How to shed light on the black box:

Transcription

How to shed light on the black box:
How to shed light on the black box:
Income approach for historical GDP estimates of the tertiary sector
with special reference to Japan, 1885-1940
Tokihiko Settsu
Draft paper for Asian Historical Economics conference
at Sano Shoin, Hitotsubashi University, Tokyo
13 September, 2012
Very Preliminary, please do not cite.
1
Introduction
Historical national accounts provided by economic historians have given us very
useful resources to draw an image of economic growth in long term perspective. At
the same time, however, we know that the figures are based on a bunch of
assumptions which are likely to be fragile and even if these assumptions are
modified slightly, the figures could be also altered dramatically.
When economic historians tackle the estimation of historical national
account, in many cases, they calculate the value added by each industry. On the
whole, we can find relatively abundant historical data on the primary and
secondary sector such as agriculture and manufacturing. It is because the
contemporary government had tendency to collect the data on production for tax
reasons or grasping the national capability. In contrast, as for the tertiary industry
or service sector, it often seems quite difficult to find the data on its production or
other relevant information. Thus, in many cases of estimation, value added of this
sector is based on very strong assumptions which seems more arbitrage or fragile
than those on the other sectors. It is the reason why the estimates on the tertiary
sector are regarded as one of the weakest points and the black box of historical
national accounts.1
Fortunately, we have had some other ways to shed light on this black box
even though the light was not bright enough to make a whole thing clear. For
estimating the value added of tertiary-sector activities, this paper brings forward
one of the promising methods, that is, income approach based on the historical
occupational data which were constructed by our project, the international network
for the comparative history of occupational structure (INCHOS). Furthermore, we also
point out the importance of taking into account the existence of by-employment for
conducting an income approach. We have already demonstrated that the effect of
by-employment can be considerable on the evaluation of sectoral labour productivity in
Saito and Settsu (2010). In this paper, we will evaluate the effect of by-employment on
the estimation of historical national account. In the next section, we briefly explain
what the income approach is and, after that, go over several cases of estimation of
tertiary sectors’ value added in previous research on some countries, and finally
consider the case of Japan and the effect of by-employment.
1
See also van Ark (1995) ‘Towards European Historical National Accounts’,
Scandinavian Economic History Review, vol.XLIII. no.1, (Stockholm: Scandinavian
Society for Economic and Social History and Historical Geography)
2
What is an income approach?
The value added of the primary or secondary sector is basically estimated as a
difference between total output value and intermediate input value (or the product
of total output value and value-added ratio) based on the production data. On the
other hand, it is obvious that we cannot apply the same approach (production
approach) to calculate the value added in the tertiary sector. In the previous studies,
they sometimes used income approach to overcome the problem of data scarcity. The
income approach is to calculate the value added in terms of the distribution side of
GDP explained by the formula as below.
GDP
Operatingsurplus
Compensationofemployees
Netindirecttax
Consumptionoffixedcapital
Operating surplus is distribution to capital for its contribution to the production
and Compensation of employees is to labour.2 Whole of the works using income
approach are based on this formula, but there are several variations of actual
calculation such as estimating each component of this formula directly or just
calculating a part of components and multiplying the inverse ratio as a percentage
of GDP which is obtained from the input/output table for later period of the country
(as the case of Taiwan). At all events, the key of the estimation is how calculate the
first two components or, at least, the compensation of employees.
Basically, the compensation of employees is estimated as the product of the
number of employees and average wage or income rate. Hence, the accuracy of the
estimation depends on the quality of data for the number of workforce by sector and
wage. In this respect, the outcome from the INCHOS project can contribute in some
measure to the estimation of historical national accounts.
Examples of estimation
The UK
It is well known that the UK has the long tradition of estimating the national
Modifying this formula, we can obtain some formulae related to the idea of GDP as
follows: Net domestic product (NDP) at factor cost = Operating surplus +
Compensation of employees; NDP at market price = Operating surplus +
Compensation of employees + Net indirect tax.
2
3
wealth such as by Gregory King, Joseph Massie and Patrick Colquhoun. And it was
the work done by Phyllis Dean and W. A. Cole (Dean and Cole 1962) that was the
first historical national accounts of the UK based on the idea of the System of
National Accounts (SNA) and this historic work had contributed to render an image
of economic growth in the industrial revolution. That image had been in common for
long years but Nick Crafts and C. K. Harley (Crafts 1985, Crafts and Harley 1992)
strikingly changed this image with their new estimates. They focused on the index
number problem, that is, the difficulty of constructing a plausible deflator due to a
temporal change of quantitative weight of each product, and they revised the
deflator for the period of industrial revolution by using a better method than that of
Dean and Cole. This revision resulted in drawing a new picture that the industrial
revolution was associated with lower economic performance than we had previously
believed. And, accordingly, it posed a meaningful question whether industrial
revolution was actually a revolution or not. After the revision by Crafts and Harley,
as the third generation, Broadberry et al. (2011) are revising the previous estimates
and constructing hyper-long-time series of GDP from 1270 to 1870.
Certainly, these efforts have improved the quality of estimates. According to
Broadberry et al. (2011), however, the method of calculation on the service sector
has just slightly changed since the attempt of Dean and Cole which was based on
wage bills or information of income tax for some bench mark years. Regarding the
period from 1700 to 1870, Broadberry et al. (2011) divided the service sector into 8
sub-sectors: commerce, government, housing, transport, distribution, finance and
other commerce. But, except government and housing, they seem to assume that
the growth of each sub-sectors is in line with other proxies such as output of other
sectors, population or employment. 3 It seems they have not found the way to
calculate directly the income or output of the service sector, yet.
Germany
Regarding the historical national account of Germany, Fremdling (1995) gives a
concise and detailed overview. For long years, Hoffman’s work (Hoffman 1965) has
been relied on and referred to, but recently its revision has started. 4 A
characteristic of the Hoffman’s estimation of NDP or NNP for 1850-1913 is that he
constructed the three types of figures using income, production and expenditure
Broadberry (2011) pp. 20-22. Using employment as proxy means assuming that
labour productivity does not change in time.
4 The works of Angus Maddison are also based on the Hoffman’s estimates (Maddison
2007).
3
4
approach. Hoffman applied income approach to not only the service sector but other
sectors as well (except for agriculture which was on the basis of output data).
According to Fremdling (1995), however, these three types of estimates are not
independent with each other and the output in the key-benchmark year (1913) is
almost determined by income approach.5 He explains Hoffman’s income approach
as follows:
6
1. For each sub-sector of the economy the number of employed people was assessed.
2. For the same sector the average yearly income per person was calculated. Thus
the labour income for the entire economy could be computed.
3. Capital income was mostly derived indirect by applying an average rate of return
on capital stock at current prices.
Eventually, Hoffman calculated NNP at factor cost by this methodology.
7
This calculation procedure was clearly derived from the formula of income approach
mentioned in the previous section. But Fremdling (1995) and Burhop and Wolff
(2005) criticize Hoffman’s estimates that the calculation of capital income is based
on some arbitrary assumptions and Burhop and Wolff (2005) tackle revisions in
terms of this point. But it seems nobody has yet to touch the revision of another
main component of the formula, compensation of employees.
India
For the historical national account of India, we just take a look at a massive work by
Sivasubramonian (Sivasubramonian 2000). He estimated national income and GDP
of India in the 19th century by sector. He divided the tertiary sector into 6
sub-sectors, that is, Railways and communications, Government services, Other
commerce and transport, Professions and liberal arts, Domestic service, and House
property, and applied an income approach to these sectors (and some small-scale
and cottage industries’ income was obtained by the income approach) except
Government activities and House property. He also conducted his calculation along
the formula of distribution side of GDP, and he assumed that Other commerce and
transport, Professions and liberal arts, and Domestic service did not earn any
For the production approach, Hoffman used NNP in 1913 as a bench mark and
deflated it by production indices of each sectors.
6 Fremdling (1995) p.85.
7 Although this does not seem to be clearly mentioned, Hoffman’s estimates included a
capital return from overseas investment.
5
5
return from their capital and, accordingly, consumption of fixed capital did not exist.
So, these three sectors’ incomes were obtained as the product of the number in the
working force and the average annual per head wage.
Taiwan
Mizoguchi ed. (2008) is provided the latest estimates of historical national account
of Taiwan. This book was published as one of the outcome of Asian Historical
Statistics Project (ASHSTAT project) organized by members of the Institute of
Economic Research at Hitotsubashi University. This project aims at constructing
historical national accounts of Asian countries. In the colonial period that is from
1895 to 1945, a number of statistical surveys including population or industrial
censuses were conducted in Taiwan by the colonial government. Despite this
relative abundance of data, Mizoguchi mentioned it was quite difficult to get
statistical information on the tertiary sector in the pre-war period.
Consequently, he applies income approach to calculate a traditional part of
this industry whose statistical information was quite scarce. His estimation
procedure on this industry is as follows:
1. For some sectors for which statistical data are relatively plentiful, production
approach is chosen for the calculation and the value added is computed using
information of an amount of sales or production and value added ratio as well as the
procedure of the primary and secondary sector. In some cases, the time-series
estimates are obtained by indirect calculation, that is, interpolation or extrapolation
using production index of agriculture and manufacturing (for wholesaling), foreign
trade index (for foreign trade merchants) or private consumption index (for
retailing).8
2. For the traditional sectors whose data are scarce such as restaurants, barbers
and other services or professions, income approach is applied using below formula:
The denominator is guessed from the information of input-output table constructed
in post-war Taiwan.
It seems, in most of the cases, estimates on 1939 are used as benchmark because
industry and commerce census was conducted in this year.
8
6
It is clear that this estimation was conducted using indirect information
and included some strong assumptions. But Mizoguchi also approached national
account from expenditure side (GDE) separately, and we can check the robustness of
the estimation in the aggregate.
Korea
According to Kim ed. (2006), value added of the service sector in pre-war Korea had
been calculated by the method including income approach which they thought had
poor reliability because the coverage of data they used was limited.9 Kim ed. (2006)
tackles to improve the quality of the estimates on this sector using production
approach as long as they could.10
For instance, to calculate the value added yielded in the retail and
wholesale sector they employed a kind of commodity flow approach; they estimated
total consumption in Korea using production and foreign trade data and divided it
into whole sale distribution value (intermediate input) and retail distribution value
(final demand) by commodity based on information derived from Korea’s
input/output table for 1966. Multiplying distribution margin calculated from the
data of wholesale and retail price to these values, they could get the total production
in the distribution sector consisted of margins of wholesale, retail, and
transportation.
They
divided
the
production
value
into
commerce
and
transportation based on the I/O table for 1966 and, finally, got the value added by
multiplying value added ratio which was also deprived from the I/O table.
This is innovative and ideal methodology to estimate the value added in the
tertiary sector, which also makes it possible to keep calculations on the production
side (GDP) and on the expenditure side (GDE) consistent because the procedure
they conducted is based on the commodity flow approach. That is to say, the private
consumption expenditure which is principal component of GDE is obtained as one of
by-products of this calculation.
11
However, this calculation is not yet perfect, of
course, because a great deal is depended on the information from the I/O table in
1966. This is a crucial point for evaluating their estimation whether they can justify
Following explanation is based on Japanese version of this book published in 2008.
Production approach is explained below fomulae:
Valueadded totaloutput intemediateinput
Valueadded totaloutput valueaddedratio
11 As a result of their calculation, a discrepancy between GDP and GDE became quite
small (2.4% on average)
9
10
7
their wide-range application of the I/O table.12
The case of Japan and its revision
From the survey of estimation methodologies, it is evident that the income approach
has been applied widely except for the case of Korea.
For estimating aggregate incomes of pre-war Japan, Kazushi Ohkawa and
others took an output approach for most production branches.13 As for agriculture,
forestry and fisheries, manufacturing, mining and construction, and also
communication and public utilities and much of transport, net domestic product
was estimated by applying value added ratios to gross output figures. So no LTES
estimates of primary and secondary incomes are affected by changes that our
sectoral estimates of labour force statistics which are outcome of INCHOS project.
As for commerce and services and also for most of transport,
communication and public utilities, however, the income approach was adopted. As
Ohkawa himself admits, the income approach is difficult to apply before the
emergence of well-articulated factor markets. Labour markets in the tertiary sector
of the Meiji period were never fully functioning. A vast majority of tertiary-sector
workers in the Meiji period were self-employed and many worked as subsidiary
workers whose principal occupations were in the primary and, to a lesser extent, the
secondary sectors. Thus Nobukiyo Takamatsu, who was responsible for much of
estimation of tertiary-sector output and incomes, applied the formula we have
already explained in the previous section to his estimation. Takamatsu aggregated
miscellaneous commerce and service sectors as one sector, called ‘Commerce and
services B’ included wholesale and retail trade, restaurant, finance, estate service,
entertainment service, and so forth. He divided it into 5 categories in terms of the
type of income: company, self-employed workers, employed workers, family workers,
and by-employment workers. He considered the net profit which is provided in tax
statistics is a return from capital and income of the self-employed should be mixed
income which includes the returns from capital and labour. He used the data of an
income survey in Tokyo in 1930 for acquiring these figures of income and some
strong assumptions were made to derive a wage or income series for tertiary-sector
Makino (2009) has criticised the adaption of post-war I/O table to pre-war period
because there is no reason that we can believe pre-war Korea had a similar economic
structure with that in 1966.
13 Ohkawa, Takamatsu and Yamamoto, Kokumin shotoku (National income). For an English
summary of the estimation procedures, see pp.159-173.
12
8
employments from that in manufacturing.
More problematic, perhaps, is to estimate the true numbers of workers in
commerce, services and transport. Takamatsu recognised the need to include
subsidiary workers whose principal occupations were in other sectors. In order to do
so, he decided to apply the following ratios of subsidiary to principal employments
in the sector of commerce and services to the four sub-periods:14
1885-1904: 0.3;
1905-1920: 0.24;
1921-1929: linear interpolation between 1920 and 1930; and
1930-1940: 0.108.
The ratios for the periods 1905-1920 and 1930-1940 come from sample tabulations
of the 1920 and 1930 national censuses, while that for the first sub-period is a mere
guesstimate.15 These ratios were applied to the number of principal workers in the
tertiary sector estimated separately by Umemrura (the series available was a
provisional one).16 The total number thus estimated stood at 640,000 in 1885, for
example. Given the paucity of data concerning actual working hours this group of
subsidiary workers spent, one half of the wage for employed workers estimated
above is used to calculate the incomes they generated.
There remain several problems, however. First, as figure 1 shows, the
time-series graph exhibits an irregular, zigzag pattern, a product of the assumption
of a fixed ratio applied for a given sub-period. Second, the estimation procedure does
not take into account any information about the size of the labour force in other
sectors, while the number of subsidiary workers whose principal occupation was in
other branches of the tertiary sector is included in the numerator of the calculation.
To put it differently, Takamatsu did not pay attention to the source of subsidiary
labour in tertiary-sector employment. Third, the ratio of 0.3 for the first sub-period
is an arbitrary one, but there is also an implicit assumption that the proportion of
those engaged in tertiary activities as by-employments declined linearly over time.
This may or may not be true for the period in question, and is a question to be
14
Note that for some unknown reason, Takamatsu did not include transport in his estimation of
subsidiary workers.
15 Naikaku Tōkei-kyoku, Chūshutsu hōhō ni yoru daiikkai kokusei chōsa no gaiyō (Tokyo: Naikaku
Tōkei-kyoku, 1924), and Chūshutsu hōhō ni yoru Shōwa 5-nen kokusei chōsa no gaiyō (Tokyo:
Naikaku Tōkei-kyoku, 1932).
16 Umemura’s provisional estimates are published in ‘Population and labor force’ in Patterns,
pp.392-395, although no breakdown figures are set out for the period before 1905.
9
settled empirically. Finally, it is probably worth pointing out that Takamatsu could
not take into account the revisions Umemura produced later for his sectoral
estimates of principal workers. 17 As far as the tertiary sector is concerned,
Umemura’s new series gives somewhat lower estimates for earlier years and higher
estimates for later years. Using the revised Umemura estimates, we calculated the
number of by-employed workers applying the methodology which was conducted in
Saito and Settsu (2010): the calculation based on matrix-formatted data of primary
and subsidiary occupations which provides the information how many labour force
working in the primary and secondary sectors engaged in the tertiary sector as
by-employment;18 this revised Takamatsu series of subsidiary workers in commerce,
services and transport is shown in table 1.
Unlike the postulated Takamatsu series, this paper's new series of the
number of persons having a subsidiary occupation in the tertiary sector, estimated
above, no longer looks like a zigzag line (See figure 1). Although its long-run
tendency is not entirely different from the Takamatsu estimates, this new series
enables us, by applying exactly the same method Takamatsu employed, to estimate
the incomes they earned and the total NDP of the tertiary sector in corresponding
years.
According to our revised estimates (set out in table 1), the number of
subsidiary workers in the tertiary sector was 720,000 in 1885, 14 per cent larger
than the Takamatsu estimate (it would be 22 per cent larger than that implied by
the revised Umemura series). Rather unexpectedly, however, the largest difference
between the old and new series is found for 1920. It is evident that Takamatsu
overstated the number substantially for this census year. We can see the same
difference in the estimated NDP between the one based on the Takamatsu
estimates and our new estimates of the number of by-employed workers.
Despite these non-negligible differences in the numbers of subsidiary
workers and the implied incomes earned by them, the two output estimates appear
similar. The largest gap is found for the year 1920 but is only by 2 per cent. This is
because wage earnings by subsidiary labour are assumed to have been half the
amount earned by the principal worker, so that the subsidiary wage portion in the
sectoral NDP could not be large. The Takamatsu estimates were not wide of the
mark.
17
18
Umemura et al.(1988).
See Saito and Settsu (2010).
10
Concluding remarks
Although our new estimates did not succeed in changing the picture depicted
by Takamatsu and Umemura, there can be no doubt that whether or not we take into
account the problem of by-employment is a critical issue, especially when we estimate
the value added using the income approach. As we took a look in the previous section,
however, existing estimates which employed the income approach did not give attention
to this issue.
One of the reasons of their ignorance is certainly scarcity of data on
by-employment. However, it is likely to come across the data we need if we check
through an archive carefully again. Actually, though our INCHOS project, we found the
matrix-form data on by-employment for pre-war Germany and India. Moreover, it has
been known that there are some data on by-employment in population censuses in
pre-war Taiwan and Korea though they are not provided as matrix format. Of course,
we know there is no panacea for the difficulties of estimation of historical national
account; however, if we use these data as an experimental light, we might be able to
know the contents hidden in the black box.
11
Figure 1. The number of by‐employment workers
900
800
700
thousand parsons
600
500
Our estimates
LTES1
400
300
200
100
0
1885 1888 1891 1894 1897 1900 1903 1906 1909 1912 1915 1918 1921 1924 1927 1930 1933 1936 1939
12
Table 1. The estimated numbers of subsidiary workers in commerce, services and
transport: LTES and our estimates compared, 1885-1940
Year
(1)
(2)
(3)
(4)
LTES1
LTES2
1885
636,300
591,300
724,341
14%
22%
1890
669,900
622,800
708,475
6%
14%
1900
721,200
685,800
692,372
-4%
1%
1910
678,240
624,480
657,647
-3%
5%
1920
789,360
811,200
595,055
-25%
-27%
1930
448,632
532,440
548,268
22%
3%
1940
535,248
551,880
540,384
1%
-2%
Our estimates [(3)-(1)]/(1)
(5)
[(3)-(2)]/(2)
Sources: See text
Note 1) LTES 1 are the Takamatsu estimates used in Ohkawa et al., Kokumin
shotoku, while LTES 2 are those which should have been if based on Umemura's
revised estimates in Umemura et al., Rōdōryoku.
2) Transport and communications are not included in the two LTES series.
13
Data sources
Dajōkan Tōkei-in (1882) Kainokuni genzai ninbetsu shirabe.
Naikaku Tōkei-kyoku (1924) Chūshutsu hōhō ni yoru daiikkai kokusei chōsa no gaiyō,
Naikaku Tōkei-kyoku.
Shizuoka-ken Chiji Kanbō Tōkei-ka (1928) Taishō juyonen kokuseichōsa narabini
shokugyō chōsa kekka hōkoku.
References
Broadberry, Stephen, Bruce Campbell, Alexander Klein, Mark Overton, Bas van
Leeuwen (2011) ‘British Economic Growth, 1270-1870’, London School of Economics.
http://www2.lse.ac.uk/economicHistory/seminars/ModernAndComparative/papers201
1-12/Papers/Broadberry.pdf
Burhop, Carsten and Guntram B. Wolff (2005) ‘A Compromise Estimate of German Net
National Product, 1851-1913, and Its Implications for Growth and Business Cycles’
The Journal of Economic History, Vol. 65, No. 3,pp. 613-657.
Crafts, N. F. R. (1985) British Economic Growth during the Industrial Revolution,
Oxford University Press.
Crafts, N. F. R. and C. K. Harley (1992) ‘Output growth and the British industrial
revolution: a reassessment of the Crafts-Harley view’, Economic History Review, 2nd
ser. Vol. 45, no.4. pp. 703-730.
Dean, Phyllis and W. A. Cole (1962) British Economic Growth 1688-1959, Cambridge
University Press.
Fremdling, Rainer (1995) ‘German National Accounts for the 19th and Early 20th
Century’, Scandinavian Economic History Review, vol.XLIII. no.1, pp. 77-100.
Hoffman, W.G. (1965) Das Wachstum der Deutschen Wirtshaft seit der Mitte des 19.
Jahrhunderts, Springer-Verlag.
Kim, Nak Nyon ed. (2006) 한국의 경제성장 1910-1945 (Economic Growth in Korea
1910-1945), Korea National University Press (translated into Japanese by Ho Il Moon
and Seung Mi Kim in 2008, Japanese title is Shokuminchi ki chōsen no
kokuminkeizaikeisan 1910-1945, Tokyo daigaku shuppankai).
Maddison, Angus (2007) The World Economy: A Millennial Perspective/ Historical
Statistics, OECD publishing.
Makino, Fumio (2009) ‘Shohyō Kim Nak Nyon hen (Moon Ho Il・Kim Seung Mi yaku,
Odaka Konosuke ・ Saito Osamu yakubuni kanshu)Shokuminchi ki chōsen no
14
kokuminkeizaikeisan 1910-1945’ (Book review: Kim, Nak Nyon ed. (translated by Ho
Il Moon and Seung Mi Kim under supervision of Konosuke Okdaka and Osamu Saito),
Shakai Keizai Shigaku (Socio-Economic History), vol.75, no.1, pp.95-97
Mizoguchi, Toshiyuki (2008) Asia chōki keizai tōkei: Taiwan (Asian Historical Statistics:
Taiwan), Toyo Keizai Shimposha.
Ohkawa, Kazushi, Nobukiyo Takamatsu and Yuzo Yamamoto (1974) Kokumin shotoku
(National income), LTES 1, Toyo Keizai Shimposha,
Saito, Osamu and Tokihiko Settsu (2010) ‘Unveiling Historical Occupational Structures
and its Implications for Sectoral Labour Productivity Analysis in Japan's Economic
Growth’, Global COE Hi-Stat Discussion Paper series No. 143
Settsu, Tokihiko (2009) ‘Kindai Nihon no keizai seichō to shōgyō saabisu-gyō (Economic
growth and the tertiary sector in modern Japan)’, PhD dissertation, Hitotsubashi
University.
Sivasubramonian, S. (2000) The National Income of India in the Twentieth Century,
Oxford University Press.
Umemura, Mataji (1969) ‘Agriculture and labor supply in the Meiji era’, in Kazushi
Ohkawa et al., eds., Agriculture and economic growth: Japan’s experience, University
of Tokyo Press, pp.175-197.
Umemura, Mataji et al. (1988) Rōdōryoku (Manpower), LTES 2, Toyo Keizai
Shimposha.
van Ark, Bart (1995) ‘Towards European Historical National Accounts’, Scandinavian
Economic History Review, vol.XLIII. no.1, pp. 3-16.
15