Getting It:
Language Standardization and the Industrial Revolution
9 March 2015
Leonard Dudley*
Université de Montréal
Abstract
Although the Industrial Revolution was preceded by a sharp rise in literacy in the West,
paradoxically, countries with the highest literacy rates were late to industrialize. Here a
new data set comprising 117 important innovations between 1700 and 1850 and 201
urban regions suggests a possible explanation. In the three states that contributed virtually
all of these innovations (Britain, France and the USA), rising literacy was merely the first
step toward the formation of large networks of people speaking standardized languages.
Such networks proved particularly important for innovations requiring collaboration.
Elsewhere, where language standardization was delayed, industrialization also came later.
JEL Codes: O3, N6
Keywords: Industrial Revolution, innovation, language standardization, cooperation,
networks
Paper to be presented at the meeting of the European Historical Economics Society, Pisa,
September 4-5, 2015.
2
One of the mysteries of the Industrial Revolution is the absence of any close
relationship between education and productivity growth rates (Mitch, 1999, 255-256).
During the seventeenth and eighteenth centuries, literacy rates rose especially rapidly in
the Protestant states of Northern Europe. By 1800 male literacy in northern Germany, the
Netherlands and Scandinavia, measured by signature rates, ranged from 70 to over 90
percent (Reis, 2005, 202). Endogenous growth theory suggests that these regions should
have been the first to experience rapid productivity growth. Yet measured by rates of
innovation, these regions contributed little before 1850. Instead, for a century and a half
after 1700, virtually all of the important innovations were developed in Britain, its
American offshoots and northern France, where literacy rates were considerably lower
than in northern Europe.1
This paper offers a possible explanation for the apparent orthogonality of
education and productivity growth in the West before 1850. In a word, literacy was
merely a first step toward creating standardized language networks that made possible
cooperation to innovate between people from different regions. The argument is based on
three propositions. First, because industrial technologies were growing in complexity,
cooperation between individuals with different skill sets was increasingly important.2
Think, for example, of Scottish engineer James Watt and English entrepreneur Matthew
Boulton. Second, cooperation between individuals required that they write and speak the
same vernacular language. Had Watt and Boulton been born a century earlier, they would
have had difficulty communicating with each other. However, by the second quarter of
the eighteenth century, grammar and spelling texts used in both England and southern
Scotland were insisting on standardized ways of spelling and pronouncing the English
language for boys “who are to be put to Trades” (Cohen, 1977, 48, 81-82; Frank, 1994,
55).Third, there was some minimum size for a language network that would allow it to
generate cooperative innovations such as the Boulton and Watt steam engine. In the first
decade of the nineteenth century, England, Scotland and Wales had a combined
1
In 1800, the male literacy rate in England was only about 60 percent (Cressy, 1980, 177) and in France 50
percent (Houdaille, 1977, 68). If slaves and Native Americans are included, American literacy levels were
probably even lower than those in England (Graff, 1991, 252).
2
In eighteenth-century Britain, as Jacob (1997, 105-108) explained, technological innovation was typically
the result of partnership between a self-educated engineer and an unrelated industrial entrepreneur.
3
population of 11 million (Maddison, 2001, 247), while France north of the Loire had a
population of 15 million. In comparison, the United Provinces had a population of only
1.9 million, while Sweden and Prussia each had only 1.5 million. Arguably, if network
effects are taken into account, these and other continental states were too small to
generate many innovations (Burnham, 2005).
Previous studies have offered several plausible explanations for the burst of
innovation that occurred during the Industrial Revolution. The dominant approach has
emphasized the supply side. North (1990, 3) stressed the importance of institutions
(“humanly devised constraints that shape human interaction”). Similarly, Acemoglu and
Robinson (2012, 82) pointed to the emergence of “inclusive” political institutions as the
keys to Britain’s economic success. A corollary suggested by Mokyr (2002, 34) was that
the innovating society must have an ideology that favors new ways of practical thinking
(“the expansion of useful knowledge”). Also on the supply side, Landes (1998, 219)
emphasized Britain's individualistic culture, while Clark (2008, 11) argued that Britain’s
industrial lead had a genetic/cultural component due to high fertility rates of the
“economically successful.”
Allen (2009), in contrast, proposed a demand-side approach to innovation,
suggesting that institutions and ideology were not enough. New technologies, he argued,
were developed through the response of entrepreneurs to the opportunities signaled by
changes in factor prices. In the case of eighteenth-century Britain, the rising costs of labor
and conventional sources of energy led skilled artisans and their backers to develop new
processes that replaced labor with machinery and charcoal, wind and water with coal
(Crafts, 2010). Human capital adds another dimension to the demand-side story, linking
education levels to growth in productivity (Galor et al., 2009, 144).
There are several problems, however, with these explanations. Relying on
differences between nations as a whole they cannot explain why even in the innovating
countries, development of new technologies tended to be confined to a small number of
regions. As will be shown below, some three-quarters of the important innovations cited
by historians of technology were developed in three narrow bands of territory in England,
4
northern France and the United States. Other regions in the West with similar institutions
and factor prices, or higher levels of literacy failed to innovate during the century and a
half prior to 1850.
What did these particular regions of Britain, France and America have that other
regions in the West lacked? Section 1 presents an original data set composed of 117
important innovations identified by the urban regions where they were developed.
Following the first proposition stated above, the approach here departs from earlier
studies in distinguishing between cooperative and non-cooperative innovations.3
A second difficulty overlooked in most previous research is the effect of
communication on strategic behavior, that is, situations where the outcome for each
individual depends on the actions of other agents. Smith (2010) observed that there are
several ways in which communication through language may facilitate cooperation. If
information is costless, “cheap talk” may allow individuals with convergent interests to
coordinate their strategies (Farrell and Rabin, 1986). Language also permits information
concerning the reputation of individuals to circulate in groups (Smith 2010, 236-237), for
example, the network of several hundred Maghribi traders studied by Greif (2006, 5861). Finally, in larger groups, if acquiring information has a cost that is negatively
correlated with an individual’s competence, players can credibly signal their capacities to
one another by investing in formal education (Spence, 1973).
Through these different mechanisms, learning a standard language permitted the
formation of social capital through the building of trust, as described by Sunderland
(2007). In accord with the second proposition above, therefore, Section 2 examines the
innovations data by language area. It is shown that in Britain, France and America there
was a clustering of cooperative innovations around the main cultural center. Data from
previous studies on the intelligibility of dialects across regions suggests that this
concentration is explained in part by the gradual diffusion of linguistic standards outward
from the cultural centers.
3
Allen (1983) studied “collective invention” among nineteenth-century American iron and steel firms that
shared the results of successful technological refinements with local rivals; however, all were price takers
and there was no explicit research and development.
5
A third difficulty with previous discussions of the Industrial Revolution is neglect
of the process of innovation. Over sixty years ago Ashton (1948, 13-15) proposed that
innovation occurred through the combination of apparently unrelated ideas. A related
consideration is Metcalfe’s Law and its extensions, which suggest that the number of
interactions in a network increases more than proportionally to the number of nodes
(Briscoe et al., 2006). If these insights are valid, it is important to model explicitly the
connections between individuals in a society.
To verify the effects of network size, following the third proposition, Section 3
proposes a model of collaboration between potential innovators in two cities. The
probability of successful innovation depends on the populations of each city and on the
distance between them. An empirical version of this model is then expanded to include
the institutional, factor-price and human-capital explanations of innovation as special
cases. The results in Section 4 along with the robustness and the sensitivity tests in
Section 5 help explain why literacy rates were unrelated to productivity growth before
1850. Literacy was a necessary first step, but rapid innovation required that the
vernacular language then be standardized among large networks of people.
6
1. Cooperating to Innovate
The introduction proposed an approach to innovation during the Industrial
Revolution that emphasized the effect of language standardization on the willingness to
cooperate. Consider now a data set that allows comparison of this linguistic approach
with those of previous studies.
Table 1 presents 117 important innovations selected by a panel of historians of
technology. To qualify for this table, an innovation had to be mentioned by two or more
experts from four different countries. The authors in question were Donald Cardwell
(1991) of Britain, Maurice Daumas (1979) of France, Joel Mokyr (1990) born in the
Netherlands and living in the United States, and Akos Paulinyi (1989), born in Hungary
and residing in Germany. The Encyclopedia Britannica served as arbitrator in 30 cases of
a reference by a single author. This methodology is subject to bias, since the Daumas
study was a primary source for the other three authors. However, the present approach
has the advantage of including important technologies never patented, such as Smeaton’s
breast wheel and Trevithick’s high-pressure engine (Mokyr, 2009, 91), and excluding
numerous many others that were patented but never applied.
A detailed study of the 117 innovations reveals that 54 of them involved
collaboration between two or more principals. An example is the partnership for the
water frame by wig-maker Richard Arkwright and clock-maker John Kay, in which the
former specialized in conception and marketing and the latter in technical development.
Since each contributed services that could not readily be bought on the market, the
withdrawal of either would have meant the end of the project (Hills, 1970, 63-66).
Defined in this way, the cooperative innovations are underlined in Table 1. Note that
these innovations tended to be technologically relatively complex. Compare, for example,
the atmospheric steam engine, machine spinning of cotton or smelting with coke with
three non-cooperative innovations from the same period such as the flying shuttle, the
quadrant or stereotyping.
Allen (1983) and Nuvolari (2004) noted that innovations within a given country
were not spread evenly across its territory but rather tended to cluster around a few
7
centers. This finding suggests that the unit of observation should be an urban region
within a state. Accordingly, the units of observation were 201 cities, each of which had at
least 7,000 inhabitants in 1700 (the population of Birmingham in that year). Since the
goal of the study was to explain variations within countries, all of the European cities
were required to be at least as close to London as the most distant innovating city – Como
in northern Italy. This assumption has the advantage of keeping the number of zero
observations within a reasonable limit. Also included were three American cities – New
York, Philadelphia and Boston.
An examination of Table 1 reveals that innovation during the Industrial
Revolution was highly concentrated geographically. As Figure 1 shows, one corridor ran
from London up to Birmingham and Manchester. In France, there was a band stretching
from Le Havre through Paris to Lyon. Finally in the United States, most of the
innovations were developed in a stretch down the east coast, from Connecticut to the
Potomac River. These three strips accounted for 74 percent of all the innovations in the
sample and 82 percent of the cooperative innovations.
It is interesting to note that the English and French bands of innovation in Figure
1 include London and Paris, the main cultural centers of the corresponding societies.
These are precisely the areas within which the English and French standard national
languages first took form (Crystal, 2003, 54; Harris, 1990, 211). The next section will
explore whether the distance from these cultural centers is related to the propensity to
innovate.
8
2. The Language-Standardization Approach
Why might three-quarters of the important innovations of the Industrial
Revolution, and four-fifths of those that were cooperative have been clustered around the
main cultural centers in three states? This section suggests that prior to 1850 there were
barriers to innovation that increased with distance from the cultural center.
Consider the distance of each innovating city in the sample of Table 1 from the
cultural center of its society. For Britain, France, the Netherlands, Denmark, Austria and
Italy, the capital city was chosen. For the United States, the cultural center selected was
New York. As for Germany, the city chosen was Leipzig, a major publishing center and
after 1632 the location of the largest book fair. For multicultural Switzerland and
Belgium, the centers chosen were Paris, Leipzig or Amsterdam depending on whether the
city's principal language was French, German or Dutch, respectively. In Figure 2, the
resulting distances are plotted against the year of the innovation.
The upper graph of Figure 2 shows the plot for the complete sample of 117
innovations. For example, in the lower left is displayed the distance of 190 km from
London to the region of Birmingham, where Newcomen and Calley invented the
atmospheric steam engine in 1712. In the middle of the graph is the distance of 750 km
from Paris to Montpellier where Lenormand invented the parachute in 1783. The vertical
and horizontal heavy dark lines indicate that the mean distance of the innovations from
their society's cultural center was 212 km and the mean year was 1794.
The lower graph displays the corresponding data for the 54 cooperative
innovations. There are several features to notice here. First, the mean distance from the
cultural center for the cooperative innovations was only 169 km. Second, the mean year
for the subsample was a relatively-late 1799. These differences in means for distance and
years compared to the sample as a whole are significant at the one and ten percent levels
respectively. Third, it is striking that all of the cooperative innovations came from
Britain, France or the United States. In other words, before 1850 there is little evidence of
successful cooperation to innovate on the Continent outside of France.
9
The graphs in Figure 2 suggest three questions. First, why were the 54 cooperative
innovations developed closer to the cultural centers than the 63 non-cooperative ones
were? The graphs are consistent with the notion that between 1700 and 1850, there was
some factor related to distance that inhibited cooperation. Second, why were the
cooperative innovations developed later than the non-cooperative innovations? This
evidence suggests that developing the capacity to cooperate took time. Finally, why were
Britain, France and the United States the only countries to develop cooperative
innovations during the century and a half from 1700 to 1850?
What was it that favored cooperation in this way? A recent experiment in
Paleolithic tool-making suggests that our ancestors’ capacity to innovate depended to a
great extent on improvements in their ability to communicate with one another. A study
with 184 participants divided into five groups according to the method of transmitting
information indicated that innovations in tool-making which began 2.5 million years ago
were related to improvements in teaching and language (Morgan et al. 2015).
But is there really any indication that prior to 1700 people from neighboring
regions had difficulty understanding one another? To take an example from the second
half of the sixteenth century, an English official arriving at the Scottish court in
Edinburgh was obliged to speak French in order to make himself understood (Bain, 1898,
322). London English and Edinburgh Scots were evidently not mutually comprehensible.
Such difficulties in communication among people from different regions of Britain
persisted well into the twentieth century (Bragg, 2003, 25). The situation was similar in
France. In the seventeenth century, Louis XIV was unable to understand people in
Picardy barely 30 miles (50 km) north of Paris who were complaining of food shortages
(Bell, 2001, 82).
During the latter half of the seventeenth century, these linguistic barriers began to
dissolve. Thanks largely to the efforts of publishers in London and Paris, a credible signal
of the “quality” of one's interlocutor appeared in both Britain and France. Standard
written and spoken forms of the English and French languages had emerged – first in the
areas around the main cultural center and then gradually in more distant regions (Crystal,
10
2003, 67, 72; Barlow and Nadeau, 2007, 94-95). By the second quarter of the eighteenth
century, despite slight differences in pronunciation, the written and spoken language of
educated people across Britain was gradually becoming standardized around the dialect
of London (Freeborn 1992, 180). Meanwhile, in north-central France, the popularity of
inexpensive volumes of the Bibliothèque bleue showed that a standardized form of the
French language was spreading beyond Paris (Bell, 2001, 85).
Ideally, to quantify the degree of language standardization spatially, one would
like to have a measure of the “distance” between the dialect of each region and the
standardized language of the country’s cultural center. Since such data do not exist for
the period under study, an alternative approach is required. Consider for each language
region the year of publication of the first monolingual dictionary. In order for a publisher
to be willing to launch such a publication, he had to be confident that there were both a
large market of potential buyers and a desire among them to master a single standard of
spelling and uniform definitions. The date of a society’s first dictionary is thus an
approximate measure of the size and degree of integration of its language network during
the period under study.
The strong central governments of England and France had long decreed that only
the vernaculars of their respective capitals be legally recognized. 4 Then in 1658, Edward
Phillips, a nephew of the poet John Milton, published the first monolingual English
dictionary that was not simply a collection of hard words to spell (Jackson, 2002, 36). In
1680, César-Pierre Richelet, a law graduate from the Champagne region north-east of
Paris, published the first monolingual dictionary in the French language (Bray, 1986,
235-242). Table 2 compares the years of these first dictionaries in England and France
with the later dates for the corresponding publications of other states. Let us assume that
a potential partner of “high” quality had a lower cost of learning to speak and write a
standardized language than one of “low” quality. Then the use of a standardized language
rather than the dialect of one’s place of origin could be interpreted as a signal of
reliability.
4
Under the Pleading in English Act of 1362, only the English tongue could be used in pleas before the
English courts. Similarly, in France, the Ordinance of Villers-Cotterêts in 1539 had insisted on the use of
French in all legal acts.
11
Supporting this anecdotal evidence from previous centuries is a series of recent
studies on the mutual intelligibility of dialects in Europe. In the Germanic countries, even
today, people from neighboring regions have considerable difficulty understanding one
other's dialects. In one recent, Dutch subjects from the region bordering on Germany
were able to comprehend only 57 percent of words spoken in Low German (Gooskens et
al., 2011, Table 11). A second study of young Danes from Copenhagen indicated that the
average intelligibility of individual words ranged from about 50 per cent for dialects from
neighboring regions to about a third for distant regions within the same language family
(Gooskens et al., 2008, 74). As the bottom line of Table 3, shows, however,
comprehension of standardized Norwegian and Swedish words averaged over 60 percent.
Table 3. Intelligibility of Scandinavian dialects to young Danes from Copenhagen
according to distance and degree of standardization
Dialect
Distance (km)
Intelligibility (%)
Neighboring regions
301
48
Distant regions
956
35
Standard Danish
0
99
Standard Swedish and Norwegian
628
62
Source: Gooskens et al. (2008; 66, 74).
The negative impact of distance on comprehension in Table 3 mirrors the negative
effect of distance on the frequency of cooperative innovation in Figure 2. In addition, the
result that standardization improves comprehension in Table 3 echoes the finding in
Figure 2 that all of the cooperative innovations were in the three countries that were the
first to standardize their languages. Therefore, it would be seem worthwhile to test the
impact of language standardization on the propensity to innovate.
12
3. Model Specification
This section begins by proposing a simple theoretical model capable of capturing
the effect of language on innovation. The result is a Poisson specification that is adapted
to accept other approaches to innovation as special cases. Missing observations of the
explanatory variables are then estimated by the stochastic-regression method.
Assume that there are two cities, with populations
and
respectively. Initially
their dialects are sufficiently different to prevent communication between their residents.
Now let a standardized language be introduced into their populations. The number of new
pairings, x, made up of one resident from each city made possible by this development is
given by:
.
If the π1 of these potential partnerships leads to a successful innovation in city 1,
the total number of innovations there, y1, is:
.
Of course, not all of these innovations need take place in city 1. More realistically, we
assume that there may be congestion, as represented by the parameters
and
, when
individuals from city 2 try to find partners in the city 1. Moreover, because of transaction
costs, the probability of a successful pairing will be a decreasing function of the distance,
d, between the two cities. The expected number of innovations in city 1 will then be:
,
Take logs:
or
, where
(1)
Equation (1) expresses the number of innovations produced after the introduction
of a standardized language as a Poisson distribution that has as explanatory variables the
13
logarithms of the population of two cities and the distance between them – in effect, a
gravity model of innovation. The next step is to integrate this networking approach into a
specification that incorporates the supply and demand approaches used in previous
studies. Define the dependent variable as the number of innovations of a given type that
occurred in the region of a given city during each half-century between 1700 and 1849.
Since such innovations may be considered rare events, we should use an estimation
method appropriate for count data. The variance of this variable in our sample (0.182) is
considerably greater than the mean (0.065); accordingly, a regular Poisson distribution as
in equation (1) is not appropriate. To allow for over-dispersion, the study will use a
negative-binomial specification.
y = exp (Xβ + ε + u),
(2)
where yijt is the number of innovations in city i of type j in period t, xijt is a vector of
explanatory variables, β is a vector of parameters,
is a random variable and exp (
)
follows a gamma distribution with parameters α and 1/α.
Consider next the explanatory variables. Unobserved supply-side influences may
be assumed to be picked up by fixed-effects variables. Accordingly, five dummy
variables correspond to the principal nations included in the sample, while 1750 and 1800
represent the half-centuries after 1750 and 1800 respectively.
A second group of variables picks up the influence of demand conditions. Data for
eight urban regions and three periods permitted estimation of the following equation:
Energy price/Wage rate = 1.044 - 0.691*Coal + 0.442*Catholic - 0.282*Port
(0.076) (0.096)
(0.089)
(0.075)
- 0.172*1800, R2 = 0.873, Root MSE = 0.2080,
(3)
where the figures in brackets are the robust standard errors. From the estimated
coefficients and root MSE, values for the 579 missing observations of Relative energy
prices were then imputed by the stochastic-regression method.
14
As mentioned earlier, Galor st al. (2009) suggested that a greater stock of human
capital would stimulate innovation. The Literacy rate is an approximate measure of the
abundance of this factor. Estimates of signature rates of marrying couples at the regional
level are available for 281 of the 603 observations in the data set. With these data, it was
possible to estimate the following regression:
Literacy = 41.14 + 11.96*1800 – 7.50*Catholic + 20.61*Dissent
(1.400) (2.985)
(3.414)
(2.050)
+ 0.0086*DistRome - 0.0370*DistMainz + 18.57*Male – 4.41*Rural,
(0.00073)
(0.00262)
(0.899)
(1.233)
R2=0.842, Root MSE = 10.20,
(4)
where Dissent indicates a religion other than Catholic, Lutheran or Anglican, DistRome is
the distance from Rome in kilometers, DistMainz is the distance from Mainz, and Male
and Rural indicate that the measure applies to males or rural residents respectively. As
Cipolla (1969, 72-73) observed, other things being equal Catholics tended to have lower
literacy rates and Dissenting religions higher rates, although German Catholics were
highly literate.5 The 322 missing observations were again imputed by the stochasticregression method.
Allen (2009, 109-111) also suggested that openness to international trade made an
important contribution to Britain’s economic success. This consideration was assumed to
be captured by a dummy variable indicating whether or not the region’s principal city
was an Ocean port in 1700. By the end of the period being studied, inland cities such as
Manchester and Birmingham had been linked to the sea by canals. However, since these
changes were in part endogenous, depending on the region’s previous innovating success,
river cities are excluded as ports.
The next step was to integrate the language-network model of equation (1) into
the specification of equation (2). Three additional variables were inserted. The first was
the log of the population of the urban region’s main center, City population. Next was the
5
Dittmar (2011, 1139) used the distance from Mainz to explain sixteenth-century European urban
population growth, explaining that this measure was positively correlated with literacy.
15
log of the population of the rest of the society, Country population. For most cities in the
sample, this variable was assumed to be captured by the population within the presentday country (less the city’s population) at the beginning of each period. The United States
was an exception. At the beginning of the first two sub-periods, the thirteen colonies were
a part of the British Empire. Even after the American War of Independence, the two
countries remained important trading partners. At the end of the eighteenth century, over
half of British exports and a third of its imports were with the Americas, including a
small portion with British North America (Deane and Cole 1962, 62). The British Isles
also remained the most important source of immigration to the new nation (Bankston,
2010). Accordingly Great Britain and the United States were assumed to form a single
society.
For the third networking variable – the distance of the city in question from the
rest of the country – two different measures were used. One factor that would reduce the
willingness to cooperate of two individuals is simply physical separation, Distance,
defined as the number of kilometers of the city in question from the country’s main
cultural center. Willingness to cooperate would also depend on the degree of
standardization of linguistic signals. For example, if Watt had been raised in the Scottish
Highlands, it is unlikely that he would have been able to speak and write Standard
English in 1764. However, growing up in the Lowlands, he had been exposed to the
Anglicization of the Scottish school system that had begun with the Schools Act of 1696.
This legislation had provided for a school in every Scottish parish that would teach
reading and writing in English (Herman, 2001 p. 22). Accordingly, an alternative
measure of distance is the degree of language standardization, as measured by Dictionary
year, the year of publication of the country’s first privately-published monolingual
dictionary (normalized with Britain in 1658 equal to zero), as shown in Table 2.
One possible explanatory variable not included is the prior existence of a
university in the region – a factor that Bosker et al. (2013) found helped explain
European urban growth before 1800. In this case, however, causality appears to have
been reversed. Despite having universities only from the third decade of the nineteenth
16
century, Birmingham and Manchester, two of the fastest-growing cities in the sample,
contributed 30 percent of all the innovations between 1700 and 1850!
17
4. Results
Does taking account of language standardization improve the explanatory power
of the supply and demand approaches used in previous studies of innovation? We shall
consider separately each of two sub-groups of the total sample; namely, the 54
Cooperative Innovations (CIs) and the 63 Non-Cooperative Innovations (NCIs).
Table 4 presents the results for the CIs. In column (1) is a reduced-form
specification that captures the demand and supply approaches. One finding is that the
country fixed effects for Britain and France are arithmetically much greater than those for
the other north-European countries. A possible explanation is that this gap reflects
important supply-side effects of institutions or culture. In addition, the significant
positive coefficient for 1750 is consistent with Mokyr’s (2009, 30) argument that the
Enlightenment promoted a dissemination of knowledge across Europe and North
America, allowing scientific discoveries to be applied by those with practical problems to
solve. As for the demand side, we see that the Relative energy price and the Literacy rate
are also significantly different from zero with the expected signs.
Column (2) adds two networking variables, Country population and Distance
from the principal cultural center. We find that the coefficients of these variables are both
highly significant with the expected signs. This result is compatible with the hypothesis
that there were significant network effects and that these effects diminished with
distance. The other results remain essentially unchanged.
Column (3) presents an alternative explanation of network effects, based on the
degree of language standardization rather than physical distance. Inclusion of Dictionary
year significantly improves the fit of the model. However, the factor-price and humancapital variables are no longer significant. Column (4) shows that these results were
insensitive to the inclusion of Distance.
The drop in statistical significance of the factor-price and human-capital variables
between columns (2) and (3) has potential implications for our understanding of the
process of innovation, helping to explain the uneven pattern of industrialization across
Europe between 1700 and 1850. Despite access to the flow of scientific knowledge across
18
national borders, despite rich coal deposits and well-educated populations, many regions
– even in Britain and France – failed to experience rapid growth before the middle
decades of the nineteenth century. A possible interpretation is that in England, France and
the USA, the creation of large networks of people speaking standard languages set off a
series of social changes in regions close to the main cultural center: encouraging the
development of energy resources, raising the real wage rate, and stimulating innovation.
Elsewhere, these changes were delayed.
Table 5 presents the same specifications for the NCIs. At first glance, the results
appear similar to those of the CIs. Looking at the most general specification in column
(4), for example, we find that city population along with the country fixed effects for all
countries but Britain are significant. In addition, neither the relative energy price nor the
literacy rate has a significant effect on the frequency for this group of innovations. The
Enlightenment variable, 1750, is highly significant, suggesting important knowledge
spillovers across countries during this half-century. However, in results not shown, when
terms capturing interaction between Britain and 1750 or 1800 were added to these
specifications, they were not significant. In other words, there was no evidence that
institutional change in Britain created a peculiarly British “Industrial Enlightenment” that
was more favorable to innovation than its continental counterpart, ceteris paribus.
A closer examination of Table 5 nevertheless reveals important differences with
respect to the preceding table. For one thing, the measure of network size, Country
population, is no longer significant. Indeed, the coefficients of this variable and the other
key networking measure, Dictionary year, are both significantly smaller in absolute value
than in Table 4. This result suggests that inventors working on their own were not as
dependent on communication with colleagues from other regions as cooperators were.
However, since Dictionary year was still significant, independent inventors did seem to
require clear communication. Undoubtedly inventors working alone depended on the
cooperation of suppliers, employees and clients with respect to contractual details too
costly to specify explicitly.
19
In short, priority in the formation of standardized language networks was highly
significant in explaining the innovation performance of Britain, its offshoots and northern
France. However, there were significant fixed effects of unobserved variable across
countries that are compatible with cultural-genetic differences on the supply side.
20
5. Robustness and Fragility
Implicit in the results of Tables 4 and 5 are a number of arbitrary assumptions
concerning the specification and estimation of equation (2). It is therefore important to
examine the sensitivity of these estimates not only to the explanatory variables included
but also to the way in which these variables are defined and to the estimation procedure
used.
The robustness of the above estimates may be assessed by regression of the
dependent variable on different combinations of the explanatory variables, leaving the
fixed effects in each specification. Each of the six principal explanatory variables thus
appears in 32 specifications. Let us consider as robust those variables that not only
always have the expected sign but also are significant at the five percent level with a twotailed test in at least 95 percent of the runs. The names of these "robust" variables appear
in bold face in Tables 6 and 7 for Cooperative Innovations and Non-Cooperative
Innovations respectively.
Consider first the networking variables. City population and Dictionary year are
robust for both types of innovations. Country population, the measure of network size, is
more fragile. It is always positive for CIs but is negative in two cases for NCIs. In the
assessment of the preceding results, an important question is whether or not the
networking measures are simply capturing conventional supply-and-demand effects. Let
us compare the coefficients for the CIs with those of the NCIs. In Tables 6 and 7, we find
that the mean coefficients of two key networking variables – Country population and
Dictionary year – are greater in absolute value for CIs than for NCIs. Furthermore, a
comparison of the corresponding mean standard errors suggests that in most cases these
differences are significant. In other words, in this sample, networking is more important
when two principals had to combine their expertise than when a single inventor was
working independently. This evidence suggests that it was communications networks
rather than factor prices or institutions that were driving the significant results for the
networking variables.
21
Another major concern in interpreting these results is whether the dates of the first
monolingual dictionaries are actually capturing the degree of language standardization. It
might be argued that for some countries, the dates of the first dictionaries are quite
similar to the years in which new scientific institutions such as the Royal Society (1660)
in Britain, the Académie des sciences (1666) in France and the Dutch Academy of Arts
and Sciences (1808) in the Netherlands were founded. Therefore Dictionary year might
actually be picking up the penetration of scientific knowledge across Western Europe.
However, there are several reasons to hesitate before accepting such an interpretation. (i)
Although the Prussian Akademie der Wissenschaften was founded in 1700, there were no
Prussian and indeed few German innovations before 1850. (ii) As Allen (2009, 252)
pointed out, in metals and textiles, the most important industrial sectors of the eighteenth
and early nineteenth centuries, the connection between innovation and science was
tenuous. (iii) There is no reason to expect that increased scientific training would benefit
collaborative innovators more than independent ones. However, as observed above, the
coefficient of Dictionary year was generally significantly greater in absolute value for
CIs than for NCIs.
A final question is the sensitivity of the results in Tables 4 and 5 to several
arbitrary assumptions used in the estimation procedure. At issue, for example, is the
negative binomial specification with correction for clusters within countries. Another
issue is the arbitrary date of 1728 for the standardization of American English. The
specifications of column (4) of Tables 4 and 5 are repeated in column (1) of Tables 8 and
9. Column (2) of the latter tables presents a Poisson distribution. Column (3) replaces the
country-cluster correction of standard errors with a simple robust correction. Column (4)
uses a period-cluster correction. Finally, Column (5) assumes that the standardization of
American English occurred in 1755 rather than 1728.6 Comparing these results with the
original estimates in column (1), we see that the principal conclusions of the preceding
analysis go through unchanged, namely, the highly significant network effects for
cooperative innovations, the lesser importance of such effects for non-cooperative
6
Delaying standardization in Scotland by 30 years and advancing Austrian standardization to the German
date left the conclusions similarly unchanged.
22
innovations and the significant differences between the country fixed effects of Britain
and France on the one hand and the remaining countries of northern Europe on the other..
These findings may explain why for a century and a half after 1700, industrial
innovation was concentrated within a few regions in Britain, northern France and the
north-eastern United States. On the one hand, the Germanic countries, Scandinavia and
the Low Countries had high literacy rates, but their non-standardized vernaculars and
small populations prevented large numbers of people with a wide range of skills from
collaborating easily with one another. On the other hand, in southern Europe, not only did
each region have its own dialect, but also literacy rates were low. Even in Britain and
France, regions distant from the main cultural centers were at a considerable
disadvantage when it came to generating innovation.
23
6. Conclusion
One possible lesson to be drawn from this study is that new institutions, cheaper
energy and rising literacy alone cannot account for the jump in the rate of innovation
observed in the West after 1700. The evidence presented here suggests that the creation
of novelty depended also on the diffusion of standardized languages which permitted the
formation of broad networks of trust within which strangers could collaborate easily. In
societies where such language network formation was delayed, there was little innovation
before 1850 even if literacy rates were high.
A second tentative conclusion is that these new language networks were
especially important for the cooperative innovations that made up almost half of the
sample. Compared to the technologies with a single inventor, these cooperative
innovations tended to be technologically relatively complex. Mutual trust between two or
more individuals would therefore have been required to span the skills that were
integrated in developing these technologies. For the generally simpler non-cooperative
innovations, network effects were statistically significant but quantitatively less
important.
A final regularity was the selective spread of the capacity to innovate within
limited territorial bands in Britain, France and the United States over the century and a
half after 1700. The evidence suggested that the rate of diffusion of standard languages
was an important factor in this geographic concentration. To have learned to speak and
write in the standard national language was to be able to signal to strangers that one was
reliable as a potential partner even when one’s reputation was unknown. There
nevertheless remained significant international variations in the propensity to innovate
that are compatible with the presence of underlying cultural or genetic differences.
24
References
Acemoglu, Daron, et James A. Robinson. Why Nations Fail: The Origins of Power,
Prosperity and Poverty. London, UK: Profile Books, 2012.
Allen, Robert C. "Collective Invention." Journal of Economic Behavior and
Organization 4 (1983): 1-24.
—. The British Industrial Revolution in Global Perspective. Cambridge, UK: Cambridge
University Press, 2009.
Allen, Robert C. "The Great Divergence in European Wages and Prices from the Middle
Ages to the First World War." Explorations in Economic History 38 (2001): 411–
447.
Ashton, Thomas S. The Industrial Revolution, 1760-1830. London: Oxford University
Press, 1948/1962.
Bain, Joseph. Calendar of the State Papers Related to Scotland and Mary, Queen of
Scots, 1547-1603. Edinburgh, Scotland: H. M. General Register House, 1898.
Bairoch, Paul, Jean Batou, et Pierre Chèvre. The Population of European Cities, 8001850. Geneva: Librairie Droz, 1988.
Bankston, Carl L. Encyclopedia of American Immigration. Hackensack, NJ: Salem Press,
2010.
Barlow, Julie, et Jean-Benoît Nadeau. La grande aventure de la langue française.
Montréal, Canada: Québec Amérique, 2007.
Barraclough, Geoffrey. The Times Atlas of World History. Maplewood, NJ, USA:
Hammond, 1984.
Baten, Joerg, et Jan Luiten van Zanden. «Book Production and the Onset of Modern
Economic Growth.» Journal of Economic Growth 13(3) (2008): 217-235.
Bell, David A. "Culture and Religion." In Old Regime France 1648-1788, by William
Doyle, 78-104. Oxford, UK: Oxford University Press, 2001.
Bosker, Maarten, Eltjo Buringh, et Jan Luiten Van Zanden. «From Baghdad to London:
Unravelling Urban Development in Europe and the Arab World 800-1800.» Review
of Economics and Statistics 95(4) (2013): 1418-1437.
Bragg, Melvyn. The Adventure of English. London: Hodder and Stoughton, 2003.
Bray, Laurent. César-Pierre Michelet (1626-1698). Tübingen, Germany: Max Niemeyer,
1986.
Briscoe, Bob, Andrew Odlyzko, et Benjamin Tilly. «Metcalfe's Law is Wrong.» IEEE
Spectrum. 2006. http://spectrum.ieee.org/computing/networks/metcalfes-law-iswrong (accès le January 15, 2015).
Burnham, Robert. «Statistical Abstract of the Napoleonic Wars: Population Statistics.»
The
Napoleon
Series.
2005.
http://www.napoleonseries.org/research/abstract/population/c_population.html (accès le January 7, 2015).
25
Cardwell, D. S. L. Turning Points in Western Technology. Canton, MA: Science History
Publications/USA, 1991.
Cipolla, Carlo M. Literacy and Development in the West. Harmondsworth, UK: Penguin,
1969.
Clark, Gregory. A Farewell to Alms: A Brief Economic History of the World. Princeton,
NJ: Princeton University Press, 2007.
Cohen, Murray. Sensible Words: Linguistic Practice in England, 1640-1785. Baltimore,
USA: Johns Hopkins University Press, 1977.
Crafts, Nicholas. "Explaining the First Industrial Revolution: Two Views." European
Review of Economic History 15 (2010): 153-168.
Cressy, David. Literacy and the Social Order: Reading and Writing in Tudor and Stuart
England. Cambridge, UK: Cambridge University Press, 1980.
Crystal, David. The Cambridge Encyclopedia of the English Language. Cambridge, UK:
Cambridge University Press, 2003.
Daumas, Maurice. A History of Technology and Invention, Volume III, The Expansion of
Mechanization, 1725-1860. New York: Crown Publishers, 1979.
Deane, Phyllis, and W. A. Cole. British Economic Growth, 1688-1959 : Trends and
Structure. Cambridge, UK: Cambridge University Press, 1962.
Dittmar, Jeremiah E. «Information Technology and Economic Change: The Impact of the
Printing Press.» Quarterly Journal of Economics 126 (2011): 1133-1172.
Farrell, Joseph, and Matthew Rabin. "Cheap Talk." Journal of Economic Perspectives 10,
no. 3 (1996): 103-118.
Frank, Thomas. "Language Standardization in Eighteenth-century Scotland." In Towards
a Standard English 1600-1800, by Dieter Stein and Ingrid Tieken-Boon van Ostade,
51-62. Berlin: Mouton de Gruyter, 1994.
Freeborn, Dennis. From Old English to Standard English. Basingstoke, UK: Macmillan,
1992.
Galor, Oded, Omer Moav, et Dietrich Vollrath. «Inequality in Landownership, the
Emergence of Human-Capital Promoting Institutions, and the Great Divergence.»
Review of Economic Studies 76(1) (2009): 143-179.
Gildea, Robert. "Province and Nation." Chap. 6 in Revolutionary France 1788-1880, by
Malcolm Crook. Oxford, UK: Oxford University Press, 2002.
Gooskens, Charlotte, Sebastian Kürschner, et Renée van Bezooijen. «Intelligibility of
Standard German and Low German to Speakers of Dutch.» Dialectologia, 2011: 3563.
Gooskens, Charlotte, Wilbert Heeringa, et Karin Beijering. «Phonetic and Lexical
Predictors of Intelligibility.» International Journal of Humanities and Arts
Computing 2(1-2) (2008): 63-81.
Graff, Harvey J. The Legacies of Literacy. Bloomington, Indiana, USA: Indiana
University Press, 1991.
26
Green, Jonathon. Chasing the Sun: Dictionary Makers and the Dictionaries They Made.
New York: Henry Holt and Co. , 1996.
Greif, Avner. Institutions and the Path to the Modern Economy. Cambridge: Cambridge
University Press, 2006.
Harris, Martin. "French." In The World's Major Languages, by Bernard Comrie, 210-235.
Oxford, UK: Oxford University Press, 1990.
Herman, Arthur. How the Scots Invented the Modern World. New York: Three Rivers
Press, 2001.
Hills, Richard L. Power in the industrial revolution. Manchester, UK: Manchester
University Press, 1970.
Hofmeister, Andrea, Reiner Prass, et Norbert Winnige. «Elementary Education, Schools
and the Demands of Everyday Life: North West Germany in 1800.» Central
European History 31 (1998): 329-384.
Houdaille, Jacques. «Les signatures au mariage de 1740 à 1829.» Population 32(1)
(1977): 65-90.
Jackson, Howard. Lexicography: An Introduction. London: Routledge, 2002.
Jacob, Margaret C. Scientific Culture and the Making of the Industrial West. Oxford, UK:
Oxford University Press, 1997.
Landes, David S. The Wealth and Poverty of Nations. New York: W. W. Norton, 1998.
Maddison, Angus. Contours of the World Economy, 1-2030 AD: Essays in MacroEconomic History. Oxford, UK: Oxford University Press, 2007.
—. The World Economy: A Millennial Perspective. Paris: OECD, 2001.
Mitch, David. «The Role of Education and Skill in the British Industrial Revolution.»
Dans The British Industrial Revolution: An Economic Perspective, Second edition,
de Joel Mokyr, 241-279. Boulder, CO: Westview Press, 1999.
Mokyr, Joel. The Enlightened Economy: An Economic History of Britain 1700-1850.
New Haven, USA: Yale University Press, 2009.
—. The Gifts of Athena : Historical Origins of the Knowledge Economy. Princeton, NJ:
Princeton University Press, 2002.
—. The Lever of Riches: Technological Creativity and Economic Progress. Oxford, UK:
Oxford University Press, 1990.
Morgan, T. J. H., N.T. Uomini, et L.E. Rendell. «Experimental Evidence for the Coevolution of Hominin Tool-making Teaching and Language.» Nature
Communications, January 2015.
North, Douglass C.l. Institutions, Institutional Change, and Economic Performance.
Cambridge, UK: Cambridge University Press, 1990.
Nuvolari, Alessandro. "Collective Invention during the British Industrial Revolution: The
Case of the Cornish Pumping Engine." Cambridge Journal of Economics 28, no. 3
(2004): 347-363.
27
Paulinyi, Akos. Industrielle Revolution: Vom Ursprung der modernen Technik. Reinbek
bei Hamburg, Germany: Rowohlt, 1989.
Reis, Jaime. "Economic Growth, Human Capital Formation, and Consumption in
Western Europe before 1800." In Living Standards in the Past : New Perspectives on
Well-being in Asia and Europe, by Robert C. Allen, Tommy Bengtsson and Martin
Dribe, 195-225. Oxford: Oxford University Press, 2005.
Smith, Eric Alden. "Communication and Collective Action: Language and the Evolution
of Human Cooperation." 31 (2010): 231-245.
Spence, Michael. "Job Market Signaling." Quarterly Journal of Economics 87, no. 3
(1973).
Stone, Lawrence. "Literacy and Education in England 1640-1900." Past & Present 42
(1969): 69-139.
Sunderland, David. Social Capital, Trust and the Industrial Revolution, 1780-1880.
Abingdon, Oxon, UK: Routledge, 2007.
van der Woude, A. M. «De alfabetisering.» Algemene Geschiednis der Nederlanden 7
(1980): 257-264.
28
Appendix – Data Sources
City population. Population estimates for European cities were from Bairoch et al.
(1988). Estimates for New York, Philadelphia and Boston were from Longman
Publishing http://wps.ablongman.com/wps/media/objects/244/250688/Appendix/12.pdf.
There were 46 cities at or near which one or more innovations occurred. To avoid
potential endogeneity, population and literacy were determined at the beginning of each
of the three half-century periods.
Country population. The source was Maddison (2007).
Coal. The identification of cities with coal deposits within 30 miles (50 km) was obtained
from Barraclough (1984: 201, 210-211).
Distance. The driving distance in kilometers of each city from the country’s major
cultural center was obtained from Google Maps.
Dictionary year. See Table 2. Note that for five countries, the dates of the first dictionary
were chosen arbitrarily. Cities in Belgium and Switzerland were assigned the dates of
French, Dutch or German dictionaries, depending on their main languages. For Austria,
the choice was the mean of Germany and the first Austrian dictionary, i.e. 1868. As for
Scotland, by the year of Union with England, 1707, educated Scots were growing
accustomed to using English rather than the Scottish dialect for formal communication
(Herman 2001, 116). In the case of the United States, the first settlers spoke the language
of their home regions. However, by the year 1725, the more popular English dictionaries
were beginning to be imported (Green, 1999, 285).
Literacy. Signature rates at marriage are not necessarily an accurate measure of people’s
ability to read and write (Mitch, 1999, 244). However, unlike the national publication
rates used by Baten and Luiten van Zanden (2008), differences in signature rates provide
an indication of changes in human capital at the regional level. By country, the sources
were England: Cressy (1980, 177); Scotland: Stone (1969, 121); France: Houdaille (1977,
68); Germany: Hofmeister et al. (1998); Italy: Reis (2005, 202); Netherlands: van der
Woude (1980, 257-264); United States: Graff (1991, 249).
29
Relative energy price. Wage rates for eight European cities in the sample were from
Allen (2001, Table 2). Prices of coal for the same cities were available in Allen (2009,
99-100). To the extent that factors were mobile within each country, these data provide
sufficient information to obtain reasonable estimates of relative factor prices by region.
30
Figure 1. Areas of high innovation density, 1700-1849
31
All innovations
Distance from cultural center (km)
800
Means
ParisMontpellier
600
400
200
LondonBirmingham
0
1700
1750
1800
1850
1800
1850
Year
Cooperative innovations
Distance from cultural center (km)
800
Means
600
400
200
0
1700
1750
Year
Figure 2. Year of innovation and distance from cultural center
32
Table 1. 117 important innovations, 1700-1849
Country
Denmark
1700-1749
1750-1799
France
Loom coded with
perforated paper
(Bouchon, 1725;Lyon)
Loom coded with punched
cards (Falcon, 728;Lyon)
Steam-powered wagon (Cugnot,
1770; Paris)
Automatic loom (Vaucanson, 1775;
Paris)
Single-action press (Didot,
Proudon,1781; Paris)
Two-engine steamboat (Jouffroy
d'Abbans, 1783; Lyon)
Hot-air balloon (Montgolfier, 1783;
Paris)
Parachute (Lenormand, 1783;
Montpellier)
Press for the blind (Haüy, 1784;
Paris)
Chlorine as bleaching agent
(Berthollet, 1785; Paris)
Sodium carbonate from salt
(Leblanc, d’Arcet, 1790; Paris)
Visual telegraph (Chappe, 1793;
Paris)
Vacuum sealing (Appert, 1795;
Paris)
Paper-making machine (Robert,
Didot, 1798; Paris)
Illuminating gas from wood
(Lebon, 1799; Paris)
Germany
Porcelain (Tschirnhaus,
1707; Dresden)
Lithography (Senefelder, 1796;
Munich)
Great
Britain
Seed drill (Tull, 1701;
Oxford)
Iron smelting with coke
(Darby, Thomas, 1709;
Birmingham)
Atmospheric engine (Newcomen, Calley,1712;
Birmingham)
Pottery made with flint
(Astbury, 1720;
Birmingham)
Quadrant (Hadley, 1731;
London)
Flying shuttle (Kay, 1733;
Manchester)
Glass-chamber process for
sulfuric acid (Ward, White,
D’Osterman, 1736;
London)
Spinning machine with
Crucible steel (Huntsman, 1750;
York)
Rib knitting attachment (Strutt,
Roper, 1755; Birmingham)
Achromatic refracting telescope
(Dollond, 1757; London)
Breast wheel (Smeaton, 1759;
York)
Bimetallic strip chronometer
(Harrison, 1760; London)
Spinning jenny (Hargreaves, 1764;
Manchester)
Creamware pottery (Wedgewood,
Wieldon, 1765; Birmingham)
Cast-iron railroad (Reynolds, 1768;
Birmingham)
Engine using expansive steam
operation (Watt, Roebuck, 1769;
Glasgow)
Water frame (Arkwright, Kay,
1800-1849
Galvanometer (Oersted, 1819;
Copenhagen)
Automatic loom with perforated cards (Jacquard, Breton,
1805; Lyon)
Wet spinning for flax (de
Girard, 1815; Avignon)
Electromagnet (Arago, Ampère,
1820; Paris)
Water turbine (Burdin, 1824;
Saint-Étienne)
Single-helix propeller (Sauvage,
1832; Le Havre)
Three-color textile printing
machine (Perrot, 1832;
Rouen)
Water turbine with adjustable
vanes (Fourneyron, 1837;
Besançon)
Photography (Daguerre, Niepce,
1838; Paris)
Multiple-phase combing
machine (Heilmann, 1845;
Mulhouse)
Machines for tackle block
production (Brunel,
Maudslay, 1800; London)
Illuminating gas from coal
(Murdock, Watt Jr., 1802;
Birmingham)
Steam locomotive (Trevithick,
Homfray, 1804; Plymouth)
Compound steam engine
(Woolf, Edwards, 1805;
London)
Winding mechanism for loom
(Radcliffe, 1805; Manchester)
Arc lamp (Davy, 1808; London)
Food canning (Durand, Girard,
1810; London)
Rack locomotive (Blenkinsop,
Murray, 1811; Bradford)
Mechanical printing press
(Koenig, Bauer, 1813;
33
Country
1700-1749
rollers (Wyatt, Paul, 1738;
Birmingham)
Stereotyping (Ged, 1739;
Edingurgh)
Lead-chamber process for
sulfuric acid (Roebuck,
1746; Birmingham)
1750-1799
1769; Birmingham)
Efficient atmospheric steam engine
(Smeaton, 1772; Newcastle)
Dividing machine (Ramsden, 1773;
London)
Cylinder boring machine
(Wilkinson, 1775; Birmingham)
Carding machine (Arkwright, Kay,
1775; Birmingham)
Condensing chamber for steam
engine (Watt, Boulton, 1776;
Birmingham)
Steam jacket for steam engine
(Watt, Boulton, 1776;
Birmingham)
Spinning mule (Crompton, 1779;
Manchester)
Reciprocating compound steam
engine (Hornblower, 1781;
Plymouth)
Sun and planet gear (Watt, Boulton,
1781; Birmingham)
Indicator of steam engine power
(Watt, Southern, 1782;
Birmingham)
Rolling mill (Cort, Jellicoe, 1783;
London)
Cylinder printing press for calicoes
(Bell, 1783; Glasgow)
Jointed levers for parallel motion
(Watt, Boulton, 1784;
Birmingham)
Puddling (Cort, Jellicoe, 1784;
London)
Power loom (Cartwright, 1785;
York)
Speed governor (Watt, Boulton,
1787; Birmingham)
Double-acting steam engine (Watt,
Boulton, 1787; Birmingham)
Threshing machine (Meikle, 1788;
Edinburgh)
Single-phase combing machine
(Cartwright, 1789; York)
Machines for lock production
(Bramah, Maudslay, 1790;
London)
Single-action metal printing press
(Stanhope, Walker, 1795;
London)
Hydraulic press (Bramah,
Maudslay, 1796; London)
High-pressure steam engine
(Trevithick, Murdoch, 1797;
1800-1849
London)
Steam locomotive on flanged
rails (Stephenson, Wood,
1814; Newcastle)
Safety lamp (Davy, 1816;
London)
Circular knitting machine (M. I.
Brunel, 1816; London)
Planing machine (Roberts,
1817; Manchester)
Large metal lathe (Roberts,
1817; Manchester)
Gas meter (Clegg, Malam,
1819; London)
Metal power loom (Roberts,
Sharp, 1822; Manchester)
Rubber fabric (Hancock,
Macintosh, 1823; London)
Locomotive with fire-tube
boiler (Stephenson, Booth,
1829; Manchester)
Hot blast furnace (Nielson,
Macintosh, 1829; Glasgow)
Self-acting mule (Roberts,
Sharp, 1830; Manchester)
Lathe with automatic cross-feed
tool (Whitworth, 1835;
Manchester; Manchester)
Planing machine with pivoting
tool-rest (Whitworth, 1835;
Manchester)
Even-current electric cell
(Daniell, 1836: London)
Electric telegraph (Cooke,
Wheatstone, 1837; London)
Riveting machine (Fairbairn,
Smith, 1838; Manchester)
Transatlantic steamer (I. K.
Brunel, Guppy, 1838; Bristol)
Assembly-line production
(Bodmer, 1839; Manchester)
Multiple-blade propeller (Smith,
Currie, 1839; London)
Steam hammer (Nasmyth, 1842;
Manchester)
Iron, propellor-driven steamship
(Brunel, Guppy, 1844;
Bristol)
Measuring machine (Whitworth, 1845; Manchester)
Multiple-spindle drilling
machine (Roberts, 1847;
Manchester)
34
Country
1700-1749
1750-1799
1800-1849
Plymouth)
Slide lathe (Maudslay, 1799;
London)
Italy
Electric battery (Volta, 1800;
Como)
Switzerland
Massive platen printing press
(Haas, 1772; Basel)
Stirring process for glass (Guinand,
1796; Berne)
United
States
Continuous-flow production
(Evans, Ellicott, 1784;
Philadelphia)
Cotton gin (Whitney, Green, 1793;
Philadelphia)
Machine to cut and head nails
(Perkins, 1795; Boston)
Sources: see Section 1 of text.
Single-engine steamboat
(Fulton, Livingston, 1807)
Milling machine (North, 1818;
New York)
Interchangeable parts (North,
Hall, 1824; New York)
Ring spinning machine (Thorp,
1828; Boston)
Grain reaper (McCormick,
Anderson, 1832; Philadelphia)
Binary-code telegraph (Morse,
Vail, 1845; New York)
Sewing machine (Howe, Fisher,
1846; Boston)
Rotary printing press (Hoe,
1847; New York)
35
Table 2. Year of first monolingual dictionary
Country
Austria
England
Belgium (French)
Belgium (Flem.)
Denmark
France (north)
France (south)
Germany
Year
1868
1658
1680
1864
1833
1680
1815
1786
Ireland
1958
Italy
1897
Netherlands
1864
Scotland
Switzerland (Fr.)
Switzerland
(German)
United States
1707
1680
1786
1728
Author(s)
Otto Back et. al.
Edward Phillips
Christian Molbech
Pierre Richelet
Johann Christoph
Adelung
Ireland, Parliament,
Translation Bureau
Emilio Broglio &
Giovan Battista Giorgini
Marcus and Nathan
Solomon Calisch
Nathan Bailey
Publication
Österreichisches Wörterbuch (1951)
The New World of English Words
Same as France (north)
Same as Netherlands
Dansk Ordbog
Dictionnaire français
Standardization delayeda
Grammatich-kritisches Wörterbuch
der hochdeutschen Mundart
Gramadach na Gaeilge agus Litriú na
Gaeilge: An Caighdeán Oifigiúil
Nòvo vocabolario della lingua
italiana secondo l'uso di Firenze
Nieuw Woordenboek der
Nederlandsche Taal
Year of Union with England
Same as France (north)
Same as Germany
An Universal Etymological English
Dictionary
a
South of a line from St. Malo to Geneva, standardization occurred through the integrating effects
of the revolutionary and Napoleonic Wars (Graff, 1991, 193).
36
Table 4. Negative binomial regressions for cooperative innovations
Group
Fixed effects
Variable
Britain
France
Germany
Belgium
Netherlands
1750
1800
Demand
Relative energy price
Literacy
Ocean port
Language networks
City population
(1)
(2)
(3)
(4)
0.866
(0.387)
-0.673
(1.123)
-17.160**
(1.176)
-16.835**
(1.176)
-18.012**
(1.113)
1.061**
(0.412)
0.034
(0.235)
-0.003
(0.351)
-2.076*
(0.867)
-17.098**
(1.141)
-13.808**
(1.699)
-14.912**
(1.605)
0.338
(0.394)
-1.462
(0.787)
-3.982**
(0.591)
-5.915**
(1.866)
-19.888**
(1.231)
-17.733**
(1.660)
-14.604**
(1.812)
0.161
(0.642)
-1.689
(1.227)
-4.303**
(0.699)
-6.258**
(1.792)
-18.330**
(1.231)
-16.021**
(1.539)
-12.539**
(2.027)
0.098
(0.716)
-1.819
(1.379)
-1.997*
(0.910)
3.339**
(0.571)
-0.327
(0.795)
-1.918**
(0.718)
3.454**
(0.885)
-0.640
(0.395)
-1.392
(0.779)
2.077
(2.513)
-1.039*
(0.449)
-1.318
(0.759)
2.105
(2.537)
-1.081*
(0.453)
1.117*
(0.521)
0.858*
(0.375)
1.792**
(0.547)
-0.204**
(0.065)
1.127**
(0.434)
1.926**
(0.577)
1.236**
(0.408)
Country population
1.992**
(0.664)
Distance
0.086
(0.074)
Dictionary year
-3.251**
-3.462**
(0.270)
0.368
Constant
1.477
2.010
6.783**
7.056**
(1.545)
(1.257)
(2.288)
(2.225)
2.590**
1.868**
1.356**
1.319*

(0.715)
(0.372)
(0.486)
(0.542)
Log pseudolikelihood
-80.70
-76.39
-71.76
-71.66
Time-series cross-section of 201 cities for 1700-1749, 1750-1799 and 1800-1849.
Dependent variable: number of cooperative innovations in region of city j in period t.
Number of observations: 603.
Standard errors adjusted for five clusters in country are shown in parentheses.
* Coefficient significantly different from zero at 0.05 level, two-tailed test.
** Coefficient significantly different from zero at 0.01 level, two-tailed test.
Coefficients in bold face are significantly different from corresponding estimates in Table
6 at 0.05 level, two-tailed test.
37
Table 5. Negative binomial regressions for non-cooperative innovations
Group
Fixed effects
Variable
Britain
France
Germany
Belgium
Netherlands
1750
1800
Demand
Relative energy price
Literacy
Ocean port
Language networks
City population
(1)
(2)
(3)
(4)
0.545
(0.725)
-0.135
(0.353)
-2.073**
(0.304)
-15.671**
(1.124)
-16.436**
(1.123)
0.722**
(0.227)
-0.058
(0.441)
0.437
(0.802)
-0.261
(0.250)
-2.181**
(0.315)
-15.680**
(1.164)
-16.453**
(1.143)
0.692**
(0.162)
-0.400
(0.574)
-1.045
(0.988)
-1.551**
(0.527)
-2.395**
(0.549)
-15.730**
(1.219)
-15.535**
(1.128)
0.707**
(0.188)
-0.173
(0.836)
-1.127
(0.935)
-1.625**
(0.549)
-2.459**
(0.434)
-15.855**
(1.310)
-15.833**
(1.131)
0.709**
(0.194)
-0.189
(0.831)
-1.092
(1.174)
1.616
(1.833)
-0.723
(0.712)
-1.544
(0.875)
1.585
(1.629)
-0.920
(0.642)
-1.384
(0.822)
-0.044
(1.285)
-0.844
(0.643)
-1.357
(0.849)
0.051
(1.176)
-0.831
(0.640)
1.017**
(0.150)
1.026**
(0.304)
0.087
(0.309)
-0.001
(0.141)
0.991**
(0.100)
0.122
(0.241)
1.043**
(0.248)
Country population
0.094
(0.324)
Distance
0.051
(0.150)
Dictionary year
-1.001**
-1.042**
(0.216)
(0.290)
Constant
1.803
2.607
4.521
4.494*
(2.205)
(1.766)
(2.108)
(2.114)
2.045
1.827
1.431
1.426

(1.077)
(1.056)
(0.825)
(0.816)
Log pseudolikelihood
-129.59
-127.55
-125.52
-125.44
Time-series cross-section of 201 cities for 1700-1749, 1750-1799 and 1800-1849.
Dependent variable: number of non-cooperative innovations in region of city j in period t.
Number of observations: 603.
Standard errors adjusted for five clusters in country are shown in parentheses.
* Coefficient significantly different from zero at 0.05 level, two-tailed test.
** Coefficient significantly different from zero at 0.01 level, two-tailed test.
38
Group
Fixed effects
Table 6. Robustness check for cooperative innovations, 1700-1850
Variable
Mean AvgSTD
%Sig
%+
%AvgT
Britain
France
Germany
Belgium
Netherlands
1750
1800
Demand
Relative energy price
Literacy
Ocean port
Language networks
City population
Country population
Dictionary year
Obs
-0.89
-3.08
-17.68
-16.79
-15.61
0.99
0.13
0.49
1.04
1.18
1.67
1.64
0.61
1.00
77
64
100
100
100
47
14
47
3
0
0
0
100
61
53
97
100
100
100
0
39
5.48
4.74
15.04
10.95
10.48
1.88
1.18
64
64
64
64
64
64
64
-1.58
4.22
-0.09
1.67
2.44
0.68
25
50
13
0
100
53
100
0
47
1.30
2.73
0.88
32
32
32
1.13
1.72
-3.37
0.41
0.79
0.51
97
69
100
100
100
0
0
0
100
2.88
2.28
14.60
32
32
32
Table 7. Robustness check for non-cooperative innovations, 1700-1850
Group
Variable
Mean AvgSTD
%Sig
%+
%AvgT
Fixed effects
Britain
0.28
0.49
31
50
50
4.13
France
-0.89
0.41
59
16
84
3.08
Germany
-2.04
0.18
100
0
100
16.63
Belgium
-16.57
1.14
100
0
100
14.63
Netherlands
-16.45
1.15
100
0
100
14.43
1750
0.98
0.24
100
100
0
4.04
1800
0.38
0.55
17
73
27
0.96
Demand
Relative energy price
-0.99
1.07
6
0
100
1.11
Literacy
1.20
1.42
25
94
6
1.21
Ocean port
-0.24
0.45
6
47
53
0.98
Language networks
City population
0.95
0.16
100
100
0
6.44
Country population
0.10
0.13
28
94
6
1.44
Dictionary year
-1.24
0.25
100
0
100
8.83
Obs
64
64
64
64
64
64
64
32
32
32
32
32
32
39
Group
Table 8. Sensitivity analysis for Cooperative Innovations, 1700-1850
Neg. bin.
Poisson
Negative binomial
Variable
Country Country
Robust
Period
Country
US:1725 US:1725 US:1725
US:1725
US:1755
(1)
(2)
(3)
(4)
(5)
Supply
Britain
France
Germany
Belgium
Netherlands
1750
1800
Demand
Relv. energy price
Literacy
Ocean port
Social networks
City population
Country population
Distance
Dictionary year
Constant

Log pseudolikehd
-4.303**
-6.258**
18.330**
16.021**
12.539**
0.098
-1.819
-1.318
2.105
-1.081*
1.236**
1.992**
0.086
-3.462**
7.056**
1.319*
-71.66
-4.227** -4.303**
-5.960** -6.258**
-21.771** -18.330**
-4.303**
-6.258**
-18.330**
-4.330**
-6.285**
-18.614**
-19.308** -16.021**
-16.021**
-16.284**
-14.483** -12.539**
-12.539**
-12.805**
0.247
-1.840
0.098
-1.819
0.098
-1.819
0.094
-1.827
-1.114
2.706
-1.397
-1.318
2.105
-1.081*
-1.318*
2.105
-1.081
-1.320
2.109
-1.079*
1.236**
1.992**
0.086
-3.462**
7.056**
1.319
-71.66
1.236**
1.992**
0.086
-3.462**
7.056**
1.319
-71.66
1.235**
2.002**
0.085
-3.459**
7.078**
1.320*
-71.68
1.031**
2.051**
0.093
-3.138**
5.723**
-80.01
* Coefficient significantly different from zero at 0.05 level, two-tailed test.
** Coefficient significantly different from zero at 0.01 level, two-tailed test.
40
Group
Table 9. Sensitivity analysis for Non-Cooperative Innovations, 1700-1850
Neg. bin.
Poisson
Negative binomial
Variable
Country
Country
Robust
Period
Country
US:1725
US:1725
US:1725
US:1725
US:1755
(1)
(2)
(3)
(4)
(5)
Supply
Britain
France
Germany
Belgium
Netherlands
1750
1800
Demand
Relv. energy price
Literacy
Ocean port
Social networks
City population
Country population
Distance
Dictionary year
Constant

Log pseudolikehd
-1.127
-1.625**
-2.459**
-15.855**
-15.833**
0.709**
-0.189
- 1.845**
- 2.497**
- 2.852**
-16.983**
-15.819**
0.903**
-0.005
- 1.127
- 1.625
- 2.459**
-15.855**
-15.833**
0.709
- 0.189
-1.127**
-1.625**
-2.459**
-15.855**
-15.833**
0.709**
-0.189
-1.129
-1.627**
-2.464**
-15.860**
-15.839**
0.708**
-0.192
-1.357
0.051
-0.831
-1.196
-1.261
-1.192**
- 1.357**
0.051
-0.831
-1.357**
0.051
-0.831
-1.359
0.058
-0.830
1.043**
0.094
0.051
-1.042**
4.494*
1.426
-125.44
1.116**
0.296
0.082
-1.398**
5.649**
1.043**
0.094
0.051
-1.042*
4.494*
1.426*
-125.44
1.043**
0.094
0.051
-1.042**
4.494
1.426*
-125.44
-132.80
* Coefficient significantly different from zero at 0.05 level, two-tailed test.
** Coefficient significantly different from zero at 0.01 level, two-tailed test.
1.043**
0.096
0.050
-1.039**
4.494*
1.428
-125.45