Paper prepared for the conference on Accounting for the Great Divergence,
The University of Warwick in Venice, Palazzo Pesaro Papafeva, 22-24 May 2014
Estimating the shares of secondary- and tertiary-sector output in the age of
proto-industrialisation:
the case of Japan, 1600-1874
Osamu Saito (Hitotsubashi University)
and
Masanori Takashima (Hitotsubashi University)
This version: 17 May 2014
Very preliminary, PLEASE DO NOT QUOTE
1. Introduction
It has long been recognised that data on industrial goods and services are not readily available for
any country in early modern times. It is against this background that Malanima (2003, 2011) has
suggested a way in which this difficulty is circumvented, i.e. the use of urbanization rate, which has
been gaining support from scholars from European countries1. For Japan too, Bassino et al. (2011)
attempts to utilise this method for the estimation of GDP over the very long run: the underlying
rationale is that much of demand for secondary- and tertiary-sector goods and services came from
urban populations.
The supposition of those researches may hold true as far as ancient and medieval periods are
concerned, but it is likely that it will not for the proto-industrial age. Despite the naming, both
industry and services grew in the countryside during that period. As a result, the relationship
between the share of non-primary sector output and the rate of urbanisation became less
straightforward. In his analysis of proto-industrialisation Saito (1983) showed, based on regional
data for early Meiji Japan, that urbanization was not the sole predictor of population density while
there was a significant correlation between population density and the weight of rural industry.
While the former is consistent with the urbanisation pattern identified by de Vries (1984) for
proto-industrial Europe, the latter finding implies that the density of population can be regarded as a
1
By referring to Malanima’s 2003 article, Álvarez-Nogal and Prados de la Escosura (2007) adopts the
same method for Spain between 1500 and 1850.
1
supply-side factor for proto-industrial activities. The implication of this early-Meiji study for the
estimation of non-primary shares in gross domestic output of the early modern period is, therefore,
that both population density and the urbanisation rate may be used to obtain better estimates of the
share of secondary- or tertiary-sector output as long as the age of proto-industrialisation is
concerned.
Thus, this new methodology will be explored with an updated set of data for the Tokugawa and
Meiji periods. First, an early-Meiji regional panel database will be constructed for the period for
1874, 1890 and 1909 from more diversified data sources at prefectural, rather than regional levels in
order to conduct regression analysis of the regional panel data with the two explanatory variables
plus several dummy variables. Then, the results will enable us to compute the shares of secondaryand tertiary-sector output by applying population from the existing estimates of population density
and the urbanisation rate for the benchmark years before 1850.
2. Methodology, data availability and the proto-industrial context
Let us begin with the discussion of difficulties associated with the use of Malanima’s method in the
Tokugawa Japanese context. The underlying idea of Malanima’s is that the level of urbanisation was
a proxy for the growth of non-agricultural activities. Although this is to be fine-tuned by the
changing non-agricultural share in the labour force in a later stage of estimation, the first step is
explore if there was a correlation between those two variables. With a regression coefficient derived
from time-series analysis of the two variables for modern Italy in the period between Unification and
World War II, he has applied the coefficient to pre-modern urbanisation estimates so as to obtain a
series of the non-agricultural share for the 1300-1861 period (Malanima 2003, p. 281: 2011, p. 181).
According to his latest specification of the formula (Malanima 2011, p. 183),
Yna(t)/Y(t) = 15.371 + 1.82U
(1)
where Yna(t)/Y(t) Yna/Y(t) stands for the share of non-agricultural product and U for the rate of
urbanisation (both in percentages), one percentage-point rise in urbanisation must have led to a 1.82
percentage-point increase in the share of the non-primary sector in the total product.
If applied to the Japanese data for the period between 1600 and 1874, however, this formula
would produce a long-run decline in the non-primary share during the latter half of the Tokugawa
period as the level of urbanisation, having reached a peak in the late sixteenth century, exhibited a
2
downward trend towards the end of the period in question. The phenomenon was brought to light
first by Tom Smith in his seminal paper on pre-modern economic growth, who found that 24 out of
35 castle towns, for which he assembled population data for the period from c.1700 to 1850, lost its
commoner population considerably, seven saw its population neither increased nor decreased
markedly, and only four grew in size (Smith 1973). According to our own estimates of urbanisation
rate, defined as the proportion of people living in settlements with population of 10,000 or over
(reported in table A.2B), the national level of urbanisation increased from 6% in 1600 to 14% in
1650, but since then started declining gradually to 13% in 1750, to 12% in 1850 and then to 10% in
1874. What we know about the period from c.1700 to 1850, however, is that rural industry and
commerce grew during that period. There is evidence that in many cases, while the provincial city
declined, the population in the rest of the province increased, suggesting that much of
non-agricultural income must have been earned by peasants in the form of by-employment (for the
importance of farm family by-employments, see for example Smith 1969). As a result of this change
or “rural-centred growth” as Smith termed it, therefore, the non-agricultural share must have reached
a considerable level towards the end of the Tokugawa period. According to an input-output table
constructed by Shunsaku Nishikawa for a daimyo domain called Chōshū (a region that stood close to
the median according to prefectural-level income estimates for 1874), the non-agricultural share of
income estimated on a value added basis stood at 39% in the 1840s, despite the fact that 75% of the
region’s households were agricultural (Nishikawa 1987).2 This effect of rural-centred growth in the
age of de-urbanisation will be missed out if the estimation of non-agricultural output be made only
on the basis of urbanisation trends.
As hinted earlier, however, Malanima does pay attention to this effect. In his 2011 article,
reference is made to the progress of the Italian silk industry, in which a number of peasant families
were involved, and its possible impact on the share of secondary- and tertiary-sector output. In order
to capture this aspect of the “progress”, he allows the non-agricultural share of the labour force to
change over time (Malanima 2011, pp. 184-85). Thus,
Yna(t)/Y(t) = (Yna(t0)/Y(t0)) × (Lna(t)/L na(t0)),
(2)
where Lna stands for the share of the non-agricultural labour force with t0 being the base year and t
varying from t-50, t-100 and so on.
2
The non-agricultural share in the total product (including intermediate inputs) was 52%. Note that
Nishikawa’s “non-agricultural” category includes forestry, fishing and marine products.
3
Will this work for the Japanese case? It will probably not, as data availability can be a serious
problem. First, little is known about the Lna levels in the Tokugawa past: no serious estimates have
been made except for status-based classifications of the population. Second, even when some
indications about the distribution of the commoner population into peasants, artisans and merchants
are available, it is difficult to quantify the significance of non-agricultural by-employment, which
was widespread in the countryside. As for the period after 1874, Saito and Settsu (2010) gives
estimates for the sectoral shares of the labour force, which has been made possible by the use of two
prefectural-level matrix tables, for 1879 and 1825 respectively, of principal and subsidiary
occupations. In 1885, for example, it is revealed that the revised estimates of the secondary- and
tertiary-sector shares were both 18%, 2 to 3 percentage points higher than the previous estimates
based only on information about principal occupations. Also revealed is that the overall percentage
by-employed of the labour force calculated from the 1879 data for a silk-producing region was
19.5% (with intra-sectoral by-employments excluded). But, unfortunately, this kind of information
about subsidiary employments is virtually non-existent in Tokugawa times.
Given these difficulties, population density may be introduced as another proxy for the
non-primary share of the labour force, which represents supply-side factors in the thesis of
proto-industrialisation. It is widely known that population density could vary significantly across
districts even in a totally rural region, and more importantly that industrialised rural districts tend to
show up as highly densely populated clusters on the population density map.3 It implies that
cross-sectional, rather than simple time-series analysis is needed in order to obtain appropriate
coefficients used for the estimation. This consideration leads us to economic geography of
proto-industrial Japan.
One of the drivers of Japan’s proto-industrialisation was, as in many other cases, textiles.
Cotton and silk were two of the most dynamic proto-industrial commodities in the period from the
late eighteenth to the early twentieth century, but cotton was to the warmer west and silk to the east.
Cotton was cultivated and woven mostly in lowlands of central and western provinces while silk
reeling and weaving took root in upland districts of central and eastern provinces. As a result, the
relationship between food production (i.e. grain growing) and proto-industry varied with this
difference in economic geography, so did the relationship between proto-industry and population
density (Saito 1983). However, the latter relationship may have been influenced also by the regional
3
See for example early nineteenth-century district-level Belgian and English population density maps
prepared by Eric Buyst and Leigh Shaw-Taylor, respectively, for O. Saito and L. Shaw-Taylor, eds.,
Occupational Structure and Industrialisation in a Comparative Perspective (forthcoming).
4
level of urbanisation, representing demand-side factors. The urbanisation rate was generally higher
in the east than in the west. Of course, the percentage urban was considerably higher in the Kinai
with both Kyoto and Osaka (30%) and also in South Kanto with Edo (19%), than in any other
regions in 1874, but the average level of urbanisation was 6.9% for the rest of the west (i.e. western
provinces other than the Kinai) whereas that was 8.4% for the rest of the east (both are weighted
averages calculated from tables tables A.1 and A.2B). This rather unexpected finding may well have
been a reflection that after the opening of the country into world trade the silk industry in the central
and eastern provinces benefited greatly from exporting their products whereas the cotton trade in the
west was under strong pressure from European imports. That said, however, there are factors that
could have closed up the observed east-west differences. First, while cotton cultivation and spinning
were hit hard by the Western impact, the weaving districts survived the challenge and many actually
started growing in the early Meiji period. Second, some of those successful cotton districts were
found on the eastern side the east-west boundary as well. The list at the end of the Tokugawa period
included Niikawa (Toyama), Echizen (Fukui), Mōka and Sano (both Tochigi). Finally, the size of
non-textile proto-industrial production should not be underestimated. Indeed, the largest single
manufacturing industry in Tokugawa times was brewing and the brewing of rice wine, soy sauce and
miso (fermented soybeans) were found in many areas on both sides of the boundary, together with
other manufacturing activities such as paper making and clothing (see Saito and Tanimoto 1999;
Shimbo and Saito 1999). All this suggests that the relationships between population density and
urbanisation, on the one hand, and the secondary- or tertiary-sector share, on the other, varied across
the eastern half as well as the western half of the country. It is therefore worth exploring the
relationships cross-sectionally.
As in the Malanima method, regression analysis is involved also in our methodology. But,
given the nature of the tasks mentioned above, probably the best strategy to estimate the regression
coefficients to be used for the estimation is to construct from Meiji statistics a prefectural-level panel
database first. It will enable us to assign, when conducting regression analysis, year dummies for
mid- to late Meiji benchmark years and prefectural dummies to Tokyo, Osaka and a few others, in
which the impact of industrialisation or westernisation is thought to have been felt particularly
strongly. The results from such panel data analysis with prefectures, not aggregated regions, as the
unit of observations will provide better estimates with fuller confidence for the period before Meiji.
3. The Meiji-period database
5
The panel dataset includes prefectural output and population figures for the three benchmark years,
1874, 1890 and 1909. For prefectural GDP we have adopted new estimates by Settsu (2014). Since
his estimates are sector specific, our non-agricultural shares are also sector specific: the
secondary-sector share and the tertiary-sector share can be separately calculated for all the
prefectures. Urban population data for 1874 are not available, which are interpolated by using the
rate of population growth between 1873 and 1879. A topographical book (Nihon chishi teiyō) gives
population figures for cities with 10,000 or more for 1873 and a statistical report compiled by the
Army’s General Staff Office (Kyōbu seihyō) for 1879. For 1890 and 1909 the Population Statistics
of the Japanese Empire (Nihon Teikoku jinkō seitai tōkei) series are used. There were boundary
changes for a number of cities including six metropolitan cities, for which the corrected figures
reported in Umemura et al. (1983) are used. The total areas of prefectures are taken from the
Population Statistics, while the corresponding population totals are from Settsu (2014).
[Insert figures 1-2 here]
The key explanatory variables in the database are population density and the rate of
urbanisation. Both distributions are highly skewed. Figure 1 displays a scatter gram of population
density set against the level of urbanisation in 1874. Tokyo is in the upper-right corner of the graph
and Osaka on top in the middle, but all others cluster within the lower-left quarter: note that the
lowest low on the urbanisation axis was 0% (Saitama). From the graph two observations may be
made. First, the patterns of the skewed distributions suggest that if transformed logarithmically or
logit transformed both will get closer to the normal, although this should not be taken to mean that
the correlation would also become closer (see figure 2). This in fact leads us to the second point.
There is no clear correlation between the two variables if Tokyo and Osaka are excluded, indicating
that in the early Meiji period urbanisation was still not the chief determinant of population density.
The finding lends support to our supposition that both population density and urbanisation variables
were separately at work in accounting for the non-agricultural shares of output. A pair of the scatter
grams of both population density and the urbanisation rate, set against the secondary and tertiary
sectoral shares, indicates that in both cases the correlation is a positive one (figures 3-4).
[Insert figures 3-4 here]
The number of prefectures that appeared in the 1874 statistics is 63 including three
6
metropolitan prefectures of Tokyo, Osaka and Kyoto, but the number of prefectures was reduced to
47 in the mid- and late-Meiji years of 1890 and 1909. With Hokkaido and Okinawa excluded from
the analysis because their economic conditions and development levels were vastly different from
other prefectures of the period in question, and also with Saitama merged into its neighbouring
prefecture of Gunma so as to avoid the zero to be logit transformed, the remaining 60 are grouped
into 44.
Among the 44 prefectures, as the discussion at the end of the previous section suggests, there
were clear east-west differences, as a result of which the two variables may have had a differential
effect on the secondary- and tertiary-sector shares between the east and west. In order to include in
regression analysis this aspect of interactions, the prefectures are divided into the east and west
zones as set out in table A.3 (which also shows how early-Meiji prefectures are matched to
Tokugawa-period provinces).
Also needed is the identification of prefectures which had already been affected by the uneven
process of industrialization and westernisation. Obviously Tokyo had received from the very
beginning of the new Meiji era the impact of the government’s modernisation drive, so did Osaka.
Then Osaka started to undergo the transformation after the late 1880s when Japan’s “industrial
revolution” took off, whose influence diffused to Aichi and Fukuoka by 1909. These “modern”
prefectures, together with the year dummies for 1890 and 1909, will be treated separately in the
regression analysis.
4. Regression analysis
On the basis of this Meiji-period database, regression analysis is conducted for the secondary-sector
and the tertiary-sector shares separately. The same set of variables will be included on the right-hand
side of the equation, i.e. the two major variables of population density and the urbanisation rate, and
the prefectural and year dummy variables, but it is expected that while the two major variables will
turn out to be at work simultaneously in both cases, the two may behave somewhat differently.
Given the nature of the sectors, it is likely that while for secondary-sector output growth the
supply-side variable (population density) had a comparatively greater effect than the demand-side
variable (urbanisation), the other way round was the case for output growth in the tertiary sector.
The secondary-sector share variable (Sshare) used here is defined in relation to the size of the
primary sector whose estimates are less problematic for the Tokugawa period. Sshare is, therefore,
measured as the proportion of secondary-sector output to the sum of primary- and secondary-sector
7
output. Thus, the basic regression equation is as follows:
!"!
!"ℎ!"#
!
= !! + !! !"# + !! !"
+ !! !! + !! !! + !! ! + !! !"1
1 − !"ℎ!"#
1−!
(3)
+ !! !"2 + !
where D stands for population density (measured in terms of the number of persons per chō (1 chō
equals to 0.99 hectares), U for the urbanization rate, XD for lnD’s interaction with the east-west
dummy (east = 1), and XU for logitU’s interaction with the east-west dummy. Also included is a set
of dummy variables, i.e. a prefectural dummy for “modernised” prefectures and two year dummies
for 1890 and 1909 (m, yr1, and yr2 respectively). Just for comparison, the dependent variable that is
not logit transformed is also regressed on the same right-hand variables. The results for these two
specifications are set out in table 1.
[Insert Table 1 here]
The basic regression specification for the tertiary sector is essentially the same as equation (3):
!"!
!"ℎ!"#
!
= !! + !! !"# + !! !"
+ !! !! + !! !! + !! ! + !! !"1
1 − !"ℎ!"#
1−!
(4)
+ !! !"2 + !
where Tshare is the proportion of tertiary-sector output to the sum of primary and tertiary sector
output. The other variables are the same as for the secondary sector. The results of this regression are
shown in table 2.
[Insert Table 2 here]
The results for the two major explanatory variables and the set of dummy variables in the two
tables are all satisfactory with the expected signs and reasonably high t ratios (except for the
urbanisation variable in the secondary sector equation, which is statistically significant only at the
10% level). This means, first, that both population density and urbanisation were joint determinants
of secondary- and tertiary-sector growth. Second, the population density effect seems to have been
8
relatively greater for secondary-sector output than for tertiary-sector output: if the results for
equation (3) in table 1 is compared with those for the same equation in table 2, the size of the density
coefficient relative to that of the urbanisation coefficient in the case of the secondary sector was
about twice as large as in the tertiary sector. To put it simply, the supply-side factor was crucial for
secondary-sector growth while it was the urban influence that was comparatively more important for
tertiary-sector growth.
5. Back projection for Tokugawa-period national accounting
Based on specification (5) in tables 1 and 2, the non-agricultural sector shares are back projected to
the Tokugawa period. The benchmark years are 1600, 1650, 1720, 1750, 1800, 1850, and 1874. It
should be remembered, however, that the age of proto-industrialisation and de-urbanisation began
after 1720, not before. It is well known that the eighth Shogun, Yoshimune, as one of his
government’s reform effort tried to encourage domestic production by lifting in 1720 some of the
restrictions on the import of foreign books (one of the “seclusion” policy measures adopted in the
1630s). One area he was particularly concerned with was horticulture, as a result of which many
Chinese books on botany and horticulture were imported. Since the list of imports in the seventeenth
century had included raw silk, sugar and ginseng, the target was to “substitute domestic goods for
imports”. This import substitution effort is said to have been particularly successful for raw silk:
“from the 1720s on, the silk output was sufficient to start driving down the price of imports”.4
Although it is naive to link the government’s effort to this output increase directly, it seems certain
that silk reeling did spread in rural villages from the early eighteenth century onwards. In other
words, when back projected to 1650 and also to 1600, we should drop the population density term
out of the estimated equation, as it was from 1720 onwards when that effect started to work
significantly.
Thus the specification used varies from period to period:
For 1720-1850,
!"!
!"ℎ!"#
=−
1 − !"ℎ!"#
.7411 + 0.434!"# + 0.098!"
4
!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!(5)
1−!
Totman (1993), p. 312, quoting Ohba and Toby’s article on “Seek knowledge throughout the world”
and Tashiro’s on “Tsushima han’s Korean trade”.
9
!"!
!"ℎ!"#
!
= −0.173 + 0.4535!"# + 0.3305!"
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!(6)
1 − !"ℎ!"#
1−!
For 1600 and 1650,
!"!
!"ℎ!"#
!
= −1.7411 + 0.098!"
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!(7)
1 − !"ℎ!"#
1−!
!"!
!"ℎ!"#
!
= −0.173 + 0.3305!"
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!(8)
1 − !"ℎ!"#
1−!
Specifications (7) and (8) implies that for periods before the proto-industrial phase we have adopted
the Malanima model.
Tokugawa-period data necessary for the backward estimation are primary-sector output,
population density, and urbanisation rates. (1) For primary-sector output, kokudaka data with
adjustments in one form or another have been utilized by many scholars. Here, however, our own
new estimates will be used (Takashima 2014), shown in table A.4. (2) Data for the calculation of
population density are taken from Kito (1996) for the Tokugawa-period benchmark years and from
Settsu (2014) for 1874. Both are region-specific, so that check can be made at regional levels. For
the seventeenth century, however, Kito’s population estimates appear to be so low that Saito’s
estimate is adopted for 1600,5 with the figure for 1650 interpolated linearly between 1600 and 1720.
(3) Urbanisation rates are estimated for regions first. For cities whose population data are available
from previous studies we adopted their estimates; for other cities, the population totals reported in
the Chishi teiyō for 1872 are extrapolated backwards by using urban population growth rates for
each region calculated from Saito (1984). Once all the estimates are made, the proportion of
settlements with population of 10,000 inhabitants or over is calculated for the eastern and the
western half, and then for the entire nation (the results are set out in table A.2).
[Insert Table 3 here]
Our back projection results for the 1600-1874 period are shown in table 3 and in figure 7
graphically. They are in 1990 international dollars and compared with Maddison’s estimates for the
same period. The Maddison estimates were made by making an assumption that Japan’s
non-agricultural output grew faster than agricultural output over the period between 1500 and 1820
5
Unpublished, but the estimates are quoted by Farris (2006), pp. 231, 262. For the background of Saito’s
new estimation, see his forthcoming article.
10
(Maddison 2001, pp. 256-58). This implies that although the growth rate differs from sub-period to
sub-period, moderate growth persisted throughout the Tokugawa period. In contrast, our estimates
are far more articulated. While our GDP per capita estimate was a little higher than Maddison’s in
1600, it declined over the seventeenth century, reaching a trough around 1720. Since then, however,
there took place growth until 1850, thanks to the emergence of rural-centred development. The
growth was moderate, but a little stronger than Maddison suggested. After the opening of the
country into world trade the growth accelerated. The rate of growth between 1850 and 1874 was not
very different between the two estimates, but the level of our GDP per capita figure exceeded that of
Maddison’s in 1874.
[Insert Figure 5 here]
Table 4 breaks down the estimated national product into the three sectors. While the
primary-sector share declined from 67.8% in 1600 to 58.7% in 1874, there was a gradual but steady
increase in the secondary-sector share from 9.1% to 10.9% and a comparatively more marked
increase in the tertiary-sector share from 23.1% to 30.4%. Tokugawa Japan’s tertiary-sector growth
was associated with the seventeenth-century rise in the level of urbanisation, but the share of
secondary-sector output hardly rose during the seventeenth and eighteenth centuries. Much of the
advance in that share was made during the nineteenth-century phase of the proto-industrialisation
process.
6. Conclusion
In our effort to estimate non-primary sector output shares in Tokugawa times, we have identified the
period of rural-centred growth and de-urbanisation in the latter half of the whole 270-year period.
While the Maddison estimates of GDP per capita failed to capture this phase of development, the
new estimates of both per capita GDP and its sectoral distribution have successfully accommodated
the emergence of such a historically distinct phase in Japan’s economic history. One may wonder if
our estimates are consistent with Shunsaku Nishikawa’s Chōshū studies because Nishikawa’s
estimate of the non-agricultural share of output was as high as 39% for the 1840s. In fact, if
re-calculated on the primary, secondary and tertiary basis, rather than the non-agricultural basis, the
Chōshū share of non-primary sector output turns out to be 31%,6 a little lower than our 1850
6
This re-calculation can be done only from his 13-sector Input-Output table. See Nishikawa (2012),
ch.12.
11
estimate of 38% in table 4 (= 9.5 + 28.8). The more substantive difference is found in that
Nishikawa’s profile of the Chōshū economy was more industrial while, according to our estimates,
the late Tokugawa economy looks less industrial but more service-oriented.
Our new series have also suggested that seventeenth-century Japan experienced negative
growth in per capita terms, which was not apparent in the Maddison estimates. Whether or not this
downturn was real, and if it was, then whether it was Malthusian in nature or something that should
be accounted for by a different set of factors – these are remaining big issues.
12
Tables and Figures
Figure 1. Population density and the urbanisation rate: 45 prefectures in 1874
6.0
5.0
Population density
4.0
3.0
2.0
1.0
0.0
0.0
0.1
0.2
0.3
0.4
Urbanisation rate
Source: Table A.2.
13
0.5
0.6
0.7
Figure 2. Population density (log transformed) and the urbanisation rate (logit transformed): 44
prefectures in 1874
2.0
1.5
Population density
1.0
0.5
0.0
-0.5
-1.0
-5.0
-4.0
-3.0
-2.0
Urbanisation rate
Source: Table A.2.
14
-1.0
0.0
1.0
Figure 3. Population density (log transformed) and the secondary- and tertiary-sector shares
(logit transformed): 44 prefectures in 1874
2.0
1.0
Share
0.0
-1.0
-2.0
-3.0
LogitSshare
LogitTshare
-4.0
0.0
1.0
2.0
3.0
4.0
Population density
Source: Table A.2A.
15
5.0
6.0
Figure 4. The rate of urbanisation (logit transformed) and the secondary- and tertiary-sector shares
(logit transformed): 44 prefectures in 1874
2.0
1.0
Share
0.0
-1.0
-2.0
-3.0
LogitSshare
LogitTshare
-4.0
-5.0
-4.0
-3.0
-2.0
-1.0
Urbanisation rate
Source: Table A.2B.
16
0.0
1.0
0.0259
[2.65]***
Urbanization rate
0.7695
0.1771
[6.23]***
0.1411
[7.73]***
0.7746
0.1861
[6.71]***
0.1203
[7.16]***
0.7793
0.1776
[6.38]***
0.1338
[7.40]***
0.0755
[4.36]***
0.2264
[6.54]***
-0.0673
[-2.56]**
-0.0143
[-1.91]*
0.0295
[3.06]***
0.0873
[5.83]***
(4)
0.7414
-1.7411
[-10.95]***
0.8265
[8.69]***
0.4738
[5.38]***
1.0521
[5.21]***
0.0980
[1.79]*
0.4340
[5.37]***
(5)
0.7440
-1.7796
[-11.10]***
0.8871
[8.62]***
0.5429
[5.48]***
1.0219
[5.06]***
-0.0646
[-1.50]
0.1146
[2.06]**
0.4469
[5.52]***
(6)
0.7432
-1.7407
[-10.98]***
0.8057
[8.39]***
0.4586
[5.18]***
1.0303
[5.10]***
-0.2075
[-1.37]
0.0977
[1.79]*
0.4758
[5.52]***
(7)
0.7455
-1.7782
[-11.12]***
0.8653
[8.33]***
0.5263
[5.29]***
1.0015
[4.96]***
-0.2010
[-1.33]
-0.0629
[-1.46]
0.1138
[2.05]**
0.4870
[5.66]***
(8)
17
Notes: The dependent valuable in columns (1)-(4) is share of the added value of secondary sector in the sum of added value of primary and secondary sectors. The
dependent valuable in columns (5)-(8) is obtained by converting (1)-(4) into a logit value respectively. *** denotes significant at the 1 per cent level. ** denotes
significant at the 5 per cent level. * denotes significant at the 10 per cent level. All observations are 132.
Sources: See text.
0.7644
0.1860
[6.55]***
Constant
Adjusted R2
0.1272
[7.49]***
Year 1909 dummy
0.0601
[3.88]***
0.0651
[4.14]***
Year 1890 dummy
0.0810
[4.61]***
0.2330
[6.59]***
0.2332
[6.51]***
0.2402
[6.67]***
Prefectural dummy
(Modernized=1)
0.0258
[2.70]***
0.0848
[5.62]***
(3)
-0.0687
[-2.59]**
-0.0149
[-1.94]*
0.0297
[3.02]***
0.0739
[5.15]***
(2)
Population density × Regional dummy
(East Japan=1)
Urbanization rate × Regional dummy
(East Japan=1)
0.0710
[4.92]***
Population density
(1)
Table 1. OLS regression for the determinant of share of the secondary sector
0.0706
[7.67]***
Urbanization rate
0.8103
0.4584
[16.88]***
0.8093
0.4549
[16.93]***
0.8088
0.4585
[16.82]***
0.0511
[3.01]***
0.0602
[3.40]***
0.1121
[3.26]***
-0.0031
[-0.12]
0.0059
[0.81]
0.0691
[7.31]***
0.0939
[6.39]***
(4)
0.8258
-0.1730
[-1.41]
0.2873
[4.23]***
0.3099
[4.23]***
0.5790
[3.72]***
0.3305
[7.84]***
0.4535
[7.74]***
(5)
0.8260
-0.1513
[-1.22]
0.2484
[3.24]***
0.2758
[3.46]***
0.5961
[3.82]***
0.0364
[1.09]
0.3212
[7.47]***
0.4463
[7.13]***
(6)
0.8244
-0.1730
[-1.41]
0.2869
[4.18]***
0.3093
[4.15]***
0.5784
[3.69]***
-0.0061
[-0.05]
0.3305
[7.81]***
0.4548
[6.80]***
(7)
0.8246
-0.1512
[-1.21]
0.2476
[3.19]***
0.2747
[3.39]***
0.5951
[3.78]***
-0.0099
[-0.08]
0.0365
[1.09]
0.3212
[7.44]***
0.4483
[6.68]***
(8)
18
Notes: The dependent valuable in column (1)-(4) is share of the added value of secondary sector in the sum of added value of primary and secondary sectors. The
dependent valuable in columns (5)-(8) is obtained by converting (1)-(4) into a logit value respectively. *** denotes significant at the 1 per cent level. All
observations are 132.
Sources: See text.
Adjusted R2
0.8108
0.4549
[17.00]***
Constant
Year 1909 dummy
0.0575
[3.83]***
0.0658
[4.04]***
0.0577
[3.89]***
0.0661
[4.13]***
Year 1890 dummy
0.0514
[3.06]***
0.0605
[3.47]***
0.1094
[3.20]***
0.1125
[3.29]***
0.1097
[3.23]***
Prefectural dummy
(Modernized=1)
0.0706
[7.64]***
0.0949
[6.50]***
(3)
-0.0025
[-0.10]
0.0059
[0.81]
0.0691
[7.34]***
0.0932
[6.80]***
(2)
Population density × Regional dummy
(East Japan=1)
Urbanization rate × Regional dummy
(East Japan=1)
0.0944
[6.94]***
Population density
(1)
Table 2. OLS regression for the determinant of share of the tertiary sector
Table 3. GDP per capita in Japan, 1600-1874: Maddison’s and our estimates compared
(1990 international dollars)
Maddison
1600
Our estimates
520
605
1650
1700
619
570
1720
570
1750
622
1800
703
1820
669
1850
679
777
1874
756
860
Rates of growth
1600-1700
1700-1820
0.09%
0.13%
1600-1650
0.05%
1650-1720
-0.12%
1720-1750
0.29%
1750-1800
0.25%
1820-1850
0.05%
1800-1850
0.20%
1850-1874
0.45%
1850-1874
0.42%
1600-1700
0.09%
1600-1720
-0.05%
1700-1874
0.16%
1720-1874
0.27%
Sources: Maddison (2001), table B-21, p. 264; for Japan, see text.
19
Figure 5. GDP per capita in Japan, 1600-1874: Maddison’s and our estimates compared
(1990 international dollars)
1000
800
600
400
New estimates
Maddison
200
0
1600
1650
1700
1750
Source: Table 3.
20
1800
1850
Table 4. Output per capita in koku and its sectoral distribution: Japan, 1600-1874
1600
1650
1720
1750
1800
1850
1874
Japan (koku per capita)
2.65
2.72
2.5
2.73
3.09
3.41
3.77
Share
Primary
0.678
0.624
0.627
0.617
0.619
0.618
0.587
Secondary
0.091
0.091
0.093
0.093
0.093
0.095
0.109
Tertiary
0.231
0.284
0.280
0.289
0.288
0.288
0.304
Sector
Source: See text.
21
Appendix tables
Table A.1. Population by region, 1600-1874
(1,000 koku)
Region
1600
1650
1700
1720
1730
1750
1800
1850
1874
1
East Tohoku
—
—
—
2,333
2,307
2,203
1,912
1,993
2,419
2
West Tohoku
—
—
—
1,043
1,041
1,016
1,030
1,117
1,263
3
North Kanto
—
—
—
2,190
2,189
2,143
1,691
1,621
1,820
4
South Kanto
—
—
—
3,902
3,927
3,914
3,508
3,737
3,782
5
Niigata / Hokuriku
—
—
—
2,593
2,608
2,593
2,739
3,093
3,457
6
Tosan
—
—
—
1,251
1,269
1,284
1,356
1,435
1,471
7
Tokai
—
—
—
2,623
2,639
2,633
2,761
2,903
2,799
8
Kinai
—
—
—
2,696
2,658
2,567
2,446
2,366
2,184
9
Around Kinai
—
—
—
3,683
3,658
3,549
3,498
3,594
3,562
10
Sanin
—
—
—
837
857
887
1,006
1,052
1,029
11
Sanyo
—
—
—
2,407
2,432
2,445
2,661
2,935
3,042
12
Shikoku
—
—
—
1,822
1,850
1,875
2,065
2,345
2,456
13
North Kyushu
—
—
—
2,364
2,387
2,393
2,411
2,606
2,996
14
South Kyushu
—
—
—
1,294
1,334
1,406
1,498
1,683
2,236
17,000
21,846
28,074
31,037
31,156
30,908
30,582
32,481
34,516
Total
Sources and notes: Year 1600 from Saito’s new estimations quoted by Farris (2006). Years 1750 from Kito
(1996). Year 1874 from Settsu (2014). Other years are derived by linear interpolation based on the population
estimates by Kito (1996).
22
Table A.2. Population density and urbanisation rate, 1600-1874
A. Population density, 1600-1874
(per cho)
Region
1600
1650
1700
1720
1730
1750
1800
1850
1874
1
East Tohoku
—
—
—
0.51
0.50
0.48
0.42
0.43
0.53
2
West Tohoku
—
—
—
0.52
0.52
0.51
0.51
0.56
0.63
3
North Kanto
—
—
—
1.22
1.22
1.19
0.94
0.90
1.01
4
South Kanto
—
—
—
2.78
2.80
2.79
2.50
2.66
2.69
5
Niigata / Hokuriku
—
—
—
1.02
1.03
1.02
1.08
1.22
1.36
6
Tosan
—
—
—
0.57
0.58
0.58
0.62
0.65
0.67
7
Tokai
—
—
—
1.36
1.37
1.37
1.43
1.51
1.45
8
Kinai
—
—
—
3.89
3.84
3.70
3.53
3.41
3.15
9
Around Kinai
—
—
—
1.43
1.42
1.37
1.35
1.39
1.38
10
Sanin
—
—
—
0.82
0.83
0.86
0.98
1.03
1.00
11
Sanyo
—
—
—
1.16
1.17
1.18
1.28
1.42
1.47
12
Shikoku
—
—
—
0.99
1.01
1.02
1.12
1.28
1.34
13
North Kyushu
—
—
—
1.39
1.40
1.41
1.42
1.53
1.76
14
South Kyushu
—
—
—
0.53
0.54
0.57
0.61
0.68
0.91
0.59
0.76
0.97
1.08
1.08
1.07
1.06
1.13
1.20
Total
B. Urbanisation rate, 1600-1874
(%)
Region
1600
1650
1700
1720
1730
1750
1800
1850
1874
1
East Tohoku
—
—
—
6.6
7.3
8.1
9.7
9.3
6.5
2
West Tohoku
—
—
—
10.2
10.3
10.7
10.6
10.8
10.7
3
North Kanto
—
—
—
2.5
2.6
2.6
3.4
4.8
4.6
4
South Kanto
—
—
—
25.3
27.7
34.2
38.2
35.7
19.0
5
Niigata / Hokuriku
—
—
—
10.3
10.7
12.2
12.4
12.5
13.0
6
Tosan
—
—
—
3.1
3.8
3.8
3.6
3.5
3.3
7
Tokai
—
—
—
7.9
7.9
8.1
7.1
6.9
8.6
8
Kinai
—
—
—
34.7
35.8
38.5
36.8
34.8
30.2
9
Around Kinai
—
—
—
7.4
7.4
7.6
7.3
6.9
6.1
10
Sanin
—
—
—
8.2
8.2
8.3
7.8
7.9
6.6
11
Sanyo
—
—
—
7.1
7.0
6.8
6.0
5.1
7.2
12
Shikoku
—
—
—
6.1
6.0
5.8
5.5
5.1
7.6
13
North Kyushu
—
—
—
6.4
6.4
6.4
5.6
4.5
6.6
14
South Kyushu
—
—
—
6.9
6.9
7.5
8.9
7.8
7.4
6.1
13.5
12.0
11.6
12.1
13.4
13.3
12.4
10.3
Total
Notes: 1 cho = 0.00991736 km2. Urban population includes persons living in settlements of at least 10,000.
Sources: See text.
23
Tosan
Tokai
Kinai
6
7
8
Sanyo
Shikoku
North Kyushu
South Kyushu
11
12
13
14
Higo, Hyuga, Osumi, Satsuma
Chikuzen, Chikugo, Hizen, Iki, Tsushima,
Buzen, Bungo
Awa, Sanuki, Iyo, Tosa
Mimasaka, Bizen, Btchu, Bingo, Aki,
Suo, Nagato
Inaba, Hoki, Izumo, Iwami, Oki
Omi, Iga, Ise, Shima, Kii, Harima, Awaji,
Tanba, Tango, Tajima
Izu, Suruga, Totomi, Mikawa, Owari,
Mino
Yamashiro, Yamato, Izumi, Kawachi,
Settsu
Kai, Shinano, Hida
Sado, Echigo, Etchu, Noto, Kaga,
Echizen, Wakasa
Musashi, Sagami, Kazusa, Shimosa, Awa
Hitachi, Kozuke, Shimotsuke
Dewa (Uzen, Ugo)
Mutsu (Rikuo, Rikuchu, Rikuzen,
Iwashiro, Iwaki)
Tokugawa period (kuni: old province)
Shirakawa, Miyazaki, Kagoshima
Myodo (ex-Awa and ex-Samuki), Ehime.
Kochi
Fukuoka, Mizuma, Kokura, Saga,
Nagasaki, Oita
Hokujo, Okayama, Oda, Hiroshima,
Yamaguchi
Tottori, Shimane, Hamada
Shiga, Mie, Watarai, Shikama, Toyooka,
Wakayama, Myodo (ex-Awaji)
Ashigara (ex-Izu), Hamamatsu, Shizuoka,
Aichi, Gifu
Kyoto. Osaka, Nara, Hyogo, Sakai
Yamanashi, Nagano, Chikuma,
Tokyo, Saitama, Niihari (ex-Shimosa),
Chiba, Kumagaya (ex-Musashi),
Kanagawa, Ashigara (ex-Sagami)
Niigata, Aikawa, Niikawa, Ishikawa,
Tsuruga
Ibaraki, Niihari (ex-Hitachi), Tochigi,
Kumagaya (ex-Kozuke)
Akita, Yamagata, Sakata, Okitama
Aomori, Iwate, Miyagi, Mizusawa,
Fukushima, Iwamae, Wakamatsu
Early Meiji (fu, ken: prefecture)
Miyazaki, Kumamoto, Kagoshima
Fukuoka, Nagasaki, Saga, Oita
Tokushima, Kagawa, Ehime, Kochi
Okayama, Hiroshima, Yamaguchi
Shiga, Mie, Wakayama, Hyogo
(ex-Tajima and some part of ex-Tanba),
Kyoto (ex-Tango, and some parts of
ex-Tanba)
Tottori, Shimane
Kyoto (ex-Yamashiro), Osaka, Hyogo
(some parts of ex-Settsu)
Shizuoka, Aichi, Gifu (ex-Mino)
Yamanashi, Nagano, Gifu (ex-Hida)
Niigata, Toyama, Ishikawa, Fukui
Tokyo, Saitama, Chiba, Kanagawa
Ibaragi, Tochigi, Gunma
Akita, Yamagata
Aomori, Iwate, Miyagi, Fukushima
Current prefecture
24
Notes and sources: Excluding Ezochi and Ryukyu (currently Hokkaido and Okinawa prefecture). Regional division is based on the regional devisions from Saito (1983),
Kito (1996). Adjustment of the prefectural border was made by using the table of the transition of prefecure/county in Meiji bunken shiryo kankokai (1964).
Sanin
10
Around Kinai
Niigata / Hokuriku
5
9
South Kanto
4
West
Japan
North Kanto
3
East Tohoku
West Tohoku
East
Japan
2
1
East/West
Table A.3. Regional division of Japan
Table A.4. Agricultural production (Incl. the forestry and fishery sector), 1600-1874
(1,000 koku)
Region
1600
1650
1700
1720
1730
1750
1800
1850
1874
1
East Tohoku
2,485
3,108
4,689
4,995
5,156
5,492
6,426
6,894
6,696
2
West Tohoku
1,802
2,110
2,437
2,528
2,574
2,670
2,923
3,328
3,640
3
North Kanto
2,559
2,731
2,951
3,034
3,077
3,164
3,391
3,893
4,364
4
South Kanto
1,676
2,795
3,594
3,794
3,897
4,114
4,704
5,663
6,414
5
Niigata / Hokuriku
4,214
4,468
5,255
5,405
5,481
5,637
6,046
6,849
7,557
6
Tosan
521
1,310
2,095
2,194
2,245
2,351
2,637
3,066
3,379
7
Tokai
1,961
2,727
3,694
3,829
3,898
4,039
4,415
5,041
5,539
8
Kinai
1,550
1,801
2,038
2,184
2,261
2,423
2,876
3,686
4,376
9
Around Kinai
4,192
4,439
5,020
5,247
5,364
5,606
6,257
7,621
8,904
10
Sanin
1,336
1,373
1,534
1,616
1,659
1,748
1,991
2,181
2,216
11
Sanyo
1,895
2,427
3,621
3,759
3,830
3,975
4,362
5,546
6,893
12
Shikoku
1,090
1,813
2,784
2,979
3,082
3,298
3,901
4,928
5,768
13
North Kyushu
2,866
3,310
3,935
4,101
4,187
4,364
4,837
5,553
6,078
14
South Kyushu
2,469
2,654
2,902
3,044
3,118
3,271
3,685
4,209
4,528
30,616
37,065
46,550
48,708
49,828
52,151
58,452
68,457
76,351
Total
Sources: See text.
25
Table A.5. Output per capita in koku for eastern and western Japan, 1600-1874
1600
1650
1720
1750
1800
1850
1874
The east
-
-
2.53
2.79
3.26
3.52
3.60
The west
-
-
2.48
2.68
2.92
3.31
3.94
2.65
2.72
2.5
2.73
3.09
3.41
3.77
Japan
Source: See text.
Table A.6. Output per capita in koku and its sectoral distribution for eastern and western Japan,
1600-1874
Sector
The east
The west
Japan
1600
1650
1720
1750
1800
1850
1874
Primary
0.640
0.624
0.625
0.621
0.614
Secondary
0.090
0.090
0.089
0.090
0.107
Tertiary
0.269
0.285
0.286
0.288
0.278
Primary
0.613
0.610
0.612
0.615
0.562
Secondary
0.096
0.096
0.097
0.099
0.112
Tertiary
0.291
0.294
0.290
0.287
0.326
Primary
0.678
0.624
0.627
0.617
0.619
0.618
0.587
Secondary
0.091
0.091
0.093
0.093
0.093
0.095
0.109
Tertiary
0.231
0.284
0.280
0.289
0.288
0.288
0.304
Source: See text.
26
References
Official sources
Rikugun Sanbōkyoku
Sanbōkyoku.
ed. (1875), Kyōbu seihyō
Naimushō Chirikyoku
Nipposha.
ed. (1874-79), Nihon chishi teiyō
, Tokyo: Rikugun
, Tokyo:
Naimushō Kangyōryō
ed. (1875), Meiji 7 nen fuken bussanhyō
, Tokyo: Naimushō Kangyōryō.
Naimushō Chirikyoku
ed. (1881), Dai-Nihon fuken bunkatsuzu
Tokyo: Naimushō Chirikyoku.
,
Naikuaku Tōkeikyoku
seitai tōkei
Tōkeikyoku.
ed. (1911), Meiji 41 nen 12 gatsu 31 nichi, Nihon Teikoku jinkō
, Tokyo: Naikaku
Naikaku Tōkeikyoku
seitai tōkei
Tōkeikyoku.
ed. (1916), Taisho 2 nen 12 gatsu 31 nichi, Nihon Teikoku jinkō
, Tokyo: Naikaku
Other sources
Álvarez-Nogal, C. and L. Prados de la Escosura (2007), "The decline of Spain (1500-1850):
conjectural estimates", European Review of Economic History 11 (3), pp.319-366.
Bassino, J.P., S. Broadberry, K. Fukao, B. Gupta and M. Takashima (2011), "Japan and the Great
Divergence, 730-1870", working paper.
http://www2.lse.ac.uk/economicHistory/pdf/Broadberry/JapanGreatDivergence.pdf
De Vries, J. (1984), European Urbanization 1500-1800, London: Methuen.
Farris, W.W. (2006), Japan’s Medieval Population: Famine, fertility, and warfare in a
transformative age, Honolulu: University of Hawaiʻi Press.
Kito, H.
(1996), "Meiji-izen Nihon no chiiki jinkō
Keizai Ronshū 41 (1-2), pp.65-79.
", Jōchi
Maddison, A. (2001), The World Economy: A Millennial Perspective, Paris: Organisation for
Economic Co-operation and Development.
Malanima, P. (2003), "Measuring the Italian Economy. 1300-1861", Rivista di Storia Economica 19,
pp. 265-95.
Malanima, P. (2011), "The long decline of a leading economy: GDP in central and northern Italy,
1300-1913", European Review of Economic History 15 (2), pp.169-219.
Meiji bunken shiryō kankōkai
zenki sangyō hattatsuhi shiryō, bessatsu 1)
27
(1964) Meiji 9 nen zenkoku nōsanhyō (Meiji
(
1), Tokyo: Meiji bunken shiryō kankōkai.
Nishikawa, S. (1987), "The economy of Chōshū on the eve of industrialization", Economic Studies
Quarterly 38 (4), pp. 323-37.
Nishikawa, S.
(2012). Chōshū no keizai kōzō: 1840-nendai no mitorizu [Structure of the
Chōshū Economy: A Schema for the 1840s]
1840
,
Tokyo: Tōyō Keizai Shimpōsha.
Saito, O. (1983), "Population and the peasant family economy in proto-industrial Japan", Journal of
Family History 8 (1), pp.30-54.
Saito, O. (forthcoming), "Climate, famine, and population in Japanese history: a very long-term
perspective", B.L. Baten and P.C. Brown, eds., Environment and Society in the Japanese
Islands, Corvallis, OR: Oregon State University Press.
Saito, O., and M. Tanimoto (1999), "The transformation of traditional industries", in A. Hayami, O.
Saito and R.P. Toby, eds., Emergence of Economic Society in Japan, 1600-1859, Oxford:
Oxford University Press, pp. 268-300.
Saito, O., and T. Settsu (2010), "Unveiling historical occupational structures and its implications for
sectoral labour productivity analysis in Japan’s economic growth", manuscript for O. Saito
and L. Shaw-Taylor, eds., Occupational Structure and Industrialisation in a Comparative
Perspective (forthcoming). An earlier version is available as: Hitotsubashi University Global
COE Hi-Stat Discussion Paper Series no. 143.
Saito, S.
pp.48-63.
(1984) "Edo jidai no toshi jinkō
", Chiiki kaihatsu, 240,
Settsu, T. (2014), "Industrial structure: contribution to inequality and convergence", K. Fukao, J.-P.
Bassino, T. Makino, R. Paprzycki, and T. Settsu, Regional Inequality and Industrial Structure
in Japan: 1874-2008, Tokyo: Maruzen (forthcoming).
Shimbo, H., and O. Saito (1999), "The economy on the eve of industrialization", in A. Hayami, O.
Saito and R.P. Toby, eds., Emergence of Economic Society in Japan, 1600-1859, Oxford:
Oxford University Press, pp. 337-68.
Smith, T.C. (1969), "Farm family by-employments in preindustrial Japan", Journal of Economic
History 29 (4), pp. 241-57.
Smith, T.C. (1973), "Pre-modern economic growth: Japan and the West", Past and Present, 60,
pp.127-60.
Takashima, M. (2014), Agricultural Production, Proto-industrialization and Urbanization:
1598-1874, in K. Fukao, J.-P. Bassino, T. Makino, R. Paprzycki, and T. Settsu, Regional
Inequality and Industrial Structure in Japan: 1874-2008, Tokyo: Maruzen (forthcoming).
Totman, C. (1993), Early Modern Japan, Berkeley: University of California Press.
Umemura, M., N. Takamatsu, and S. Itoh,
(1983), Regional Economic
Statistics [Estimates of Long-term Ecnomic Statistics of Japan since 1868, vol.13]
(
13 ), Tokyo: Tōyō Keizai Shimpōsha.
28