin (1999) have stressed and which enters significantly and positively.
Goldin & Katz (1999) also discuss the importance of the expansion in higher education
across the earlier 1890-1940 period and in Column 6 we include several variables to ensure
that engineers is not picking up higher order human capital more generally: the share
of individuals who report attending at least one year of college, the density of lawyers
(see Murphy et al., 1991), and density of physicians. The first two enter significantly and
positively. Finally, column 7 includes all the human capital variables together. Literacy and
lawyers remains significant although now neither secondary education, the college proxy,
nor physicians do. Either literacy or college attendance can eliminate secondary education
suggesting that, at this point in history, it is broadly proxying for literacy, or perhaps college
readiness. The same remains true in column 8 which controls for state fixed effects. Slavery
now enters positively and significantly. This may imply that it was picking up a negative

                                               16
“Southern” effect which fixed effects now eliminate, but also that its negative impact works
importantly through that channel.


   In all cases, engineering retains its significance at the 1% level although losing perhaps
50% of its magnitude from the initial specification. The relative robustness of engineering
vs. secondary education suggests that a small well-educated technological class was, at this
time, more important than broad based post-literacy skills. Perhaps counter intuitively,
the consolidation of higher education in engineering occurred before the major expansion of
secondary education which Goldin & Katz (2011); Goldin (1999) document rose from only
18% in 1910 to 71% in 1940. Across our period, the corresponding average for the country
is 7% with a maximum of 26%.


   Table 5 next explores the interaction of the engineering term with patents. Annex Table
A1 suggests that engineering density is significantly correlated with patenting, but it explains
perhaps 5% of the variance across counties and is not robust to instrumenting. Further, even
with a complete specification including all measures of formal human capital, we explain
50% of the variance. This suggests that, as in Sokoloff (1988) there may be an particular
entrepreneurial or inventive human capital not captured by our engineering term. Columns
1 and 2 of Table 5 repeat the last two full specifications but replace engineering density with
patenting density as the innovation capacity measure. Patenting enters significantly at 1%
again with and without fixed effects. The human capital variables enter similarly to before.
Columns 3 and 4 combine engineering and patenting in the full specification and both
remain significant at at least the 5% level with and without fixed effects. Both measures
decline 15-20% in magnitude suggesting that one channel through which engineering density
operates is through patenting (as suggested by table A1) and that some of patenting’s
effect is as a proxy for engineering density perhaps working through non-patenting channels.
However, that each continues to enter very significantly and independently suggests that they
are capturing different types of human capital: perhaps more technological adoption in en-

                                              17
gineering and inventive activity in patenting. We fully accept the argument of Moser (2013)
that much innovation including patentable invention is done outside the patent system and
hence such distinctions cannot be drawn too sharply. However, our results suggest that there
is a particular type of human capital captured by patents not captured by our other measures.



4.1     Instrumenting engineering

Though we have attempted to control for other channels through which our engineering
measure may be correlated with present income beside the capacity to manage and generate
new technologies, we attempt to control for residual upward bias by instrumenting engineers.
There may be additional bias in the opposite direction since the large share of counties
reporting zero engineers is more likely to reflect measurement error, rather than that there
was absolutely no capacity of any kind to adopt technologies.18


    In the spirit of Moretti (2004), we also attempt to instrument engineering density using
the log distance to the nearest Morrill Land Grant colleges and universities. As discussed
earlier, the Morrill program was introduced in 1868 precisely to remedy the perceived
shortfalls in regional technical assistance in agricultural and mechanical innovation. It was
to an important degree supply driven. Its motivation was driven by both democratic ideals
of educating the working man, and by observation that the US had none of the agricultural
and technical schools found in Europe (Nevins et al., 1962; Nienkamp, 2010). Early years
saw struggles over definition of what form that education should take, and precarious finance
suggesting little initial impulse from the private sector. Further, prior to the Civil War, the
South had actively opposed the bill, fearing greater interference in matters such as universal
primary education and only the withdrawal of the Confederate States from the US Congress


  18
     50% of the countries report zeros. The vast majority are found in small counties which, in 1900, may not
have had the scale to generate an engineer. The maximum county size is 1,240,403 with the 99th percentile
at 140,000. 88.9% of zeros are found in counties below 20,000 of of counties with 5,000 inhabitants, 75%
are zeros. The zeros also pose an issue in comparability of the OLS and IV estimates since IVs offer a local
estimate over the support where there is variance in both the instrument and the variable to be instrumented.
In the present case, this may also push up the IV estimates.
                                                     18
allowed the bill to be passed. However, during Reconstruction, recognizing its technological
lag, the South started privately institutes such as Georgia Tech, and actively embraced the
Morrill Program. Hence, while public functionaries may have broadly felt that the structure
of their economies would reward such investment (although the long absence of agricultural
institutes suggest this is far from obvious), there is little evidence of organized industry
lobbying, for example, that would suggest causality from industry demand for engineers to
the establishment of these universities.


   Though the promoters of the Morrill program did not envisage it as fomenting engineering
per se, the program would eventually finance the first engineering departments in the West,
Midwest and Especially the South and was more fundamentally a central driver in developing
the higher order scientific and Engineering capacity of the country. Nienkamp (2010) in
Land-Grant Colleges and American Engineers argues that, especially the Mid-Western
schools “provided the foundation, both in training and number, for twentieth-century
American professional engineering” and were critical in defining their identity and ushering
in the modern scientific, laboratory-based approach to technical education. Nevins et al.
(1962) argues that the Morrill Program “promoted the emergence of the most effective
engineering schools on the globe.” Goldin & Katz (1999) notes that while the majority of
new universities of the time were privately started and Morrill cannot explain the expansion
of higher education, as of 1908 they also note 60% of the nation’s engineering students were
found in public universities and the “geographical dispersion of engineering students came
mainly from those enrolled in public schools” the rise of which they attribute to the Morrill
program (Goldin, 1999). Hence, the instrument is conceptually closely linked to engineering
density, while the its amorphous early motivation and supply driven nature can explain, for
instance, the lack of a significant positive correlation with manufacturing activity, at the
county level.


   The first stage of a basic 2SLS reveals a strong positive correlation between engineering

                                             19
density and the Morrill proxy with very high Cragg-Donald F-tests (1% level) and acceptable
F-tests suggesting a strong instrument. Column 5 presents the second stage results for
the complete specification with fixed effects and engineers enter again at the 1% level with
a value an order of magnitude higher than the OLS estimates. We do not present all
previous specifications simply because, as with the OLS specifications, both engineering and
patenting retain their significance and broad magnitudes throughout.


   Column 6 presents the results of a Buchinsky quantile regression model with selection
into being a zero on the instrument stage. This specification is also desirable because it
relaxes the assumption of joint normality implicit in the standard Heckman model. The
first stage again suggests a strong instrument at the 1% level. The coefficient on engineering
falls to half the 2SLS, 3.51, and maintains its strong significance. As before, the first stage
is strongly significant.


   In the spirit of the next section that works at the “state” level, we replicate the
exercise for the 1900 5% subsample (again, which does not permit us to work at the county
level), with the number of covariates reflecting the reduced degrees. This higher level of
aggregation allows for more mobility of labor among unit of analysis, although, clearly, this
is not preventing the emergence of very significant coefficients. Second, it allows another
20 years of impact of the Morrill program and the full consolidation of the engineering
infrastructure. Third it allows introducing a measure of mining activity which is available at
the state, but not country level. Finally, the aggregation also offers an alternative approach
to the zeros. The OLS and instrument estimates are 1.2 and 2.1 respectively and significant,
even with 48 observations (Annex Tables A2 and A3).


   Taken together, the US data suggests that innovative capacity, spanned by engineering
and patenting density, has a strong and independent effect on future income levels. More
specifically, running specification 6 in standard deviations suggests that a one standard

                                             20
deviation rise in engineering density leads to a 16% increase in income.          Though not
instrumented, patenting capacity leads to another 9.7%. Hence higher level human capital
plausibly accounts for large differences in state level income per capita. This does not imply
the unimportance of lower level capital. A 1 SD increase in literacy accounts for 51%. A
comparable increase in secondary schooling leads to only a 4.5% increase although, again,
the high school movement began substantially later and the variance in 1880 is very small.
As a final back-of-the-envelope exercise, taking the coefficient from the well saturated US
regressions and applying it to the international data in Figure 1, a one standard deviation
rise in engineers in 1900 leads to an difference of 36% in 2000 income. Clearly, we cannot
take this result too literally, but again, it suggests that differing endowments of higher level
scientific capital potentially contribute importantly to income differences.




5     International sub-national engineering data
    To see whether innovative capacity retains its predictive power beyond the US, we collect
engineering data at the sub-national (state) level for Argentina, Chile, Colombia, Mexico,
Venezuela and the US for which census data are available. To give a feel for the disparities,
Figure 2 maps this data for the US and Mexico by decile of engineering density and
strikingly confirms that the border divided worlds apart. In fact, the data likely understate
the true difference since our calculated stocks in Figure 1 suggest substantial overstatement
in the Mexican census data. Perhaps predictably, the advanced New England states and
the heavily mining dependent and generally less populated Western states show the highest
density while the emerging industrial centers of the Midwest are close behind. The South is
concentrated in the lower ranks with South Carolina, Georgia, Arkansas and Alabama in the
bottom deciles- the lowest density in the US. What is striking, however, is that the country
that was the principal mining center of the Spanish empire is almost entirely concentrated
in the first and second quintiles with Sonora and the two Baja Californias appearing in the
third and fourth deciles. Taking out Mexico City and the border states, Mexico is almost

                                              21
uniformly below even the American South in density of engineers. As we discuss next,
despite four centuries of mining, it had not acquired a corpus of trained professionals in the
field compared to relative newcomer, the American West. The other countries show similar
patterns.


   Table 6 employs the sub-national data and comes to similar conclusions to those previous.
We estimate:


            Y2005,ij = α + γE Eng1900ij + γpop P op1900,ij + γL Lit1900,ij + µi +   ij (3)



Engineering density is persistently significant and broadly of the magnitude of the OLS
estimates from the US. The finding of continued significance even after controlling for
a measure of population density (column 2) is consistent with Comin et al. (2010a). In
column 3, Literacy again enters positively and significantly and reduces the coefficient on
engineering by 40%, again, broadly consistent with the US results. Though we cannot
control for a more complete set of human capital measures, the county level exercises above
suggest that, in fact, engineering is capturing innovation related human capital and not
human capital accumulation more generally. We also run the regressions with a complete set
of geographical controls but these absorb substantial degrees of freedom and do not enter
significantly so we omit them.


   In sum, the various estimations suggest that our county level US findings resonate
internationally: both literacy and higher order human capital related to engineering and
science and technology are important determinants of long run income.



5.1    Mechanisms of influence today

Table 7 offers some suggestive mechanisms through which these engineering densities in 1900
could continue to affect output today. We collect five indicators at the country level that


                                                22
broadly capture modern day innovation-related inputs and outputs and calculate the simple
correlation. The first is the dynamism of the system of research and development measured
as total R&D expenditures as a share of GDP. The second asks about firm capacity for
innovation ranging from pure licensing to pioneering their own new products and processes.19
These two, arguably, correspond most closely to Howitt & Mayer-Foulkes (2005)’s R&D
model. The third draws from a globally consistent measure of management quality from
Bloom & Van Reenen (2010) and in particular, the sum of the scores on the two questions
dealing with how firms identify new production processes to adopt. On the output side,
we have Comin & Hobijn (2010); Comin & Ferrer (2013); Comin et al. (2008)’s measure
of technological adoption at the extensive margin, averaging their industrial and sectoral
scores,20 and finally patent applications filed under the Patent Cooperation Treaty (PCT)
per million population as tabulated by the World Economic Forum (World Economic Forum
et al., 2012). Together, these give an impression of measure of national absorptive and
inventive capacity in the present.             In virtually every case, the correlation between our
engineering numbers in 1900 and these indicators today is above .9.




6         Support from History: Case Studies
This section offers historical evidence that confirms that innovative capacity was a critical
barrier to taking advantage of the advances of the Industrial Revolution. The interaction
of lost learning by doing, weak higher level human capital, and underdeveloped technical
institutions emerges in explanations of the lag of the US South as well as Latin America.
However, it also offers support for more complex view of how innovative capacity affects
steady state growth. Numerous models exist for modeling the micro economics of adoption.
Comin et al. (2010b); Comin & Hobijn (2010); Comin et al. (2010a) for instance are closely
aligned with the opening stylized facts (theirs) about divergence at the intensive margin.

    19
     “In your country, how do companies obtain technology? [1 = exclusively from licensing or imitating
foreign companies; 7 = by conducting formal research and pioneering their own new products and processes].
    20
         Their data on the arguably more relevant intensive margin is not yet available
                                                        23
Human capital shortfalls are embedded in a scalar that reflects barriers to adoption for
the agent that adapts the technology to the idiosyncrasies of the country or for individual
producers that find a profitable use for the technology. The models of Howitt (2000);
Aghion et al. (2005); Howitt & Mayer-Foulkes (2005) add a feature central to this paper-
that as the technological frontier shifts out, the skill level required to maintain the same
level of absorptive capacity increases as well. Howitt & Mayer-Foulkes (2005) further add
that 1. the efficacy of education is a function of the level of technological advance and 2.
the introduction of a new method of technological change, loosely termed “modern R&D”
such as culminated in the late 19th century with the modern R&D laboratory (the rise of
institution such as government research agencies, scientific academies, universities with close
ties to industry etc.) gives rise to the possibility of an important and discrete shift in the
productivity of innovation technology. This gives rise to three equilibria: countries with with
a threshold level of skill could undertake this “modern R&D”’ and innovate; countries with
less may still adopt and grow but with a persistent income gap relative to the innovators;
countries with insufficient absorptive capacity cannot even adopt and stagnate (see also
Howitt (2000) for a model of complete stagnation). Further, advances in the frontier not
accompanied by a rise in local human capital can cause a country to lose its absorptive
capacity and slide to a worse equilibrium.


   Our data do not permit testing explicitly for convergence clubs and we must be satisfied
with a correlation with present income. However, the historical case studies do suggest the
importance of these dynamics and these ideas help organize what happened around 1900
in the US South and Latin America vs. the US North. The initial conditions in terms of
skills broadly defined allowed the latter to fully adopt modern R&D technologies, while in
the former, they did not. In the South they were unable in some cases to undertake the
necessary R&D to adapt some new technologies to local conditions, for instance, in steel.
In Latin America, the erosion of their frontier adjusted human capital was so severe when
faced with new technologies in metallurgy and chemistry that they were forced to abandon

                                              24
critical industries, in our case mining, altogether and could not, in many cases, mount a
manufacturing effort.



6.1     The American South

An extensive literature deals with Southern development and we present only a brief
summary to draw the parallel with other cases. Wright (1986) casts much of his work
explaining the US South’s persistent lag exactly in terms of an innovative capacity frame-
work. “The fundamental reason for [protracted lack of uptake of technologies] is that early
industrialization is a matter of learning in the broadest sense of that term: in management,
in technology, in marketing and certainly-though this is often underestimated- in learning
on the part on the part of the labor force”(pp. 124-125). The South came relatively late
to industrialization and lacked the indigenous technical capacity necessary for rapid catch
up.21 The emblematic failure was of the Birmingham Steel industry, which Carnegie referred
to. The problems were manifold-high labor costs, product quality and marketing- all of
which reflect the low level of collective, accumulated learning by doing. But also central was
that the low iron, high phosphorous nature of Alabama red hematite required substantial
adaptations of technology to the Southern context which the local innovative capacity was
not able to engineer. Nor was it able to develop Southern versions of new inventions in the
paper and textile industries. By way of contrast, Japan also initially imported technology
and processes, but over time it generated distinctive technologies and, by the 1920s,
was making its own textile machinery. In lumbering and iron making, as well, Southern
producers of the 1920s were not only not innovative, they were using methods phased out
decades earlier elsewhere. Arguably, the big push by the federal government, ranging from
the Land Grant colleges to selective location of advanced industries, and migration of higher
order human capital, raised innovative capacity toward the frontier and permitted catch up.


  21
     Wright (1986) argues that “Having missed the formative phases of the ‘American System’, the South
was lacking a machine-tools and capital-goods sector almost entirely and therefore was bypassed by the kind
of adaptive, dynamic, path-breaking series of technological breakthroughs that made ‘the American system’
distinctive.”(p. 124-125).
                                                    25
6.2     Multiple equilibria in mining: Latin America vs. the US

The potentially catastrophic impact of a dearth of innovative capacity is nowhere more in
evidence than in the industry in which Latin America for centuries had a true comparative
advantage, yet by the turn of the 20th century had completely stagnated: mining. Close
observation by numerous historians offers a particularly compelling window on how a
similar inability to adapt new technologies in the face of declining ore quality nearly
destroyed the Chilean mining industry as well. As Figure 3 illustrates, Chile saw its world
market share of copper fall from 40% to under 4 percent by 1911, and even as early as
                          ıa (Mining Association) wondered openly whether Chile’s
1884 the Sociedad de Miner´
copper mines would survive at all (Collier & Sater, 1996). Chilean historians date this
technological slippage to the beginning of the nineteenth century, when “the work of mining
was not very systematic” and the “receipt of industrial innovations [from abroad] was
slow and without visible influence”(Villalobos et al., 1990, p. 95-96).22 One of Chile’s
most venerated historians, Francisco Encina noted that “from the point of view of capital
and of technical and administrative aptitude, the copper industry is as demanding as the
most complicated manufacturing industry” (Encina, 1972, p. 62). However, his studies
revealed “an extraordinary economic ineptitude in the national population consequence of
an education completely inadequate to meet the demands of contemporary life.”(p. 17).
Another prominent historian, Pinto Santa Cruz (1959) argued that Chileans failed to take
advantage of opportunities for learning by doing and to evolve the innovative capacity
required to confront the technological revolution in mining and hence became dependent on
foreign firms.23

   22
      Charles Lambert, a representative of a British mining company in La Serena who was trained in the
 ´
Ecole Polytechnique in Paris, noted in 1819 the primitive mining practice, scarce knowledge of minerals, and
inefficient smelting, all of which represented poor technique relative to that employed in Europe. See also
Maloney (2002).
  23
     “The technological demands of the period, in contrast to what is occurring today in some areas of mining
or industry, were relatively modest and thus not too costly. What could and had to be done in the national
mining companies and in agriculture was perfectly compatible with the resources accumulated in the long
                                                     26
    Figure 3 captures this importing of technical ability, first in the 1870s and 80s as Britain
expanded in the nitrate and coal industry, and then after 1905 when major foreign copper
companies took over the copper industry. The census data shows the density of foreign
entrepreneurs increased from almost zero in 1870 to 80 in 1920, the last census which
permitted disaggregating by nationality/profession, or roughly 4 times Chile’s domestic
density in 1900. Local capacity was increasing as well, however, the share of foreign engineers
in the country total increased sharply from 1% of total engineers in the country to around
40% across this period. These movements emerged concomitant with Chilean production
reaching new highs and market share of an expanding global supply staging and important
recovery. As in Mexico, below, and consistent with Howitt & Mayer-Foulkes (2005)’s edu-
cation externality, both the quantity and the quality of the engineering graduates produced
by local universities were thought inferior to the talent imported from Europe and the US
                               na, 1993). Increasing frustrated by the creeping influence of
(Serrano, 1993; Bazant de Salda˜
foreigners across the major industries, it is perhaps not surprising that Chileans developed a
                                                                      on, 1982, p. 62 and
self-perception that they were perhaps “unfit for the modern era”(Monte´
35).24 By 1918, American interests controlled 87% of Chilean copper output (O’Brien, 1989).


    Similar stories of an inability to exploit new technologies leading to decline in the mining
industry can be found throughout the region. The engineering data in Figure 1 supports
the historical evidence that in Mexico local entrepreneurs lost share in the industry they
had dominated for centuries precisely due to lacking the capacity to master emerging


periods of bonanza. If the process had been initiated and maintained adequately, without doubt it would
have created the means to confront more challenging tasks, such as those posed by copper mining when it
was necessary to exploit less rich veins. However, faced with the technological revolution, the local mining
companies did not have either sufficient accumulated resources or organizational and administrative capacity-
both of which were indispensable. In these circumstances, there was no other option but the introduction of
foreign capital and expertise.”Pinto Santa Cruz (1959)(p 71)
  24
     Prominent intellectual Tancredo Pinochet Le-Brun(1909), granting that Chileans were inferior to
Europeans, still wondered, “Don’t we have minds in this country that can go to Europe to learn what
professors, whom we have imported and continue importing, have studied? Are we truly incapable of steering
our own ship?” La Conquista de Chile en el Siglo XX, Santiago, La Ilustracion, page 81 cited in Monte´  on
(1982)).
                                                    27
technologies (Ruiz Larraguivel, 2004; Brading, 1971; Marichal, 1997). Even in Zacatecas,
                 ı, and Guanajuato, long centers of mining, engineering density was at low
San Luis de Potos´
levels compared to the newcomers in the US West. Around 1900, abandoned, underexploited
and newly discovered mines fell to foreign hands that could bring new global technologies
to bear. As an example, the Guggenheim interests opened smelters in Monterrey (1892)
and Aguascalientes (1894) purchased the largest Mexican Smelting and Refining company
in 1906, introduced modern methods of extracting and refining silver ores and in addition,
started the production of lead and zinc mining (Bernstein, 1964; O’Brien, 1989). By the
early 20th century, the Americans absolutely dominated the industry.25 In Upper Peru
(now Bolivia), the decline of silver production in some of the most famous mines, like
              ı, arose from the “failure to apply new mining techniques, heavy mortality
those at Potos´
among Indian laborers and the exhausting of previous rich veins” (Scobie, 1964, p.59). In
Ecuador, Hurtado (2007) argues that the discovery of new mineral deposits was hindered by
a resistance to scientific methods.26


    The US eventual dominance of the Chilean copper and Mexican mining industries
strikingly illustrates the road that could have been taken with the same homogenous
product. Not only does Wright (1999) argue that US in the 19th century “parlayed its
[natural] resource-based industrial prosperity into a well-educated labor force, an increasingly
sophisticated science-based technology, and world leadership in scientific research itself”
(Wright, 1987, p. 665), but he uses precisely the US copper industry as an example of
national learning and of innovation as a network phenomenon. Wright stresses that in

  25
       See Maloney (2015) for data on nationality of owners.
   26
      More generally, Di Tella (1985) argues that Argentina proved unable to move beyond a state of exploiting
the pure rents of a frontier or extraction of mineral riches, and beyond the “collusive rents” offered by state-
sanctioned or otherwise imposed monopolies to tap the “unlimited source of growth” found in exploiting the
quasi-rents of innovation, as the US, Canada and Australia were able to do. They remained in an adoption or,
perhaps, stagnation equilibrium. The same phenomenon is seen in agriculture where the absence of innovative
infrastructure was recognized by contemporary Argentines as key to explaining the lagging performance of
the wheat industry compared to that in Canada and Australia. Fogarty et al. (1985) attributes to weak
innovative capacity the outcome of a quasi-experiment whereby the Spanish Merino sheep were introduced
into New South Wales, Australia, and Argentina’s River Plate region in the same year and had the same
access to European capital, but by 1885 showed yields of only half the wool per sheep in Argentina.
                                                      28
the post civil war period, the US became the foremost location for education in mining
engineering and metallurgy. It was the revolution in metallurgy (e.g. the Bessemer process
and the introduction of electrolysis on a commercial scale for the refining of copper)
overwhelmingly an American achievement, that propelled the copper industry during the
last decades of the 19th century. The transference of these technologies by US firms to their
mines and smelters in Chile and Mexico revolutionized the antiquated industries in both
countries, dramatically increased production, and left them dominant in both.27


    The US reaped high growth dividends from inventing new technologies in mining, and its
engineering and scientific capital became the best in the world. At the same time, and with
exactly the same goods, Latin America’s declining human capital relative to the expanding
technological frontier left it unable even to adapt new technologies, and the domestic
industries stagnated. The success of the Chilean State Copper Company, established in
1976, testifies that this need not be permanent and at least an adoption equilibrium was
reachable by nationals with the accumulation of the necessary human and knowledge capital.



6.3     Retrogression in industry in Brazil: Minas Gerais

In iron manufacturing, Baer (1969); Rogers (1962); Birchal (1999) argue that despite a
tradition of iron smelting dating from the mid-sixteenth century, the techniques used at the
end of the nineteenth century were primitive. While particularly the northern US colonies
engaged in a sustained process of learning by doing and innovation in both iron and steel
(Swank, 1965) from the early 18th century on, from 1830 to 1880 Brazil actually experienced
a “retrogression in technique”(Rogers, 1962, p. 183). Unable to innovate, Brazilian firms




  27
     The Guggenheim’s El Teniente mine was the first in the world to apply the flotation process in
concentrating low-grade ores. Mechanizing digging made Chuquicamata in the north the largest open pit
mine and, again, a new concentration process was introduced using sulfuric acid and electrolytic precipitation
to treat the mine’s ore. From 1912-1926, copper production in Chile quintupled as a result, reversing a 25
year period of stagnation (O’Brien, 1989).
                                                     29
instead lobbied for protection from cheaper iron imports.28               The critical innovation for
the development of the native steel industry was the foundation in 1879 of the Escola de
Minas (Mining School) at Ouro Preto, Minas Gerais, which, as the Escuela de Minas de
Antioquia was a major diffuser of new practices and led to the establishment of the first
new blast furnace since the failures at the beginning of the century. On the other hand,
the textile industry for which we have entrepreneurship data above was less fortunate. As
(Birchal, 1999, p. 183) notes “the existing informal and spontaneous technological innovative
system was not developed enough to take the process of technological assimilation farther
in the direction of a profound modification of existing foreign technologies, or to create a
more complex indigenous technological alternative” again, much as Wright argues for the
American South.


   This raises the larger question of, given the sheer disparity of our innovative capacity
numbers, if it was so hard for the American South to start competitive industries with
its innovative capacity captured by 60 engineers per 100,000 individuals, how did Latin
America’s industry expand so quickly at the turn of the 20th century with an average of 10?
Much also may lie in the Potemkin nature of much of Latin American industrialization that,
as in the American South, was dependent on imported capital goods and innovated little.
Across the region, blocks in the input-output table were rapidly filled in but, as Haber (2005,
2006) notes, these industries were not modern in the sense of being at the technological
frontier or being able to export to other countries. Much as Wright notes in the American
South, Latin America relied heavily on foreign technology imports, and was sluggish in
developing an indigenous local machine and capital goods industry, a result attributed to
low innovative capacity.29 The fact that Latin America evolved the highest levels of tariffs in

  28
     Of the thirty ironworks in the headwater region of the Rio Doce in 1879, only seven used Italian
forging methods, while the rest used the old African cadinho (crucible) technique. Graduates of the Escola
de Engenharia do Exercito (Military Engineering School) established in 1930 led the steel industry as it
developed through the 1960s.
  29
    “Since... capital goods industries now required well-developed scientific and engineering capabilities,
Mexico had little choice but to import its capital equipment. The blast furnaces and rolling mills came
from the United States, the high-speed cigarette machinery from France, the paper-making machinery from
                                                   30
the world prior to World War I –on average five times the rates in Europe– is arguably the
result of the need to protect an industry erected on weak foundations of entrepreneurship,
accumulated learning by doing, and higher level human scientific capital.30 This supports
Haber’s theory that the standard view of protectionism stimulating industrialization is
potentially backwards.          An emerging, but technologically backward and uncompetitive
industry demanded protection. In turn, lack of exposure to foreign competition blunted the
need to upgrade quality and technical capacity.




7         Drivers of innovative capacity

7.1         The US

Though a detailed examination of the factors determining innovative capacity at the turn of
the century are beyond the scope of the paper, we briefly explore select potential drivers.
Table 8 presents, first, the results for the US county patents for both engineers and patents.
As a predetermined measure of agglomeration or economic activity, we introduce population
density prior to colonization. We further introduce slavery as well as a southern dummy (at
the state level), plus geographical controls. Because of the zeros in engineers, we present
both OLS (Column 1) as well as quantile estimates with selection to report a zero (Column
2). Column 3 reports OLS for patents and Column 4 the corresponding quantile regressions.


     Somewhat strikingly, pre-colonial population enters strongly significantly and positively
in all specifications with both engineers and patents. This offers a channel through which
agglomerations and arguably economic activity persist over time (see Maloney & Valencia
(2015). Slavery enters negatively and very significantly in all specifications confirming that
reduced investment in higher level human capital is a channel through which slavery had


Switzerland, the textile looms, spindles, and other equipment from England, Belgium, the United States, or
Germany.” Haber (1997), p. 18
    30
         Coatsworth and Williamson and Clemson and Williamson, cited in Haber (2005).
                                                     31
a depressing impact on future incomes. South continues to enter significantly in all except
the engineers selection model suggesting an impact beyond the set of factors associated with
slavery. The geographical controls tell a mixed story. For engineers, only ruggedness enters
positively while for patenting it enters negatively, perhaps reflecting the importance of
engineers to mining, but not necessarily the likelihood that mining would generate patents.
Temperature enters positively and significantly, and rainfall negatively in almost all specifica-
tions. For patents, altitude enters positively and significantly, and coastal distance negatively.



7.2    The role of colonial heritage?

The degrees of freedom available in the international data are insufficient to generate
comparable results with any degree of confidence.          However, the remarkable similarity
between the engineering density in Spain and Portugal and their colonial dependencies
is consonant with the inheritance of values and institutions related to educational and
entrepreneurial capital formation from the Peninsular powers to their colonies, and in
striking parallels in development outcomes among them. Spain and Portugal, confronted
with the Industrial Revolution in England, failed to develop a scientific vocation nor the
complementary innovation promoting institutions essential to industrialization.         In fact,
strikingly Spain’s own development path shows the exact same issues of inability to manage
new technologies and external dominance characterizing Latin America (Tortella Casares,
2000) suggesting a high degree of commonality in underlying factors. For example, the
evolution of the Spanish mining industry precisely foreshadows the subsequent experience
in Chile and elsewhere discussed earlier. Though Spanish mines were rich, and mercury had
been worked for a thousand years, the lack of technical capacity and capital, and the slow
growth of domestic metallurgical know how led Spanish entrepreneurs to work close to the
surface and then sell out to foreigners once easy veins had been exhausted. In 1873, a UK
and German led conglomerate purchased the mines on the Rio Tinto river in Andalucia,
introduced new technologies, and from 1877-1891, became the world’s largest producer of
copper, contributing the to fall in Chile’s global share in Figure 3. As (Tortella Casares,
                                               32
2000, pp. 96, 213-215)) summarizes: “extraction and processing constitute a classic example
of the failure of Spanish entrepreneurs to confront the problems of developing an indus-
trial sector with complex technology, intensive use of capital, [and] a fast-expanding horizon.”


    The colonies, legally cut off from direct commerce and interaction with Northern Europe-
even trade with Cadiz was through peninsular intermediaries- were unlikely to absorb an
alternate set of institutions and values from the industrializing center or evolve indigenous
analogues, as the US did. Mass education was not prioritized on either side of the Atlantic:
in both mother countries literacy lagged even the black (often freed slave) population in
the American South, and is of similar magnitude to that of the Latin American colonies.31
The Peninsular tradition of higher education was largely religiously based, focused on law,
philosophy, and theology, proved resistant to attempts to introduce enlightenment-informed
technical curricula,32 and was exported to the colonies. As Safford continues (p. 3)“The
Latin American upper classes have been noted for their devotion to the study of law, the
humanities, and the arts and their lack of interest in the natural sciences and technology. In
the hands of the upper classes, Latin America’s educational systems, at least in times past,
have been dedicated to forming and maintaining the political elite and have been only mildly
effective in furthering such economically practical aims as the broad diffusion of literacy
and technical capacities.”33 In sum, in the patterns of human capital accumulation and

    31
       Tortella (1994) notes that in 1900 Spain had a literacy rate of 44%. This is slightly below that of
Argentina (48 %) and blacks in the American South (49%) (Collins & Margo, 2006)), and not dramatically
ahead of Peru (38%), Chile (37%), Colombia (32 %), or Mexico (22%). Portugal’s low level of literacy, 22%,
is essentially identical to that of its colony Brazil (20%).
   32
     In Portugal, the flagship university at Coimbra briefly established a more technical curriculum in 1759
under the Marquis of Pombal in the context of radical political, education and ecclesiastical reform to
modernize the state and energize the economy. However, in 1777 there was a reaction against the reforms
and the emphasis on natural science was abandoned and civil law regained its ancient prestige. The Spanish
enlightenment in the same period saw the establishment of groups of autonomous sociedades econ´       omicas
(economic societies) that sought to diffuse technology from abroad and establish libraries throughout the
country, as well as some Royal Societies emphasizing applied science. But Spain began training engineers
seriously only in the 1850s, and by 1867 the country had only one functioning School of Industrial Engineers,
located in Barcelona (Riera i Tu´ ebols, 1993).
   33
     Spanish America saw universities established from the moment of conquest, yet they were largely
committed to the training of ministers to convert Indians, and lawyers to staff the empire (Benjamin, 1965).
As (Will, 1957, p.17) documents for Chile, although it applies with greater generality “With the exception
of the inadequate facilities provided by a few religious organizations, there did not exist before the middle of
                                                      33
of development trajectories, the similarities are so close as to suggest a dominant role for
inherited attitudes and institutions.



8     Conclusion
    Much of the growth literature stresses differences in the capacity to generate, import, and
apply new technologies as central to explaining relative growth performance. In particular,
Aghion et al. (2005) document that different frontier adjusted endowments of human capital
broadly construed dictate very distinct growth trajectories and Howitt & Mayer-Foulkes
(2005) argue that at the time of a radical shift in technological progress, such differences
lead to convergence clubs of high growth and stagnant economies. To date, however, there
has been little data generated on the higher level human capital and institutions required
to make this possible, especially in a historical context, and none to attempt to verify its
impact on income differentials.


    This paper first generates national level engineering data for the Americas in 1900
that suggests that differences in innovative capacity plausibly can explain the disparate
development trajectories of countries which, at the time, had very similar levels of income.
This measure, compiled using graduates of local engineering universities, professional
associations and censuses, is arguably the first to offer international comparability for this
key period around the Second Industrial Revolution. We see it as a measure of higher level
scientifically oriented human capital per se and supporting institutions, as well as arguably
a more general measure of technological sophistication of the entrepreneurial class.


    To more conclusively establish the link, we employ county level data from the US and


the eighteenth century an institution capable of furnishing the youth of the colony with the barest essentials
of a secular education.” The majority of the relatively few university educated members of the Brazilian
elite at independence trained at Coimbra in the reinstated legal theological tradition (Carvalho, 1982). In
Ecuador, (Hurtado, 2007, p. 115) argues that education quality even among the elite was poor and unfocused
on practical elements. Again, he cites the American ambassador to Quito in the 1850s who found “convents
instead of presses, barracks in place of schools.” See comparable accounts in Will (1957) for Chile; Safford
(1976) for Colombia; and Lopez Soria (2012) for Peru.
                                                     34
establish the robust explanatory power of engineering density and patenting, controlling
for a variety of economic and geographic variables; measures of secondary and tertiary
education that may be correlated, as well as attempting to control directly for endogeneity.
Finally, we show the persistence of the impact of engineering density in a multi-country
panel working at the “state” level.


   We then provide historical evidence that broadly confirms the ranking of countries
emerging from our engineering data and which suggests that precisely an inability to
manage the new technologies emerging from the Second Industrial Revolution slowed
growth in the Southern US, and led to a loss of competitiveness in a sector long an
area of comparative advantage in Latin America– mining.           The fact that the US was
able to leverage this sector into a well-educated workforce, and world leadership in
scientific research while Latin America lost the industry confirms the importance of
innovative capacity, and the persistent effect of early human capital investments. It also
lends support to the idea of multiple equilibria arising from differences in innovative capacity.




                                              35
References
Acemoglu, D., Akcigit, U., Bloom, N., & Kerr, W. R. (2013). Innovation, reallocation and
  growth. Technical report, National Bureau of Economic Research.

Acemoglu, D. & Zilibotti, F. (2001). Productivity Differences. The Quarterly Journal of
  Economics, 116(2), 563–606.

Adkins, D. (1975). The great American degree machine: an economic analysis of the
 human resource output of higher education : a technical report sponsored by the Carnegie
 Commission on Higher Education. The Commission.

Aghion, P. & Howitt, P. (1992).      A Model of Growth Through Creative Destruction.
  Econometrica, 60(2), 323–51.

Aghion, P., Mayer-Foulkes, D., & Howitt, P. (2005). The Effect of Financial Development
  on Convergence: Theory and Evidence. Quarterly Journal of Economics, 120, 173–222.

      om, G. (1982). Engineers and Industrial Growth: Higher Technical Education and
Ahlstr¨
 the Engineering Profession during the Nineteenth and Early Twentieth Centuries: France,
 Germany, Sweden and England. Croom Helm.

      om, G. (1993). Technical Education, Engineering and Industrial Growth: Sweden in
Ahlstr¨
 the Nineteenth and Early Twentieth Centuries. In R. Fox & A. Guagnini (Eds.), Technology
 and Human Capital in Historical Perspective (pp. 115–142). Cambridge University Press.

Akcigit, U., Kerr, W. R., & Nicholas, T. (2013). The Mechanics of Endogenous Innovation
  and Growth: Evidence from Historical US Patents. Technical report, Working Paper.

Almada, M. & Zalduendo, E. (1962). El Estudio de los recursos humanos de nivel
                   ecnico en la Rep´
  universitario y t´               ublica Argentina.

Baer, W. (1969). The Development of the Brazilian Steel Industry. Vanderbilt University
  Press, [Nashville, Tenn.].

Barro, R. J. (1991). Economic growth in a cross section of countries. The Quarterly Journal
  of Economics, 106(2), 407–443.

Basu, S. & Weil, D. (1998). Appropriate Technology and Growth. The Quarterly Journal of
  Economics, 113(4), 1025–1054.

Baumol, W. J. (1990). Entrepreneurship: Productive, Unproductive, and Destructive.
  Journal of Political Economy, 98(5), 893–921.

Baumol, W. J. (2010). The Microtheory of Innovative Entrepreneurship. Princeton: Princeton
  University Press.

Bazant de Salda˜                                     on Durante el Porfiriato. Serie Historia
                na, M. (1993). Historia de la Educaci´
               on. Colegio de M´
  de la educaci´                                              oricos.
                                exico, Centro de Estudios Hist´

Behrman, J. & Birdsall, N. (1983). The Quality of Schooling: Quantity Alone is Misleading.
  The American Economic Review, (pp. 928–946).
                                           36
Benhabib, J. & Spiegel, M. (1994). The Role of Human Capital in Economic Development
  Evidence from Aggregate Cross-country Data. Journal of Monetary economics, 34(2),
  143–173.

Benhabib, J. & Spiegel, M. (2005). Human Capital and Technology Diffusion. Handbook of
  economic growth, 1, 935–966.

Benjamin, H. R. (1965). Higher Education in the American Republics. McGraw-Hill, New
  York ; Sydney.

Bernstein, M. D. (1964). The Mexican Mining Industry, 1890-1950 : A Study of the
  Interaction of Politics, Economics, and Technology. State University of New York.

Birchal, S. d. O. (1999). Entrepreneurship in Nineteenth-century Brazil: the formation of a
  business environment. London.

Bloom, N. & Van Reenen, J. (2010). Why do Management Practices Differ across Firms and
  Countries? The Journal of Economic Perspectives, 24(1), 203–224.

                                                        exico, 1763-1810. University
Brading, D. A. (1971). Miners and Merchants in Bourbon M´
  Press, Cambridge.

Braunerhjelm, P., Acs, Z., Audretsch, D., & Carlsson, B. (2010). The Missing Link:
  Knowledge Diffusion and Entrepreneurship in Endogenous Growth. Small Business
  Economics, 34(2), 105–125.

Bruhn, M. & Gallego, F. A. (2011). Good, Bad, and Ugly Colonial Activities : Studying
  Development Across the Americas. Review of Economics and Statistics.

Campante, F. & Glaeser, E. L. (2009). Yet Another Tale of Two Cities: Buenos Aires and
  Chicago. NBER Working Papers 15104, National Bureau of Economic Research, Inc.

Carvalho, J. M. d. (1982). Political Elites and State Building: The Case of Nineteenth-
  Century Brazil. Comparative Studies in Society and History, 24(3), pp. 378–399.

Caselli, F. (2001). Cross-Country Technology Diffusion: The Case of Computers. The
  American Economic Review.

       o-Climent, A. & Mukhopadhyay, A. (2013). Mass Education or a Minority Well
Castell´
  Educated Elite in the Process of Growth: The Case of India. Journal of Development
  Economics.

                                                             ıa Argentina: Edici´
Centro Argentino de Ingenieros (1981). Historia de laIngenier´                  on Especial
  Publicada. Centro Argentino de Ingenieros.

Ciccone, A. & Papaioannou, E. (2009). Human capital, the structure of production, and
  growth. The Review of Economics and Statistics, 91(1), 66–82.

Cohen, W. & Levinthal, D. (1989). Innovation and Learning: the Two Faces of R & D. The
  economic journal, 99(397), 569–596.


                                            37
Collier, S. & Sater, W. F. (1996). A History of Chile, 1808-1994. Cambridge Universtiy
  Press. 1996.

Collins, W. J. & Margo, R. A. (2006). Historical perspectives on racial differences in schooling
  in the united states. Handbook of the Economics of Education, 1, 107–154.

Comin, D., Easterly, W., & Gong, E. (2010a). Was the Wealth of Nations Determined in
  1000 BC? American Economic Journal. Macroeconomics, 2(3), 65–97.

Comin, D., Ferrer, M., & Mestieri, M. (2010b). TheIntensiveMargin of Technology Adoption.
  Technical report, Harvard Business School.

Comin, D. & Hobijn, B. (2004). Cross-Country Technology Adoption: Making the Theories
  Face the Facts. Journal of Monetary Economics, 51(1), 39–83.

Comin, D. & Hobijn, B. (2010). An Exploration of Technology Diffusion. The American
  Economic Review, 100(5), 2031–59.

Comin, D., Hobijn, B., & Rovito, E. (2008). Technology Uusage Lags. Journal of Economic
  Growth, 13(4), 237–256.

Comin, D. A., Dmitriev, M., & Rossi-Hansberg, E. (2012). The Spatial Diffusion of
  Technology. Technical report, National Bureau of Economic Research.

Comin, D. A. & Ferrer, M. M. (2013). If Technology Has Arrived Everywhere, Why Has
  Income Diverged? Technical report, National Bureau of Economic Research.

Contreras, M. E. (1993). The bolivian tin mining industry in the first half of the twentieth
  century. ISA Research Papers, (32).

Di Tella, G. (1985). Rents, Quasi Rents, Normal Profits and Growth: Argentina and the
  Areas of Recent Settlement. In Argentina, Australia, and Canada : studies in comparative
  development, 1870-1965. D.C.M. Platt and Guido di Tella.

Diaz Alejandro, C. F. (1985). Argentina, Australia and Brazil Before 1929. In Argentina,
  Australia and Canada: Studies in Comparative Develpment, 1870-1965. D.C.M Platt and
  Guido di Tella.

Diogo, M. (2007). Aprender a ser ingeniero: La ense˜                       ıa en el portugal de
                                                       nanza de la ingenier´
  los siglos xviii y xix. In A. Lafuente, A. Cardoso, & T. Saraiva (Eds.), Maquinismo ib´erico,
  Visiones hispanas. Doce Calles.

Duncan, T. & Fogarty, J. (1986). Australia and Argentina: On Parallel Paths. The Hispanic
 American Historical Review, 66(2), pp. 353–355.

Easterly, W. (2002).    The Elusive Quest for Growth:           Economists’ Adventures and
  Misadventures in the Tropics. MIT Press.

Eaton, J. & Kortum, S. (1999).       InternationalTechnology Diffusion:           Theory and
  Measurement. International Economic Review, 40(3), 537–570.


                                              38
Eaton, J. & Kortum, S. (2001). Technology, Trade, and Growth: A Unified Framework.
  European Economic Review, 45(4), 742–755.

Edelstein, M. (2009). The production of engineers in new york colleges and universities,
  1800-1950. In D. Eltis, F. Lewis, & K. Sokoloff (Eds.), Human Capital and Institutions:
  A Long-Run View. Cambridge University Press.

Encina, F. A. (1972). Nuestra Inferioridad Econ´omica : Sus Causas, Sus Consecuencias.
  Universitaria , Santiago de Chile, 3 edition.

                    ıa (2011). Innovaci´
Facultad de Ingenier´                                        on: Facultad de Ingenier´
                                       on, Excelencia Tradici´                       ıa
  1861-2011. Universidad Nacional, Bogot´a.

Fogarty, John Di Tella, G., Platt, D. C. M., & of Americanists, I. C. (1985). Staples,
  Super-Staples and the Limits of Staple Theory: The Experiences of Argentina, Australia
  and Canada Compared. In Argentina, Australia, and Canada : studies in comparative
  development, 1870-1965. D.C.M. Platt and Guido di Tella.

Foster, A. D. & Rosenzweig, M. R. (1995). Learning by Doing and Learning from Others:
  Human Capital and Technical Change in Agriculture. Journal of Political Economy,
  103(6), 1176–1209.

Foster, A. D. & Rosenzweig, M. R. (1996). Technical Change and Human-Capital Returns
  and Investments: Evidence from the Green Revolution. The American Economic Review,
  86(4), pp. 931–953.

Fox, R. & Guagnini, A. (1993). Education, Technology and Industrial Performance in Europe,
  1850-1939. Cambridge University Press.

Galor, O. (2011). Unified Growth Theory. Princeton Univ Press.

Galor, O. & Weil, D. N. (2000). Population, technology, and growth: From malthusian
 stagnation to the demographic transition and beyond. American economic review, (pp.
 806–828).

Gancia, G., Zilibotti, F., & Acemoglu, D. (2008). Technology-Skill Complementarity and
 Competition Policy. Technical report.

Gennaioli, N., La Porta, R., Lopez-de Silanes, F., & Shleifer, A. (2013). Human Capital and
 Regional Development. The Quarterly Journal of Economics, 105, 164.

Glaeser, E. L. (2007). Entrepreneurship and the City. NBER Working Papers 13551, National
  Bureau of Economic Research, Inc.

Glaeser, E. L., Kerr, W., & Ponzetto, G. A. (2010). Clusters of Entrepreneurship. Journal
  of Urban Economics, 67(1), 150–168.

Glaeser, E. L., Rosenthal, S. S., & Strange, W. C. (2009). Urban Economics and
  Entrepreneurship. NBER Working Papers 15536, National Bureau of Economic Research,
  Inc.

Goldin, C. (1999). A brief history of education in the united states.
                                             39
Goldin, C. & Katz, L. (1999). The shaping of higher education: The formative years in the
 united states, 1890 to 1940. Journal of Economic Perspectives, 13(1), 36–61.

Goldin, C. & Katz, L. F. (2009). The Race between Education and Technology. Cambridge,
 MA: Belknap of the Harvard U.

Goldin, C. & Katz, L. F. (2011). Mass secondary schooling and the state: the role of state
 compulsion in the high school movement. In Understanding Long-Run Economic Growth:
 Geography, Institutions, and the Knowledge Economy (pp. 275–310). University of Chicago
 Press.

Graham, R. (1981). Slavery and Economic Development: Brazil and the United States
  South in the Nineteenth Century. Comparative Studies in Society and History. Cambridge
  University Press.

Griffith, R., Redding, S., & Reenen, J. (2004). Mapping the Two Faces of R&D: Productivity
  Growth in a Panel of OECD Industries. Review of Economics and Statistics, 86(4), 883–
  895.

Griliches, Z. (1990). Patent statistics as economic indicators: A survey. Journal of Economic
  Literature, (pp. 1661–1707).

Haber, S. (1997). Financial Markets and Industrial Developoment: A Comparative Study
  of Government Regulation, Financial Innovation and Industrial Structure in Brazil and
  Mexico, 1840-1930. Stanford, CA:Stanford University Press.

Haber, S. (2005). Development Strategy or Endogenous Process? The Industrialization of
  Latin Americ. Technical report.

Haber, S. (2006). The Political Economy of Industrialization. The Cambridge Economic
  History of Latin America, 2, 537–84.

Hanushek, E. & Woessmann, L. (2007). The Role of Education Quality for Economic Growth.
  World Bank Policy Research Working Paper, (4122).

Hanushek, E. & Woessmann, L. (2012). Schooling, Educational Achievement, and the Latin
  American Growth Puzzle. Journal of Development Economics, 99(2), 497–512.

Hanushek, E. A., Ruhose, J., & Woessmann, L. (2015). Human Capital Quality and Aggregate
  Income Differences: Development Accounting for U.S. States. NBER Working Papers
  21295, National Bureau of Economic Research, Inc.

Harnow, H. (1997). The Role of the Engineer in Danish Modernisation, 1850–1920.
  Scandinavian Economic History Review, 45(3), 224–243.

Howitt, P. (2000). Endogenous Growth and Cross-country Income Differences. American
  Economic Review, (pp. 829–846).

Howitt, P. & Mayer-Foulkes, D. (2005). R&D, Implementation, and Stagnation: A
  Schumpeterian Theory of Convergence Clubs. Journal of Money, Credit and Banking,
  37(1), 147–77.
                                             40
Hurtado, O. (2007). Las Costumbres de los Ecuatorianos. Quito: Editorial Ecuador FTB.

ICC (1899). Annual Report of the Interstate Commerce Commission. United States Interstate
  Commerce Commission.

Iyigun, M. F. & Owen, A. L. (1998). Risk, entrepreneurship, and human-capital
  accumulation. American Economic Review, (pp. 454–457).

Iyigun, M. F. & Owen, A. L. (1999). Entrepreneurs, professionals, and growth. Journal of
  Economic Growth, 4(2), 213–232.

Judson, R. (1998). Economic Growth and Investment in Education: How Allocation Matters.
  Journal of Economic Growth, 3(4), 337–359.

Keller, W. (2004). International Technology Diffusion. Journal of Economic Literature,
  42(3), 752–782.

Klenow, P. J. & Rodriguez-Clare, A. (2005). Externalities and growth. Handbook of economic
  growth, 1, 817–861.

Krueger, A. B. & Lindahl, M. (2001). Education for Growth: Why and for Whom? Journal
  of Economic Literature, 39(4), 1101–1136.

Landes, D. S. (1998). The Wealth and Poverty of Nations : Why Some are so Rich and Some
  so Poor. W.W. Norton, New York, 1 edition.

 opez, C., Carreras, A., Tafunell, X., & Fundaci´
L´                                                                     ısticas hist´
                                                on, B. F. (2005). Estad´           oricas de
  Espa˜na: Siglos XIX-XX. Fundaci´  on BBVA.

Lopez Soria, I. (2012). Historia de la Universidad Nacional de Ingenier´
                                                                       ıa, los a¨ı¿ 1
                                                                                    2
                                                                                      os
                                                                          ıa, Proyecto
  Fundamentales, 1876-1909. Vol. 1. Lima: Universidad Nacional de Ingenier´
  Historia UNI.

Lucas, R. E. (1993). Making a Miracle. Econometrica, 61(2), 251–72.

Maloney, W. F. (2002). Missed Opportunities - Innovation and Resource-Based Growth in
 Latin America. Policy Research Working Paper Series 2935, The World Bank.

Maloney, W. F. (2015). Techno-literate Entrepreneurship and Development in Latin America.
 World Bank Mimeo.

Maloney, W. F. & Valencia, F. (2015). The Persistence of (Subnational) Fortune: Geography,
 Agglomeration, and Institutions in the New World. Economic Journal.

Mankiw, N. G., Romer, D., & Weil, D. N. (1992). A Contribution to theEempirics of
 Economic Growth. The Quarterly Journal of Economics, 107(2), 407–437.

Mann, C. (1918). A study of engineering education: prepared for the Joint committee on
 engineering education of the national engineering societies. Bulletin (Carnegie Foundation
 for the Advancement of Teaching). Merrymount Press.


                                            41
Marichal, C. (1997). Avances Recientes en la Historia de las Grandes Empresas y su
 Importancia para la Historia Econ´omica de M´exico. In C. Marichal & M. Cerutti (Eds.),
                                      exico, 1850-1930. Universidad Aut´
 Historia de las grandes empresas en M´                                onoma de Nuevo
 Le´on - Fondo de Cultura Econ´omica.

Mariscal, E. & Sokoloff, K. (2000). Schooling, Suffrage and the Persistence of Inequality in
 the Americas, 1800-1945. In Political Institutions and Economic Growth in Latin America:
 Essays in Policy, History, and Political Economy, ed. Stephen Haber. Stanford, CA: Hoover
 Institution Press.

McInnis, M. (2004). Engineering Expertise and the Canadian Exploitation of the Technology
 of the Second Industrial Revolution. In J. Ljungberg & J.-P. Smits (Eds.), Technology
 and Human Capital in Historical Perspective (pp. 49–78). Palgrave Macmillan.

Mendez, N. (2013).                            ıa en Venezuela.
                       Historia de la Tecnolog´                    Universidad Central de
 Venezuela.

Mitchener, K. J. & McLean, I. W. (2003). The Productivity of US States since 1880. Journal
 of Economic Growth, 8(1), 73–114.

Mokyr, J. (1998). The Second Industrial Revolution, 1870-1914. Northwestern University
 Manuscript. Evason, 1st edition.

Mokyr, J. (2005). The intellectual origins of modern economic growth. In A. Quadrio Curzio
 & M. Fortis (Eds.), Research and Technological Innovation (pp. 17–80). Physica-Verlag
 HD.

     on, M. (1982). Chile in the Nitrate Era: The Evolution of Economic Dependence,
Monte´
 1880-1930. University of Wisconsin Press Madison, Wisc.

Moretti, E. (2004). Estimating the Social Return to Higher Education: Evidence from
 Longitudinal and Repeated Cross-Sectional Data.

Moser, P. (2013). Patents and innovation: evidence from economic history. The Journal of
 Economic Perspectives, (pp. 23–44).

Murphy, K. M., Shleifer, A., & Vishny, R. W. (1991). The allocation of talent: Implications
 for growth. The Quarterly Journal of Economics, 106(2).

Nelson, R. R. (2005). Technology, Institutions, and Economic Growth. Harvard University
  Press, Cambridge, Mass.

Nelson, R. R. & Phelps, E. S. (1966). Investment in Humans, Technological Diffusion, and
  Economic Growth. The American Economic Review, 56(1/2), pp. 69–75.

Nevins, A. et al. (1962). The state universities and democracy. University of Illinois Press
  Urbana.

Nienkamp, P. (2010). Land-grant colleges and american engineers: Redefining professional
  and vocational engineering education in the american midwest, 1862-1917. American
  Educational History Journal, 37(1/2), 313.
                                            42
 un
N´ ˜ez, J. (2005). Signed with an x: Methodology and data sources for analyzing the evolution
 of literacy in latin america and the caribbean, 1900-1950. Latin American Research Review,
 40(2), pp. 117–135.

Nunn, N. (2008). Slavery, Inequality, and Economic Development in the Americas: An
 Examination of the Engerman-Sokoloff Hypothesis. In E. Helpman (Ed.), Institutions and
 Economic Performance (pp. 148–180). Harvard University Press.

Nunn, N. & Puga, D. (2012). Ruggedness: The Blessing of Bad Geography in Africa. Review
 of Economics and Statistics, 94(1), 20–36.

O’Brien, T. (1989). Rich beyond the dreams of avarice: the guggenheims in chile. Business
  History Review, 63(1), 122–159.

Pach´            ırez, M. T. (2006). La Infraestructura de Transporte en Colombia Durante
     on, A. & Ram´
                    a : Fondo de Cultura Econ´
  el Siglo XX. Bogot´                           omica.

Parente, S. L. & Prescott, E. C. (1994). Barriers to Technology Adoption and Development.
  Journal of political Economy, (pp. 298–321).

Phelps, E. S. (2013). Mass flourishing: How grassroots innovation created jobs, challenge,
  and change. Princeton University Press.

Pinto Santa Cruz, A. (1959).       Chile, un Caso de Desarrollo Frustrado.        Santiago :
  Universitaria, 1959.

                                                                              ıa e historia
Poveda Ramos, G. (1993). Historia Social de las Ciencias en Colombia: Ingenier´
          ecnicas, Vol. IV y V. Colciencias.
  de las t´

Prados de la Escosura, L. (2000). International Comparisons of Real Product, 1820-1990:
  An Alternative Data Set. Explorations in Economic History, 37(1), 1–41.

Pritchett, L. (1997). Divergence, Big Time. Journal of Economic Perspectives, 11(3), 3–17.

Pritchett, L. (2001). Where Has All the Education Gone?         The World Bank Economic
  Review, 15(3), 367–391.

          ebols, S. (1993). Industrialization and Technical Education in Spain, 1850-1914.
Riera i Tu´
  Cambridge University Press.

Rogers, E. (1962). The Iron and Steel Industry in Colonial and Imperial Brazil. The Americas,
  1(2), 172–184.

Romer, P. M. (1990). Endogenous technological change. Journal of Political Economy, 98(5
  pt 2).

Rosenberg, N. (2000). Schumpeter and the Endogeneity of Technology : Some American
  Perspectives. Routledge, London.

                                                                                  on,
Ruiz Larraguivel, E. (2004). Los Ingenieros en la Industria Manufacturera. Formaci´
        on y Actividad Laboral. CESU-Plaza y Valdes.
 Profesi´

                                             43
Safford, F. (1976). The Ideal of the Practical : Colombia’s Struggle to Form a Technical
  Elite. University of Texas Press, Austin.

Sandberg, L. G. (1979). The Case of theIimpoverished Sophisticate: Human Capital and
  Swedish Economic Growth before World War i. Journal of Economic History, 39(1),
  225–241.

Santa-Maria Alvarez, P. (1994). Origen, Desarrollo y Realizacion de la Escuela de Minas de
        ın. Ediciones Dike LTDE.
  Medell´

Schumpeter (1934). The Theory of Economic Development. (trans R. Opie). Boston: Harvard
  University Press.

Scobie, J. R. (1964). Argentina : A City and a Nation. Oxford University Press, New York.

Self, S. & Grabowski, R. (2004). Does education at all levels cause growth? india, a case
  study. Economics of Education Review, 23(1), 47 – 55.

Serrano, S. (1993).    Universidad Y Nacion:.              on Imagen de Chile. Editorial
                                                    Colecci´
  Universitaria.

Sianesi, B. & VanReenen, J. (2003). The Returns to Education: Macroeconomics. Journal
  of economic surveys, 17(2), 157–200.

Sokoloff, K. L. (1988). Inventive activity in early industrial america: evidence from patent
  records, 1790–1846. The Journal of Economic History, 48(04), 813–850.

Stevens, P. & Weale, M. (2004). 4 Education and Economic Growth. International handbook
  on the economics of education, 1830(1850), 164.

Swank, J. M. (1965). History of the Manufacture of Iron in all Ages, and Particularly in the
  United States from Colonial Times to 1891. B. Franklin New York,, [2d ed.] edition.

                                oria da Engenharia no Brasil, S´
Telles, P. C. d. S. (1994). Hist´                              eculos XVI a XIX. Clavero,
  Rio de Janeiro.

Temple, J. & Voth, H.-J. (1998). Human Capital, Equipment Investment, and Industrializa-
  tion. European Economic Review, 42(7), 1343–1362.

                                             on. Una Historia Econ´
Thorp, R. (1998). Progreso, Pobreza y Exclusi´                               erica Latina
                                                                  omica de Am´
 en el siglo XX. Washington D.C. BID- Uni´   an Europea.

Tortella, G. (1994).    Patterns of Economic Retardation and Recovery in South-
  WesternEurope in the Nineteenth and Twentieth Centuries. The Economic History Review,
  47(1), pp. 1–21.

Tortella Casares, G. (2000). The Development of Modern Spain : An Economic History of
  the Nineteenth and Twentieth Centuries. Harvard University Press, Cambridge, Mass. :.

                            an, L. (1990). Historia de la Ingenier´
Villalobos, S. R., , & Beltr´                                     ıa en Chile. Hachette.
  Santiago.

                                            44
      ander, N. & Squicciarini, M. (2014).
Voigtl¨                                             Knowledge elites, enlightenment, and
  industrialisation. NBER.

Waldinger, F. (2012). Bombs, brains, and science.

Will, R. M. (1957). Some Aspects of the Development of Economic Thought in Chile (ca.
 1775-1878). Technical report.

Wolff, E. & Gittleman, M. (1993). The Role of Education in Productivity Convergence: Does
 Higher Education Matter? Contributions to Economic Analysis, 214, 147–147.

World Economic Forum, Schwab, K. et al. (2012). TheGlobal Competitiveness Report 2010–
 2011. Geneva, Switzerland.

Wright, G. (1986). Old South, New South : Revolutions in the Southern Economy since the
 Civil War. Basic Books, New York.

Wright, G. (1987). The Economic Revolution in the American South. Journal of Economic
 Perspectives, 1(1), 161–78.

Wright, G. (1999). Can a nation learn? american technology as a network phenomenon. In
 Learning by doing in markets, firms, and countries (pp. 295–332). University Of Chicago
 Press.

Young, A. (1993). Invention and Bounded Learning by Doing. Journal of political economy,
  (pp. 443–472).

Yuchtman, N. (2014). Teaching to the tests: An economic analysis of traditional and modern
  education in late imperial and republican china.




                                           45
9        Annex I: Construction of Engineering Data
9.1      Argentina
                                               ıa Argentina (Centro Argentino de Ingenieros,
The principal source is Historia de la Ingenier´
1981). At the end of the 19th century, there were three universities that granted the title of
civil engineer which was their omnibus term for engineers- Buenos Aires, Cordoba and La
Plata as well as a school of mining engineers in San Juan. The CAI documents that from
1870, the year when the first engineers graduated in the country, until 1900, 250 engineers
received their diplomas. We do not know the distribution of these degrees across years so we
impute uniform graduation rates after which the attrition adjustment leaves 196 or a density
of 12.34
9.2      Bolivia
In Bolivia, the first engineering school, the Escuela Nacional de Ingenieria in Oruro, began
in 1917 and hence Bolivia has 0 locally trained engineers by 1900. As late as 1937,
84% of engineers in the Pati˜  o Tin mines (accounting for 10 of world production) were
foreign(Contreras, 1993).
9.3      Brazil
Telles (1994) Historia da Engenharia no Brasil, Seculos XVI a XIX is the principal source.
In 1858 the Royal Academy of Artillery, Fortification and Drawing, established in Rio
de Janeiro in 1792, dedicated itself to civil engineering for the first time, studying steam
engines and railroads, and in 1874 it became independent of the Military and became the
Polytechnical School of Rio (today the School of Engineering of the Federal University of
Rio de Janeiro). This was the dominant institution for training engineers. Brazil’s second
engineering school was founded around Mining in Ouro Preto. However, the low motivation
for technical teaching of the time the school’s isolation, among other social factors, made it
difficult to recruit students and it graduated few. From 1894 to 1896 four new schools were
started in Sa˜o Paolo, Pernambuco, Porto Alegre and Salvador. Telles (1994) suggestion that
these schools would eventually end the Polytechnical School of Rio’s monopoly confirms the
dominance of the latter in the production of engineers up to that point. We do not, however,
have a long time series on graduates from any program. Telles reports the average annual
number of domestic engineering graduates in Brazil as a whole for the period after 1890 at
45 per year, half of them produced in Rio by 1900. To estimate graduation rates for the
1860-1890 period, we rely on evidence from reported stocks. Telles tabulates the number of
engineers in Rio as reported in Almanaque Laemmert, a periodical dealing with governance,
commerce and industry in Rio, for 1854, 1870, 1883 and from the official Almanak dos
Engenheiros an official publication of the government for the whole country in 1906. In Rio
in 1854, there were 6 engineers and in Sa˜  o Paolo, the other principle locus of engineering
talent, in 1857, there were 5. We therefore set 11 as our initial stock for the country in 1860.
In 1870, the Almanaque Laemmert notes that Rio had grown to 28 engineers and by 1883,
126 (page 593). Given the rough earlier parity of Rio and Sa˜    o Paolo in 1854-57, we double
the Rio numbers figures to get national figures for these periods. While the Almanak may

    34
     A later data point is offered by Almada & Zalduendo (1962), which, when adjusted to be compatible
with our data, yields a density of 41.25 in 1925. Given the rapid increase in output of engineers in the
beginning decades of the 20th century in most countries, this supports our 1900 estimate.
                                                  46
be overstating the stock by including non degreed engineers, the implicit graduation rate
leading up to 1883 is roughly 15 per year which is substantially below Telles’ documented
graduation rate of 45 beginning in 1890. On the other hand, the consolidation of the Ouro
Preto School of Mines and the new schools established after 1890 doubled whatever Rio’s
capacity was and that was likely substantially more in 1890 than prior. Hence, a three-fold
increase over the last two decades seems plausible. We interpolate an average value between
the known values of 1883-1890. Together, these lead to a total stock in 1900 of 786. If
we extrapolate at the same graduation rate, the terminal stock in 1906 is 968 or slightly
above the value reported by the official Almanak (941) suggesting that we may, again, be
overstating the stock somewhat. Density 12.


9.4       Canada
McInnis (2004) is perhaps the most complete of a thin literature. Substantial engineering
curricula had been introduced at King’s College (UNB), at McGill College, and the Uni-
versity of Toronto in the 1850s although demand for engineering education gained traction
only in the 1870s. McGill offered a full diploma course by 1863 although the first five
students graduated only in 1874. Four year courses were implemented in Civil Engineering,
Mechanical Engineering, Practical Chemistry and Mining by 1878, Electrical engineering
in 1891, Chemical Engineering and Metallurgical Engineering in 1908. The University of
Toronto School of Practical Science opened in 1878 and offered the degree of civil engineer
in 1885. In 1874, Laval University established an Ecole Polytechnique which emitted its
first graduates in 1877. Other smaller programs also emerged at the same time. Also of
importance, The Royal Military College in Kingston Ontario, established in 1876, with
West Point as a model, explicitly had the dual object of providing scientific training to
military officers as well as producing civilian engineers.35 If we take the discounted sum of
the licensed graduates plus half the military graduates 36 by 1900 we reach a total density of
about 41. This is relatively low by US standards especially given the number of institutions
offering engineering courses, as well as the articulation of the different fields of engineering
at a relatively early phase. The development of, for instance, mechanical engineering as
a separate course about 20 years after in the US, Electrical Engineering 10-15 years after,
but still far ahead of any of the Latin Universities in the sample. Electrical engineering
appears more or less at the same time as in the US. It is worth noting, however, that the
four principle Canadian Universities emitting graduates lay within a circle of 350 mile radius
with Cornell at its center and including many of the principle US universities of the time.
Density: 41


9.5       Chile
As Serrano (1993) notes in Universidad y Naci´    on: Chile en el Siglo IX, training at the
Universidad de Chile (University of Chile), the principal source of engineers in the 19th
century, began in the mid-1850s. Prior to this, there were effectively no schools in Chile and
those engineers trained abroad were very few. From 1846-50 there had been 2 fellowships to

  35
     http://www.warmuseum.ca/education/online-educational-resources/dispatches/the-royal-military-
college-of-canada-1876-to-the-present
  36
       our thanks to Marvin McInnis for discussions on this
                                                     47
study abroad with uneven results. Serrano notes (p. 216) that between 1856 and 1879, 100
geographical engineers (surveyors/geographers), 61 mining engineers and 4 civil engineers
plus 11 general assayers graduated. This gives us a graduation rate for the first 20 years of
our exercise. To anchor the subsequent 20 years, Villalobos et al. (1990) collaborating with
                                  ıa en Chile, offers that “in the 19th century there were 130
Serrano in Historia de la Ingenier´
Chilean engineers and toward 1938, the country had a list of 270 professionals graduating
from the University of Chile. They created, in 1930, the Institute of Mining Engineers of
Chile ” (p. 198). The context may be taken to suggest that we are talking exclusively about
mining engineers, although in personal communication, Serrano confirms that it is total
graduated engineers. This is consistent with the fact that the implicit graduation rates from
the pre -1879 period, 2.3 mining engineers per year, respectively accumulates to only half
of the 130 number cited by Villalobos at end century. Clearly, this gap could be made up
by a rapid expansion in number of graduates from 1879 on, but as Serrano notes, in 1867
the government expressed concern that the numbers of graduates in physical studies and
mathematics was actually decreasing (page 212) so this would have represented a reversal
in trend. Geographical engineers translate broadly as surveyors/geographers which are not
generally treated as engineers per se and hence, to the degree that they are included in the
130 number, this overstates installed capacity. Density: 17.


9.6    Colombia
As Safford (1976) notes in his The Ideal of the Practical: Colombia’s Struggle to Form and
Technical Elite, the process of establishing a technical class was undermined by recurrent
civil wars which often whipsawed the ideological foundations of the schools when they were
not closing them, and perennial shortages of funding. The Universidad Nacional (National
University), founded in 1867, was the dominant source of degreed engineers. It was built on
the Colegio Militar (Military College) which operated over two brief periods, 1848-1854, and
then in 1861. In the early 1880s, the Congress also authorized the creation of mining schools
in Antioquia, Rionegro, Popayan and Ibague but most were aborted by the civil war of 1885.
Safford (1976). The exception was the Escuela Nacional de Mineria (National Mining School
in Antioquia set up in 1887 to 1895, that would eventually close due to a lack of financing,
among other factors, and become part of the Escuela de Ingenier´      ıa de la Universidad de
Antioquia (Engineering School of the University of Antioquia). Poveda Ramos (1993) in
Historia Social de la Ciencias en Colombia: Ingenier´                      ecnicas summarizes
                                                     ıa e Historia de las T´
“By the end of the century, there were only three schools of engineering in Colombia: the
Universidad Nacional, the Escuela Nacional de Mineria, and the Universidad Republicana
(Republican University) in Bogota. The number of students was small, so much so that
the National University, the largest of all, the number of students fluctuated from one year
to another between 25 and 50.”(p.55). The Universidad Republicana (now the Universidad
Libre Free University) was begun in 1890 but it nearly collapsed financially by 1910 and its
contribution to our accumulated stock is likely to be small. Facultad de Ingenier´   ıa (2011)
tabulates that from 1868-1870 enrollments in the National University averaged 35 per year,
yet graduates in the 1871-1875 period average about 4 per year. Though the authors note that
their tabulations may not be complete, the virtual absence of graduates from 1876 to 1888 is
plausible as from 1876 to 1884 the school was again taken over by the military and oriented
away from industry related training. In 1880, despite 56 enrolled, higher mathematics and
engineering classes contained only 4 students. (p. 195). In the National School of Mines,
                                               48
from 1887 to 1890 average enrollment was 25 students. Safford notes 63 alumnai of the 1888-
1894 period, which is confirmed by Santa-Maria Alvarez (1994) as the number of ”egresados”
(exiters) of the program. However the same text notes that only five of these had graduated
with thesis across the period (in 1893 and 1894) and none again until 1906 (Annex 5 page
103). Poveda Ramos (1993) confirms the lower numbers noting that the first 3 degrees of
Mining Engineer were granted in 1893. The two schools together yield an accumulated stock
of 75 Engineers by 1900. This is broadly consistent with Safford’s finding of “more than 200
Colombian engineers and surveyors ”in 1887 derived from the Anales de Ingeneir´  ıa (Annals
of Engineering), the organ of the Colombian Engineering Association Safford (1976) page
219) which, again does not discriminate by whether or not the inscribed had completed a
degree, nor separate out surveyors. However, we also know that both the University of Cauca
as well as the Republican University in Bogota were generating some unknown quantity of
graduates. We round to 100 as number that would incorporate these and missing graduates
from Antioquia and the National University. Density: 5.
9.7    Denmark
The Polyteknisk Laereanstalt was founded in 1829 as the first university level technical school
in Copenhagen and was one of the first of its kind in Europe and was heavily influenced by
the French Ecole Polytecnique. Harnow (1997) in his study of the impact of engineers in
Denmark only focuses on this school, arguing that from 1850 to 1920 it was by far the most
important Danish technical institution. He tabulates the number of graduates across the
period 1832-69 and then for roughly 10 year periods after. Taking the yearly graduation rate
as the average of each period and then applying the usual discounting yields a density of 92.
9.8    Ecuador
The Escuela Polit´ecnica Nacional (National Polytechnical School) was founded in 1869 by
the President Gabriel Garc´ıa Moreno with the aim of establishing a center for research and
training of engineers and scientists at a high level. German Jesuits were brought for the
purpose but the school was closed in 1876 for political reasons and was only reopened in
1935. This trajectory is not so different from that of Colombia’s School of Mines, although
that country had two other universities for more or less three times the population. We
can’t know the number of graduates of the program, over these seven years, but the density
would have to be less that than of Colombia, including that until 1935 there was effectively
no local training capacity which is part of what we’re trying to capture here. http :
//www.epn.edu.ec/index.php?option = comc ontentview = articleid = 1129Itemid = 378.
We assign a value of 2.
9.9    Mexico
The earliest technical training in Mexico was the Colegio de Mineria formerly the Real
Seminario de Mineria (College of Mining, Royal Seminary of Mining) in Mexico city which
opened in 1792 and was perhaps the most secular and highest quality technical institution
in the hemisphere at the time. Bazant de Salda˜    na (1993) in her Historia de la Educaci´ on
Durante el Porfiriato has best documented the subsequent evolution. Wars of independence,
foreign invasion, and perilous fiscal situations led to a steady decline and by the time it was
transformed into the into the Escuela Nacional de Mineria (National School of Mines) in
1867 under Benito Juarez, the number of students was so low that the government considered
closing it and sending the 8-10 students abroad. Porfirio Diaz would subsequently put great
                                             49
emphasis on engineering as part of his modernization compaign. Despite this, by 1902, still
only 18 engineers were graduating per year. Flows from the National School of Mines from
1876-1901 total 327. From 1876 to 1880 (41); 1881-1890 (106); 1891-1901 (180). Most other
universities in other areas contributed very few. Allowing for another 16 years prior at the
1877 rate, which likely overstates the case, gives a total stock in 1900 of 336 or a surprisingly
low density of 5. Other figures broadly corroborate. The census reports 884 engineers for
Mexico city or roughly half the total that it reports for the entire country. Applying that
ratio to the stock above gives 159. By comparison, Bazant cites the Massey Blue Book,
an English language director of Mexico (City) as giving a total of 91 engineers and the
Directorio de Vecinos de la Ciudad de Mexico as 183, both including some unspecified number
of foreigners. The Association de Ingenieros (Engineering Association) in 1910 counted 255
members which, again, is not clear on the level of education of its members and may also
include both the acceleration in graduation at the turn of the century in many countries. In
all, the magnitudes do not suggest that our stocks are importantly underestimated. Density
of 5.
9.10     Peru
Although there were institutions teaching technical skills in various parts of the country,
modern engineering began in Peru in 1852 two French and one Polish engineer to design
and undertake public works of engineering. The need to import talent for these tasks,
as was the case elsewhere in Latin America, testifies to the dearth of locally generated
qualified human capital. The first school of engineers was discussed in the early 1850s,
but only became reality when the Peruvian state in 1876 invited Polish engineer, Edward
John Habich, to advise on irrigation, railways and other projects as well as the founding of
a school of mines. Lopez Soria (2012) in Historia de la Universidad Nacional de Ingenieria,
los A˜nos Fundamentales, 1876-1909 notes that the resulting School of Civil Construction
and Mining Engineers (now the National Engineering University-Universidad National de
Ingenieria) opened in 1876 and graduated its first class of 4 in 1880. The school was heavily
damaged when used by the invading Chilean forces in 1880 and took several years to rebuild,
only graduating one more student by 1882. Lopez Soria (2012) tabulates annual list of
graduates going forward, disaggregated by specialty and allowing us to take out surveyors
and include only industrial, mining and civil engineers, giving a total net of attrition of
100. This broadly confirms the statement by the Sociedad de Ingeneiros Del Per`   u (Peruvian
Engineering Society) (established 1898) of ”more than 80” engineers in the country. This
gives us a density of 5.
9.11     Portugal
Formal training of non military engineering in Portugal did not begin until the turn of
the 20th century with the Instituto de Lisboa (Institute of Lisbon) which started training
industrial engineers in 1903 (Heitor et al), and the Instituto Superior T´  ecnico (Higher
Technical Institute) founded in 1917 Diogo (2007)). Hence, we are unable to generate a
stock of graduates as in many of the other cases. Diogo argues, however, that military
engineers were responsible for most civil engineering projects and hence military engineers
should be counted in this case. The Associa¸ ao dos Engenheiros Civis Portuguezes - AECP
                                            c˜
(Portuguese Association of Civil Engineers) also did register the majority of those who
considered themselves non-military engineers. Though registration in the AECP was not
mandatory to be a practicing engineer, it was mandatory in the organization that followed,
                                             50
the Ordem dos Engenheiros (OE) (Order of Engineers). In 1870 the AECP reports 150
inscribed; in 1926, 733. We take the average growth rate between the two points and impute
the value for 1900. After 1900, we are able to compare the rates between the AECP and
the mandatory OE: in 1930, there were 845 members (AECP) and in 1936, 1127 members
(OE). Imputing the same growth rate between 1930 and 1936 as previous suggests that the
AECP is understating the stock of practicing engineers by roughly 20%. We apply this to
the stock value generated using the AECP data for 1900 to yield 579 or a density of 22. This
is likely to be an overstatement since we do not know what fraction of these had any higher
educational training.
9.12     Spain
We offer two estimates of the stock of Spanish engineers derived from Riera i Tuebols (1993)
from 1867 Industrialization and Technical Education in Spain, 1850-1914 and L´     opez et al.
(2005) Estadisticas Hist´oricas de Espa˜na: Siglos XIX-XX from 1857. The estimates differ
in scope. Riera i Tuebols reports graduates of escuelas de ingenier´ ıa engineering schools as
such, starting with Spain’s first, founded in Barcelona in 1867, to train industrial engineers
(see also Riera, 2008). He also offers data from mining and civil engineer graduates primarily
from institutions in Madrid. Though Riera’s tabulations are the most accurate count of
certifiably degreed engineers from university programs available, the resulting stock, 892,
may be a lower bound. L´     opez et al. (2005) casts a broader net, including information
from all technical schools (including Escuelas Nacionales, Escuelas Superiores, Escuelas
Especiales, Escuelas Centrales, Escuelas Profesionales and Escuelas Elementales ). Although
this compendium is more comprehensive geographically, the estimates include graduates
from other technical disciplines potentially miscategorized as engineers as well as including
graduates of indeterminate level of training. We treat the resulting estimate of 3,089 as an
upper bound. The Riera number is roughly half of the number of engineers and architects
combined reported in the 1900 census. The Lopez is about 50% higher, which makes it the
only case among our countries where the accumulated estimate is above that reported in the
census. Since, as noted, self-reported census definitions are looser than documented degrees
conferred, we find this improbable. Density either 12 or 42 respectively and we plot the
average of the two. In our regressions, the Spanish influence is accounted for by a dummy so
our results are unaffected by these estimates.
9.13     Sweden
The reference here is Ahlstr¨   om (1993) who tabulates graduates of the two principal
engineering programs. The Kungl Tekniska H¨    ogskolan (KTH) or Royal Technical University
in Stockholm has roots in the Laboratorium Mechanicum founded in 1697, which later became
the Mechanical school (1798). The Chalmers Institution in Gothenburg, founded in 1829,
                                                             om argues that in the mid 19th
provided technical education equal to that of the KTH. Ahlstr¨
century, ”...anyone in Sweden who sought an internationally reputable technical education
could find it in these institutions.” Density is 99.
9.14     United States
9.14.1   National data
We draw on several sources for the US engineering numbers. First, Mann (1918), in his
Study of Engineering Education done for the Joint Committee on Engineering Education,

                                             51
tabulated graduates from US schools until 1915. As of 1900 this gives a total of 14,679, which
gives a density per 100,000 workers of 50. However, as Adkins (1975) in The Great American
Degree Machine: An Economic Analysis of the Human Resource Output of Higher Education
notes, before 1940, the Office of Education made no effort to maintain comparability across
years or completeness of coverage of educational institutions. It is not clear how they
identified the universe of relevant institutions and, if an institution did not respond to their
survey two years in a row, it was dropped from the interview rolls. Hence, Mann’s estimates
underestimate the true stock by a potentially significant amount. To bring to bear other
sources of information, we use more reliable graduation data in select states or periods to
calibrate the Census numbers, and then impute engineering stocks for the country in 1900.
First, Adkins’ tabulations for the US in 1930 yield a stock that is .53 of the census declaration
of occupations in engineering at that time. Second, Edelstein (2009) in The Production of
Engineers in New York Colleges and Universities, 1800-1950 offers a full and comprehensive
stock accounting for the state of New York for our time period 1900. His tabulations yield a
density of 179 which is .67 of the US census number corresponding to New York that year.
It is likely that New York’s number may have a higher density of fully degreed engineers self-
reporting in the census than the country as a whole so this number may be somewhat high.
Similarly, Adkins’ estimates for the whole country in 1930 may reflect that, in the ensuing
30 years, a higher share of self-declared engineers actually had degrees. Mann’s numbers
yield a ratio of .39 which we expect to be too low for the reasons outlined above. Hence, we
take an intermediate value of .5 as the national ratio of actual graduates to Census declared
engineers in 1900 and the data for the South and the North are projections based on this ra-
tio. This yields a density of 84 for the entire country, 160 for the North, and 60 for the South.

9.14.2   County level data
Innovative capacity: As elsewhere we calculate density by engineers per 100.000 male
workers. We use OCC1950 variable of IPUMS USA for 1880 census and we aggregate all
categories for set up engineers: Engineering, chemical engineers, civil engineers, electrical
engineers, industrial engineers, mechanical engineers, metallurgical and metallurgists
engineers, mining engineers, and other engineers. We use as proxy of male workers the
40% of population because there are inconsistencies in labor force variable. Patenting is
the collected number of patents between 1890 and 1910 like proportion to 1880 population.
Number of patents was collected from Akcigit et al. (2013) and include all patents granted
by the USPTO.

Sub-national Income in 2005 US Dollars is the household mean income drawn from 2000
USA Census.

Geographical Controls: we include temperature, altitude and rainfall taken from Worldcrim.
We use distance to river (distance between the county centroid to the near medium or big
size river) derived from HydroSHEDS (USGS 2011). We employ a measure of distance to
the coast calculated like distance between centroid and the near coast similar to Gennaioli
et al. (2013). We further include a measured of ruggedness of terrain from Nunn Puga (2012).

Growth variables: Population Density in 1880 as a measure of agglomeration is taken
from the 1880 census data. As a measure of lagged economic activity, we calculate two
                                               52
measures using manufacturing output, taken from the 1870 NHGIS. Per capita yields a
measure of structural transformation, per manufacturing output a measure of productivity.
The extreme values and high variance of manufacturing employment suggests some lack of
confidence in the 1880 labor allocation data so we report only the first measure. However,
the results do not change appreciably using the other. We use slavery as a measure of
institutions taken from the 1860 Census as well as Nunn (2008). For railroads, we employ
two variables from the Nebraska-Lincoln University railroads data. An identifier variable if
the county has railroad or a railroad density variable.

Human capital: Aggregate literacy rates we take from 1880 Census. As with engineers,
we measured lawyers and physicians density per 100.000 habitants. We use the OCC1950
variable of IPUMS USA for 1880 census. Lawyers and Physicians categories are used from
OCC1950.

Instrument: we compute the distance between the county centroid to the near the 57 1862
Land Grant Colleges.


9.15    Venezuela
Mendez (2013) in Historia de la Tecnolog´ıa en Venezuela notes that the Universidad Central
de Venezuela (UCV) (Central University of Venezuela), as it would eventually be known,
become the primary source of engineering graduates from 1867 on: 8 from 1867 to 1879 ; 80
from 1880 to 1889; 102 from 1890 to 1899. Other universities that graduated engineers were
la Universidad del Zulia (University of Zulia) (1 in 1892) and the Universidad de Valencia
(University of Valencia) (4 between 1892 y 1904); Colegio Federal de Maracaibo (Federal
College of Maracaibo) 1886 (5) who submitted their these to the UCV for approval. To fill
in the 1860-1866 period, we take the average of graduates from Academia de Matem´    aticas
de Caracas (Academy of Mathematics of Caracas)from 1831 to 1872 (97 graduates) perhaps
half of which were employed in civil or industrial work. Applying our usual discounting
gives about 185 engineers. The engineering association gives 196 although this may include
foreigners and members of undetermined educational attainment. Density: 11.




                                            53
10      Annex II: The US at the State Level
We now look more closely at the US sub-sample alone since the 1900 census offers several
other correlates that allow us to test the robustness of the engineering results (Table
4). The cost of these additional covariates is the reduced number of observations (51).
The geographic variables for this sample are not significant as a block and we drop them
to preserve degrees of freedom.37 The magnitudes of the remaining coefficients are not
sensitive to this omission. In columns 1-3, Engineering enters significantly freestanding,
with population density, and literacy included sequentially and entering significantly and
positively as before.

    We test the robustness of this result in two ways. First, the US sample allows progres-
sively adding a richer group of controls. Following Murphy et al. (1991) for the present
day, in column 4 we include the density of lawyers. This does not alter the magnitude of
the engineering variable suggesting that we are not picking up simply higher order human
capital. In column 5 we add state-level measures of railroad density and mining activity
which, again, capture infrastructure or industrial activity that engineering may be proxying
for. The former enters significantly and positively although mining does not and engineering
continues to remain significant and of similar magnitude.

    It is also possible that engineering is proxying for the institutional differences across
states, and in particular, the legacy of slavery. We employ slavery data from Nunn (2008)
which reduces our sample to 38 observations. Consistent with Nunn (2008), slavery alone
(not shown) enters negatively and significantly. However adding engineering eliminates
its importance, while engineering density itself remains strongly significant (Column 6).
This suggests that weak engineering capacity is one possible channel through which the
institution of slavery depressed Southern incomes.38

   Including all variables (column 7) renders many insignificant although engineering
prevails. Though we are pushing the data hard given the limited degrees of freedom, engi-
neering retains its significance after controlling for agglomeration effects, other higher order
human capital, sectors using engineers that may have an independent effect, and institutions.

    Second, though we have controlled for the two activities most related to engineering, in
the spirit of Moretti (2004), we also attempt to instrument engineering density using the
number of Morrill Land Grant colleges and universities found in each state. As discussed
earlier, the Morrill program was introduced in 1868 precisely to remedy the perceived
shortfalls in regional technical assistance in agricultural and mechanical innovation. In
practice, this program financed the first engineering departments in the emerging West and
Midwest and especially in the South. It was to an important degree supply driven. Prior
to the Civil War, the South had actively opposed the bill, fearing greater interference in
  37
    Given the small number of observations, we also bootstrap the standard errors. The results remain
unchanged.
  38
     Taking the parameter value on engineering from the most complete specification (column 7), the
difference in engineering density between the North and the South could account for a log difference of
.12 or roughly 13 percent which is, in fact, larger than the difference that currently exists between the two
regions.
                                                    54
matters such as universal primary education. The withdrawal of the Confederate States from
the US Congress allowed the bill to be passed. However, during Reconstruction, recognizing
its technological lag, the South started privately some universities such as Georgia Tech,
and actively embraced the Morrill Program.

    The first stage reveals a strong and negative correlation between engineering density
and the Morrill uptake, suggesting the role of the program as a remedial supply side
effort.39 As discussed earlier, Morrill financed programs in Texas, Virginia, Kentucky, and
North Carolina began awarding degrees in the 1880s and 1890s which means that their
accumulated engineering stock would still be low in 1900. Table 5 presents the second
stage results. Though we have few degrees of freedom, engineering enters significantly in all
specifications despite sequential addition of agglomeration, literacy and lawyers as controls.
The coefficient is significantly higher in all specifications suggesting that the instrument is,
in fact, helping to overcome an important downward bias.

   In sum, the various estimations offer support to the idea that higher order human capital
and institutions related to engineering and science and technology at the turn of the 20th
century plausibly are important to explaining present prosperity.




   39
     With 51 observations, this exercise is somewhat heroic and diagnostics should be taken as suggestive.
Both the first stage Cragg-Donald F test and Anderson-Rubin test of joint significance of regressors are
marginally acceptable suggesting that attempting to instrument is informative. Using the second wave 1890
Morrill grants yields stronger diagnostics although there is less clarity on the selection criteria and hence we
prefer using the 1868 wave. Both yield similar second stage results.
                                                      55
11     Figures and Tables


                     Figure 1: Income 1900 and Engineering Density 1900




     Note: Plot of GDP per capita in 1900 from Maddison. Engineering Density is accumulated graduates of
     engineering programs per 100,000 male workers around 1900 as described in the Annex.




                                                     56
                                       Figure 2: Sub-national Engineering Density, US and Mexico, in 1900
57




     Note: Engineering Density at the subnational level for North America. Derived from census reported engineers per 100,000 male workers around 1900.
             Figure 3: Copper Production in Chile and Foreign Engineers




Note: The figure shows the decline in Chilean production and share of the world market for Copper approaching
1900 and then the sharp recovery with the entrants of foreign mining companies and engineers in 1905.
Engineering density per 100,000 male workers calculated from National Censuses available until 1920 when
Chile ceased to divide by profession-origin.




                                                    58
                           Table 1: Summary Statistics (State Level, Americas)


             Variable                                   mean            p50           sd    min    max
             Ln Income                                    9.03          8.92        0.91    7.13   11.18
             Engineers                                   25.34         11.00       31.49    5.00   84.00
             Engineers (sub)                             88.30         44.81      108.69    0.00 472.59
             Population Density (1900)                   42.78          4.54      248.42    0.00 3319.27
             Population Density (1500)                    8.88          2.00       26.13    0.00 392.34
             Literacy                                    41.93         34.00       24.41   11.30   86.70
             Literacy (sub)                              53.86         46.57       30.97    7.64   98.31
             Railroads                                    3.15          1.80        2.71    0.30    9.30
             Railroads (sub)                             65.12         48.54       57.46    5.16 309.20
             South                                        0.15          0.00        0.36    0.00    1.00
             Slavery                                     20.67          3.28       25.16    0.00   72.66
             Lawyers                                    218.92        139.22      210.93    1.64 1156.44
             Mine Output ($)                              .466          .123        1.06 .000075    6.49
             Spain                                        0.81          1.00        0.39    0.00    1.00
             Land Suitability                             0.56          0.58        0.28    0.00    1.00
             River Density                                3.28          3.29        1.23    0.00    6.92
             Average Temperature                         19.97         20.40        5.83    2.38   29.00
             Rainfall                                     1.28          1.10        0.95    0.00    8.13
             Altitude                                     0.66          0.19        0.92    0.00    4.33
             Dist. from Coast                             0.87          0.91        0.12    0.45    1.00
             Ruggedness                                 126.89         99.33      103.53    0.00 474.34
Notes: Log Income per capita in 2000 (PPP 2005 US dollars). Engineering density measured by engineers per 100.000 male
workers. Engineering density measured by engineers per 100.000 male workers, sub-national. Engineering density measured
by engineers per 100.000 male workers, sub-national, scaled by national estimates of engineering stock. Population density is
number of individuals per 100 square kilometers in 1900. Pre-colonial population density measures the number of natives per
square kilometer in 1492. Literacy share of the population that is literate in 1900. Literacy share of the population that is literate
in 1900, sub-national. Railroad density measured as miles of track per 1000 square kilometers. Railroad density measured as
miles of track per 1000 square kilometers, sub-national. Colonial Good and Bad Institutions as generated by Bruhn and Gallego
(2011); none is excluded category. South is a dummy variable for whether the US state is a Southern state according to the US
census. Slavery is measured as a fraction of the population and is taken from Bergad (2008) and Nunn (2008). Lawyer density
measured by lawyers per 100.000 individuals. Mining is total mining output in 1860 in hundred thousand dollars. Spain is a
dummy for whether the country was a Spanish colony: Argentina, Chile, Mexico and Venezuela. Agriculture Suitability is an
index of probability of cultivation given cultivable land, climate and soil composition, from Ramankutty, Foley and McSweeney
(2002). Rivers captures the density of rivers as a share of land area derived from HydroSHEDS (USGS 2011). ; Temperature is
a yearly average in degrees celsius; Altitude measures the elevation of the capital city of the state in kilometers; and Rainfall
captures total yearly rainfall in meters, all are from Bruhn and Gallego (2011). Distance from the Coast from (Gennaioli et al.,
2013); Ruggedness of Terrain from Nunn & Puga (2012) .




                                                                 59
                          Table 2: Summary Statistics (County Level, US)


                  Variable                mean                       p50         sd    min   max
                  Ln Income               10.043                    10.032      0.219 9.165 11.360
                  Engineers                0.022                    0.000       0.038 0.000 0.299
                  Patents                  0.453                     0.274      0.602 0.000 6.041
                  Ln LGC distance          0.176                    0.281       0.683 -5.450 1.765
                  Rainfall                 0.077                    0.076       0.042 0.009 0.393
                  Altitude                 0.025                     0.023      0.019 0.000 0.175
                  Ruggedness              0.005                      0.003      0.006 0.000 0.043
                  Distance to river        0.017                     0.014      0.014 0.000 0.141
                  Average Temperature     -0.002                    -0.006      0.061 -0.160 0.188
                  Dist. from Coast         0.027                     0.023      0.020 0.000 0.102
                  Population Density      0.003                      0.001      0.049 0.000 2.055
                  Manufacturing GDP        0.042                    0.018       0.066 0.000 0.576
                  Slavery                  0.153                    0.019       0.215 0.000 0.925
                  Railroads               0.584                      1.000      0.493 0.000 1.000
                  Literacy                 0.759                    0.829       0.203 0.151 1.000
                  School Assistance 12-17 0.051                      0.055      0.018 0.000 0.090
                  Educational Score        0.073                     0.073      0.019 0.035 0.168
                  Lawyers                 0.001                      0.001      0.001 0.000 0.005
                  Physicians               0.002                    0.002       0.001 0.000 0.004
Notes: Log Income per capita in 2000. Engineering density measured by engineers per 100,000 male workers. Patents density
measured by patents per 100 habitants. Dist. to LGC is distance to Land Grant College measured by distance between the near
LGC and the county centroid. Population density is number of individuals per 100 square kilometers in 1880. Literacy share of
the population that is literate in 1880. Secondary is share of 14-17 years olds who report that they are attending school. Tertiary
is- percentage of workers who report completing one or more years of college. Railroad is a identifier variable if the county has
railroad. Slavery to county level is taken from Nunn (2008). Lawyer density measured by lawyers per 100.000 individuals.
Physicians density measured by physicians per 100.000 individuals. Mining is total mining output in 1860 in hundred thousand
dollars. Rivers captures the distance between the near river and the county. Rivers taken from HydroSHEDS (USGS 2011). ;
Temperature is a yearly average in degrees celsius; Altitude measures the elevation of the capital city of the state in kilometers;
and Rainfall captures total yearly rainfall in meters. Dist. to Coast is distance between the coast and the county centroid;
Ruggedness of Terrain from Nunn & Puga (2012). Manuf. Output is the manufacturing output per capita in 1880 taken from
NHGIS.




                                                               60
Table 3: Summary Regressions: Innovation Capacity (1900) vs Income per capita (2000)
(National Level, Americas)

                                             (1)           (2)         (3)          (4)           (5)
                      Engineering           28.4∗∗       25.5∗∗       27.0∗        15.7         25.0∗
                                           (14.26)      (12.87)      (14.35)      (27.09)      (14.71)
                      Pop Density                        0.03∗                                  0.03∗∗
                                                         (0.01)                                 (0.01)
                      Railroads                                        0.07                      0.06
                                                                      (0.06)                    (0.07)
                      Literacy                                                      0.02
                                                                                   (0.02)
                      Constant              8.7∗∗∗       8.8∗∗∗       8.5∗∗∗       8.3∗∗∗       8.6∗∗∗
                                            (0.17)       (0.22)       (0.17)       (0.42)       (0.32)
                      N                      273          225          273          273          225
                      N Countries             11            9           11           11            9
                      R2                     0.77         0.76         0.81         0.81         0.79

Notes: Dependent Variable is log subnational income per capita (2000). Engineering density measured by engineers per 100,000
male workers (coef scaled by 1000). Population density is number of individuals per 100 square kilometers in 1900. Railroad
density measured as miles of track per 1000 square kilometers. Literacy share of the population that is literate in 1900. Robust
and clustered SE in parenthesis. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.




                                                              61
        Table 4: Engineering Capacity as a Determinant of Income (County Level, US)
                           (1)      (2)               (3)           (4)           (5)           (6)           (7)           (8)
                          OLS      OLS               OLS           OLS           OLS           OLS           OLS           OLS
  Engineers            0.224*** 0.289***          0.188***     0.147***       0.184***     0.118***       0.112***     0.0938***
                         (0.03)   (0.04)            (0.03)        (0.03)        (0.03)        (0.03)        (0.03)        (0.03)
  Rainfall               0.0263   0.0328           0.0304         0.0557       0.0290         0.0739        0.0828         0.131
                         (0.06)   (0.07)            (0.07)        (0.07)        (0.07)        (0.06)        (0.07)        (0.08)
  Altitude             -0.00325  -0.0195           0.0335        0.00472        0.0418        0.0168       -0.0108       -0.0716
                         (0.03)   (0.08)            (0.07)        (0.08)        (0.07)        (0.07)        (0.07)        (0.06)
  Ruggedness            -0.0817  -0.0813            -0.104       -0.0295        -0.115       -0.0780       -0.0342       -0.0193
                         (0.08)    (0.1)            (0.09)        (0.09)        (0.09)        (0.08)        (0.08)        (0.08)
  River Dist.            0.0236   0.0148           0.0220         0.0187       0.0187         0.0352       0.0296         0.0293
                         (0.03)   (0.03)            (0.03)        (0.03)        (0.03)        (0.03)        (0.03)        (0.03)
  Temperature         -0.281*** -0.310***         -0.124**       -0.0113       -0.0781       -0.131*       -0.0361       -0.0784
                         (0.05)   (0.05)            (0.06)        (0.05)        (0.06)        (0.07)        (0.07)         (0.2)
  Coast Dist.         -0.171*** -0.236***        -0.168***      -0.128**     -0.174***     -0.176***     -0.129***       -0.0725
                         (0.05)   (0.05)            (0.05)        (0.05)        (0.05)        (0.05)        (0.05)        (0.09)
  Pop. Density                                   0.0930***     0.0952***     0.0934***     0.0831***     0.0896***     0.0993***
                                                   (0.007)       (0.007)       (0.007)       (0.009)       (0.008)       (0.008)
  Manuf. Output                                  0.193***       0.173***     0.212***       0.169***      0.161***     0.122***
                                                    (0.04)        (0.03)        (0.04)        (0.04)        (0.03)        (0.04)
  Slavery                                           -0.102         0.179       -0.0419       -0.0216        0.161        0.178**
                                                    (0.07)         (0.1)        (0.08)        (0.08)         (0.1)        (0.07)
  Railroad                                       0.161***       0.112***     0.144***       0.124***      0.106***     0.0957***
                                                    (0.04)        (0.03)        (0.03)        (0.03)        (0.03)        (0.02)
  Literacy                                                     0.463***                                   0.395***      0.502***
                                                                   (0.1)                                     (0.1)        (0.07)
  Secondary                                                                   0.120**                      0.00482        0.0374
                                                                               (0.05)                       (0.05)        (0.05)
  Tertiary                                                                                  0.142**         0.0471       0.0588
                                                                                             (0.05)         (0.04)        (0.04)
  Lawyers                                                                                   0.131***      0.112***      0.119***
                                                                                             (0.04)         (0.04)        (0.03)
  Physicians                                                                               -0.000377       -0.0463        0.0179
                                                                                             (0.05)         (0.04)        (0.03)
  N                      2380         1905          1905          1905          1905          1905           1905          1905
  R2                     0.188        0.238         0.313         0.362         0.319         0.345         0.371          0.476
  FE                      No           No            No            No            No            No             No            Yes

Notes: Dependent Variable is log subnational income per capita (2000). Engineering density measured by engineers per 100,000
male workers (coef. scaled by 1000). Patents density measured by patents per 100 inhabitants. Rainfall captures total yearly
rainfall in meters; Altitude measures the elevation of the capital city of the state in kilometers; Ruggedness of Terrain from
Nunn & Puga (2012); Rivers captures the distance between the near river and the county. Rivers taken from HydroSHEDS
(USGS 2011). Temperature is a yearly average in degrees celsius; Dist. to Coast is distance between the coast and the county
centroid. Population density is number of individuals per 100 square kilometers in 1880. Manuf. Output is the manufacturing
output per capita in 1880 taken from NHGIS. Slavery to county level is taken from Nunn (2008). Railroad is a identifier variable
if the county has railroad. Literacy is share of the population that is literate in 1880. Secondary is share of 14-17 years olds who
report that they are attending school. Tertiary is percentage of workers who report completing one or more years of college.
Lawyer and Physician density measured per 100 individuals.Engineers instrumentd using log distance to nearest Land Grant
College measured by distance between the near LGC and the county centroid, estimated by 2SLS and Buchinsky quantile 2SLS.
Robust and clustered SE in parenthesis. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.




                                                                62
Table 5: Engineering and Patenting Capacity as Determinants of Income: Patents & IV
(County Level, US)
                                              (1)           (2)         (3)       (4)      (5)         (6)
                                              OLS           OLS        OLS       OLS       IV    IV-Buchinsky
   Engineers                                                       0.0959*** 0.0817*** 0.111***    0.161***
                                                                      (0.03)    (0.02)   (0.03)      (0.04)
   Patents                                 0.125***       0.114***   0.105**  0.101*** 0.120***    0.0974**
                                              (0.04)        (0.04)    (0.04)    (0.04)   (0.04)      (0.04)
   Rainfall                                  0.0860         0.142     0.0815    0.145*   0.145*      0.147*
                                              (0.08)        (0.08)    (0.07)    (0.08)   (0.08)      (0.08)
   Altitude                                  -0.0370       -0.0964   -0.0359   -0.0932  -0.105*     -0.0873
                                              (0.07)        (0.06)    (0.08)    (0.06)   (0.05)      (0.06)
   Ruggedness                               -0.00177       0.00274   -0.0115 -0.000973 -0.00529    -0.000402
                                              (0.08)        (0.07)    (0.08)    (0.07)   (0.07)      (0.07)
   River Dist.                               0.0306        0.0268    0.0276     0.0256   0.0310      0.0292
                                              (0.03)        (0.03)    (0.02)    (0.03)   (0.03)      (0.03)
   Temperature                               -0.0248       -0.0681   -0.0385   -0.0731  -0.0507     -0.0750
                                              (0.07)         (0.2)    (0.06)     (0.2)    (0.2)       (0.2)
   Coast Dist.                              -0.119**       -0.0537  -0.118**   -0.0493  -0.0411     -0.0740
                                              (0.05)        (0.09)    (0.05)    (0.09)   (0.09)      (0.08)
   Pop. Density                            0.0889***     0.0983*** 0.0891*** 0.0985*** 0.0946***   0.102***
                                             (0.007)       (0.008)   (0.008)   (0.008)  (0.008)     (0.008)
   Manuf. Output                           0.135***       0.0886** 0.119***    0.0770*  0.0829*      0.0503
                                              (0.04)        (0.04)    (0.04)    (0.04)   (0.04)      (0.04)
   Slavery                                    0.175        0.189**     0.174 0.193*** 0.175***      0.166**
                                               (0.1)        (0.07)     (0.1)    (0.07)   (0.07)      (0.07)
   Railroad                                0.117***       0.106*** 0.114*** 0.103*** 0.189***      0.155***
                                              (0.03)        (0.02)    (0.03)    (0.02)   (0.04)      (0.04)
   Literacy                                0.408***       0.512*** 0.401*** 0.516*** 0.509***      0.506***
                                               (0.1)        (0.07)     (0.1)    (0.07)   (0.07)      (0.07)
   Secondary                                 0.0147        0.0448    0.0148     0.0470   0.0285      0.0445
                                              (0.05)        (0.05)    (0.05)    (0.05)   (0.05)      (0.05)
   Tertiary                                  0.0397        0.0533    0.0290     0.0446   0.0502      0.0297
                                              (0.04)        (0.04)    (0.04)    (0.04)   (0.04)      (0.04)
   Lawyers                                 0.113***       0.117*** 0.0916** 0.0986*** 0.112***     0.102***
                                              (0.04)        (0.04)    (0.04)    (0.03)   (0.04)      (0.03)
   Physicians                                -0.0496       0.0152    -0.0441   0.0161    0.0127      0.0155
                                              (0.04)        (0.03)    (0.04)    (0.03)   (0.03)      (0.03)
   N                                           1905          1905      1905      1905     1905        1905
   R2                                         0.370         0.476     0.376      0.480
   FE                                           No            Yes       No        Yes      Yes         Yes
   F statistical (robust)                                                                14.023       25.45
   F statistical (robust and cluster)                                                     7.755
   F Cragg-Donald                                                                        20.34

Notes: Dependent Variable is log subnational income per capita (2000). Engineering density measured by engineers per 100,000
male workers (coef. scaled by 1000). Patents density measured by patents per 100 inhabitants. Rainfall captures total yearly
rainfall in meters; Altitude measures the elevation of the capital city of the state in kilometers; Ruggedness of Terrain from Nunn
& Puga (2012); Rivers captures the distance between the near river and the county. Rivers taken from HydroSHEDS (USGS
2011). Temperature is a yearly average in degrees celsius; Dist. to Coast is distance between the coast and the county centroid.
Population density is number of individuals per 100 square kilometers in 1880. Manuf. Output is the manufacturing output
per capita in 1880 taken from NHGIS. Slavery to county level is taken from Nunn (2008). Railroad is a identifier variable if the
county has railroad. Literacy share of the population that is literate in 1880. Secondary is share of 14-17 years olds who report
that they are attending school. Tertiary is percentage of workers who report completing one or more years of college. Lawyer
and Physician density measured per 100 individuals.Engineers instrumentd using log distance to nearest Land Grant College
measured by distance between the near LGC and the county centroid, estimated by 2SLS and Buchinsky quantile 2SLS. Robust
and clustered SE in parenthesis. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01. In Column 5, F statistical of instrument’s relevance is
14.023 with robust standard errors and 7.755 with robust and clustered to state level standard errors. Cragg-Donald F statistical
is 20.34. Using quantile regression in Column 6 F statistical of instrument’s relevance is 25.45 with robust standard errors.




                                                               63
   Table 6: Innovative Capacity as a Determinant of Income (Fixed Effect, Subnational)

                                                           (2)
                                                          (1)      (3)
                                     Engineering         1.1∗∗∗ 0.6∗∗∗
                                                         1.0∗∗∗
                                                         (0.28) (0.17)
                                                         (0.33)
                                     Pop Density         0.03∗∗ 0.03∗∗∗
                                                         (0.01) (0.01)
                                     Literacy                    0.01∗∗∗
                                                                 (0.00)
                                                     ∗∗∗     ∗∗∗
                                     Constant    9.5     9.4     8.9∗∗∗
                                                 (0.41) (0.38) (0.36)
                                     N            170     166      166
                                     N Countries    6       6       6
                                      2
                                     R            0.11    0.18    0.37

Notes: Dependent Variable is log subnational income per capita (2000). Full sample is Argentina, Chile, Colombia, Mexico,
US and Venezuela. Engineering density measured by engineers per 100,000 male workers (coef. scaled by 1000). Population
density is log number of individuals per 100 square kilometers in 1900. Literacy is share of the population that is literate in
1900. Bootstrapped clustered SE in parenthesis. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.




                                                             64
       Table 7: Mechanisms through which Engineering in 1900 Drives Income in 2000.

                                                                      All     New World
                               Inputs
                               R&D/GDP                  0.94                       0.96
                               Firm Innovative Capacity 0.94                       0.94
                               Modern Management        0.93                       0.93

                               Outputs
                               Patents                                0.95         0.98
                               Technological Adoption                 0.84         0.94

Notes: Table reports the correlation between engineering density in 1900 and three inputs and two outputs in the 2000. Inputs:
1. R&D expenditures as a share of GDP. 2. Firm capacity for innovation ranging from pure licensing to pioneering their own
new products and processes. “In your country, how do companies obtain technology? [1 = exclusively from licensing or imitating
foreign companies; 7 = by conducting formal research and pioneering their own new products and processes]. 3. A globally
consistent measure of management quality from Bloom & Van Reenen (2010) and in particular, the sum of the scores on the
two the questions dealing with how firms identify new production processes to adopt. On the output side, we have 1. Comin
& Hobijn (2010); Comin & Ferrer (2013); Comin et al. (2008)’s measure of technological adoption at the extensive margin,
averaging their industrial and sectoral scores and 2. patent applications filed under the Patent Cooperation Treaty (PCT) per
million population as tabulated by the World Economic Forum (2008-9)(World Economic Forum et al., 2012).




                                                             65
                 Table 8: Determinants of Innovative Capacity (County Level, US)
                                                         Engineers                 Patents
                                                    OLS       Buchinsky      OLS       Quantile
                                 Pop. Density    0.02∗∗∗        0.01∗∗∗     0.3∗∗∗       0.2∗∗∗
                                                  (0.004)       (0.002)     (0.06)       (0.02)
                                 South           -0.02∗∗∗        -0.001    -0.2∗∗∗      -0.2∗∗∗
                                                  (0.004)       (0.002)     (0.05)       (0.02)
                                 Slavery         -0.02  ∗∗∗    -0.009∗∗    -0.7∗∗∗      -0.3∗∗∗
                                                  (0.005)       (0.004)     (0.07)       (0.05)
                                 River dist.        -0.09         -0.03       -1.7         -0.7
                                                   (0.08)        (0.04)      (1.3)        (0.5)
                                 Temperature      0.09∗∗∗          0.02        0.4       0.4∗∗
                                                   (0.03)        (0.01)      (0.6)        (0.2)
                                 Rainfall          -0.08∗         -0.03      -1.2∗      -1.3∗∗∗
                                                   (0.04)        (0.02)      (0.7)        (0.3)
                                 Altitude           -0.09         -0.05      3.8∗        2.4∗∗∗
                                                   (0.10)        (0.05)      (2.1)        (0.6)
                                 Coast Dist.        -0.06          0.02    -4.5∗∗∗        -0.7∗
                                                   (0.06)        (0.03)      (0.9)        (0.4)
                                 Ruggedness         0.5∗           0.2    -16.0∗∗∗      -9.1∗∗∗
                                                    (0.3)         (0.1)      (5.3)        (1.9)
                                 N                  1912          1912       1907         1907
                                 R2                 0.140                   0.231


Notes: Dependent Variable is innovative capacity measured by engineers per 100,000 male workers and patents per 100000
inhabitants. Pre-colonial population density measures the number of natives per square kilometer in 1492. South a dummy
capturing southern US states; slavery as tabulated from the 1860 Census as compiled in Nunn (2008). Rainfall captures total
yearly rainfall in meters; Altitude measures the elevation of the capital city of the state in kilometers; Ruggedness of Terrain
from Nunn & Puga (2012); Rivers captures the distance between the near river and the county. Rivers taken from HydroSHEDS
(USGS 2011). Temperature is a yearly average in degrees celsius; Dist. to Coast is distance between the coast and the county
centroid. Population density is number of individuals per 100 square kilometers in 1880.Geographical controls include river
density, average temperature, rainfall, altitude, distance from a coast, and ruggedness of terrain. Robust SE in parenthesis. ∗
p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.




                                                              66
                                               Table A1: Engineering Density as a Determinant of Patents & IV (County Level, US)
                                                                              (1)           (2)          (3)           (4)           (5)             (6)          (7)           (8)            (9)
                                                                             OLS           OLS          OLS           OLS           OLS             OLS          OLS            IV      IV-Buchinsky
                                                        Engineers          0.308***     0.295***     0.188***      0.188***      0.187***        0.124***     0.120***       -0.0316       0.133***
                                                                            (0.04)        (0.04)       (0.03)        (0.03)        (0.03)          (0.03)       (0.03)        (0.03)         (0.04)
                                                        Rainfall                          -0.167       -0.192        -0.193        -0.192          -0.134       -0.141        -0.149         -0.142
                                                                                           (0.2)        (0.2)         (0.2)         (0.2)           (0.1)        (0.1)         (0.1)          (0.1)
                                                        Altitude                          0.211*      0.215**       0.215**       0.212**        0.217**       0.213**      0.215***       0.217***
                                                                                           (0.1)       (0.09)        (0.09)        (0.09)          (0.08)       (0.08)        (0.07)         (0.07)
                                                        Ruggedness                        -0.184      -0.187*       -0.189*       -0.183*         -0.175*     -0.181**      -0.177**       -0.180**
                                                                                           (0.1)        (0.1)         (0.1)         (0.1)          (0.09)       (0.08)        (0.08)         (0.08)
                                                        River Dist.                      -0.0221       0.0283        0.0286        0.0297         0.0323       0.0362         0.0374         0.0401
                                                                                          (0.05)       (0.04)        (0.04)        (0.04)          (0.03)       (0.03)        (0.03)         (0.03)
                                                        Temperature                      -0.0269     -0.00366      -0.00887       -0.0203         0.0183       -0.0519       -0.0503        -0.0505
                                                                                           (0.2)        (0.1)         (0.1)         (0.1)           (0.1)        (0.1)         (0.1)          (0.1)
                                                        Coast Dist.                     -0.253**     -0.202**      -0.202**      -0.195**       -0.227***    -0.229***     -0.244***      -0.254***
                                                                                          (0.10)       (0.08)        (0.08)        (0.08)          (0.08)       (0.08)        (0.08)         (0.07)
                                                        Pop. Density                                 0.0182**      0.0182**      0.0183**         0.0103       0.00874       0.00971        0.0117*
                                                                                                      (0.008)       (0.008)       (0.008)         (0.007)      (0.007)       (0.008)        (0.007)
                                                        Manuf. Output                                0.507***      0.508***      0.495***       0.459***      0.442***      0.469***       0.431***
                                                                                                       (0.06)        (0.06)        (0.06)          (0.06)       (0.06)        (0.06)         (0.06)
                                                        Slavery                                     -0.0802***    -0.0854***    -0.0985***      -0.0692**    -0.144***     -0.149***      -0.169***
                                                                                                       (0.03)        (0.02)        (0.03)          (0.03)       (0.03)        (0.03)         (0.03)
                                                        Railroad                                      -0.0483       -0.0473       -0.0440      -0.0833***    -0.0734**     -0.0660**     -0.0938***
                                                                                                       (0.03)        (0.03)        (0.03)          (0.03)       (0.03)        (0.03)         (0.03)
                                                        Literacy                                                    -0.0128                                   -0.144**      -0.152**       -0.155**
                                                                                                                     (0.04)                                     (0.06)        (0.06)         (0.06)
67




                                                        Secondary                                                                 -0.0672*                   -0.0950**    -0.0954***     -0.0991***
                                                                                                                                   (0.04)                       (0.04)        (0.03)         (0.04)
                                                        Tertiary                                                                                 0.104**     0.141***      0.157***        0.135***
                                                                                                                                                  (0.04)        (0.04)        (0.04)         (0.04)
                                                        Lawyers                                                                                 0.203***     0.203***      0.236***        0.219***
                                                                                                                                                  (0.06)        (0.05)        (0.06)         (0.06)
                                                        Physicians                                                                              -0.00798       0.0181         0.0178         0.0171
                                                                                                                                                  (0.03)        (0.04)        (0.04)         (0.03)
                                                        N                    1905        1905          1905          1905          1905            1905          1905          1905           1905
                                                        R2                   0.316       0.340         0.471         0.471         0.472          0.505          0.511
                                                        FE                    Yes         Yes           Yes           Yes           Yes            Yes            Yes         Yes             Yes


     Notes: Dependent Variable is patents per 100,00 inhabitants (1900). Engineering density measured by engineers per 100,000 male workers (coef. scaled by 1000). Patents density measured by patents per 100 inhabitants. Rainfall captures total
     yearly rainfall in meters; Altitude measures the elevation of the capital city of the state in kilometers; Ruggedness of Terrain from Nunn & Puga (2012); Rivers captures the distance between the near river and the county. Rivers taken from
     HydroSHEDS (USGS 2011). Temperature is a yearly average in degrees celsius; Dist. to Coast is distance between the coast and the county centroid. Population density is number of individuals per 100 square kilometers in 1880. Manuf. Output
     is the manufacturing output per capita in 1880 taken from NHGIS. Slavery to county level is taken from Nunn (2008). Railroad is a identifier variable if the county has railroad. Literacy share of the population that is literate in 1880. Secondary
     is share of 14-17 years olds who report that they are attending school. Tertiary is percentage of workers who report completing one or more years of college. Lawyer and Physician density measured per 100 individuals.Engineers instrumentd
     using log distance to nearest Land Grant College measured by distance between the near LGC and the county centroid, estimated by 2SLS and Buchinsky quantile 2SLS. Robust and clustered SE in parenthesis. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗
     p < 0.01. In Column 5, F statistical of instrument’s relevance is 14.023 with robust standard errors and 7.755 with robust and clustered to state level standard errors. Cragg-Donald F statistical is 20.34. Using quantile regression in Column
     6 F statistical of instrument’s relevance is 25.45 with robust standard errors.
                  Table A2: Innovative Capacity as a Determinant of Income (US)

                               (1)          (2)         (3)           (4)          (5)          (6)           (7)
          Engineering         0.5∗∗∗      0.8∗∗∗       0.4∗∗∗        0.4∗∗       0.6∗∗∗        1.2∗∗∗       1.2∗∗∗
                              (0.15)      (0.15)       (0.14)       (0.18)       (0.11)        (0.40)       (0.33)
          Pop Density                     0.04∗∗∗                                                           -0.008
                                          (0.01)                                                            (0.03)
          Literacy                                     0.003∗                                                0.01
                                                       (0.00)                                               (0.01)
          Lawyers                                                    0.2                                      -0.2
                                                                    (0.21)                                  (0.23)
          Railroads                                                             0.002∗∗∗                    0.001
                                                                                 (0.00)                     (0.00)
          Mining                                                                  -14.7                    -23.1∗∗∗
                                                                                (10.58)                     (7.12)
          Slavery                                                                             -0.0002        0.006
                                                                                               (0.00)       (0.00)
          South                                                                                -0.004        -0.05
                                                                                               (0.10)       (0.07)
          Constant            10.6∗∗∗     10.4∗∗∗      10.4∗∗∗     10.6∗∗∗       10.5∗∗∗      10.6∗∗∗       9.3∗∗∗
                              (0.02)      (0.06)       (0.15)      (0.07)        (0.03)        (0.06)       (0.91)
          N                     51          51           51          51            44            38            34
          R2                   0.14        0.40         0.17        0.17          0.47          0.41         0.60

Notes: Dependent Variable is log subnational income per capita (2000). Engineering density measured by engineers per 100,000
male workers (coef. scaled X 1000). Population density is log number of individuals per square kilometer in 1900. Literacy share
of the population that is literate in 1900. Lawyer density measured by lawyers per 100.000 individuals (coef. scaled X 1000).
Institutional Controls: Slavery and South” ”Railroad density measured as miles of track per 1000 square kilometers. Mining is
total mining output in 1880 in US 100,000 dollars. Robust SE in parenthesis. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.




                                                              68
        Table A3: Innovative Capacity as a Determinant of Income (US, instrumented)

                                             (1)         (2)          (3)         (4)         (5)
                       Engineering           0.9∗      1.6∗∗∗        0.8∗∗       0.9∗∗       2.1∗∗
                                            (0.50)     (0.51)       (0.36)      (0.40)      (1.06)
                       Pop Density                     0.07∗∗∗                              0.08∗∗
                                                       (0.02)                               (0.03)
                       Literacy                                     0.002                   -0.002
                                                                    (0.00)                  (0.00)
                       Lawyers                                                   -0.02       -0.3
                                                                                (0.25)      (0.34)
                       Constant            10.6∗∗∗     10.3∗∗∗     10.4∗∗∗      10.6∗∗∗     10.5∗∗∗
                                           (0.08)      (0.11)      (0.16)       (0.08)      (0.25)
                       N                     51          51          51            51         51
                       R2                   0.02        0.06        0.09          0.02         .

Notes: Dependent Variable is log subnational income per capita (2000). Engineering density measured by engineers per male
workers (coef. scaled X 1000). Population density is log number of individuals per square kilometer in 1900. Literacy share
of the population that is literate in 1900. Lawyer density measured by lawyers per 100.000 individuals (coef. scaled X 1000).
Robust SE in parenthesis. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01. R2 in specification 5 is close to 1 but not calculated.




                                                            69