The Sources of Long-term Economic Growth for Turkey, 1880-2005

Abstract

This article considers the sources of long-term economic growth for Turkey over the period 1880–2005. The period in question covers the decline and eventual dissolution of the former Ottoman Empire and the emergence of the new Turkish Republic in 1923. Hence, the article provides a unique look at the growth experience of these two different political and economic regimes. The article examines in detail the evolution of factors that led to growth in output across broad periods, including the post-World War II period and the era of globalization beginning in the 1980s. It also considers output growth in the agricultural and non-agricultural sectors separately and allows for the effects of sectoral re-allocation. The lessons from this exercise have important implications for Turkey's future economic performance, for its ability to converge to per capita income levels of developed countries, and for the viability of its current bid for European Union membership.

Full text

DISCUSSION PAPER SERIES
ABCD
www.cepr.org
Available online at:
www.cepr.org/pubs/dps/DP6463.asp
www.ssrn.com/xxx/xxx/xxx
No. 6463
THE SOURCES OF LONG-TERM
ECONOMIC GROWTH FOR TURKEY,
1880-2005
Sumru G. Altuğ, Alpay Filiztekin
and Şevket Pamuk
INTERNATIONAL MACROECONOMICS

ISSN 0265-8003
THE SOURCES OF LONG-TERM
ECONOMIC GROWTH FOR TURKEY,
1880-2005
Sumru G. Altuğ, Koc University and CEPR
Alpay Filiztekin, Sabanci University
Şevket Pamuk, Bogazici University
Discussion Paper No. 6463
September 2007
Centre for Economic Policy Research
90–98 Goswell Rd, London EC1V 7RR, UK
Tel: (44 20) 7878 2900, Fax: (44 20) 7878 2999
Email: cepr@cepr.org, Website: www.cepr.org
This Discussion Paper is issued under the auspices of the Centre’s research
programme in INTERNATIONAL MACROECONOMICS. Any opinions
expressed here are those of the author(s) and not those of the Centre for
Economic Policy Research. Research disseminated by CEPR may include
views on policy, but the Centre itself takes no institutional policy positions.
The Centre for Economic Policy Research was established in 1983 as a
private educational charity, to promote independent analysis and public
discussion of open economies and the relations among them. It is pluralist
and non-partisan, bringing economic research to bear on the analysis of
medium- and long-run policy questions. Institutional (core) finance for the
Centre has been provided through major grants from the Economic and
Social Research Council, under which an ESRC Resource Centre operates
within CEPR; the Esmée Fairbairn Charitable Trust; and the Bank of
England. These organizations do not give prior review to the Centre’s
publications, nor do they necessarily endorse the views expressed therein.
These Discussion Papers often represent preliminary or incomplete work,
circulated to encourage discussion and comment. Citation and use of such a
paper should take account of its provisional character.
Copyright: Sumru G. Altuğ, Alpay Filiztekin and Şevket Pamuk

CEPR Discussion Paper No. 6463
September 2007
ABSTRACT
The Sources of Long-term Economic Growth for Turkey, 1880-
2005*
This paper considers the sources of long-term economic growth for Turkey
over the period 1880-2005. The period in question covers the decline and
eventual dissolution of the former Ottoman Empire and the emergence of the
new Turkish Republic in 1923. Hence, the paper provides a unique look at the
growth experience of these two different political and economic regimes. The
paper examines in detail the evolution of factors that led to growth in output
across broad periods, including the post WWII period and the era or
globalization beginning in the 1980's. It also considers output growth in the
agricultural and non-agricultural sectors separately and allows for the effects
of sectoral re-allocation. The lessons from this exercise have important
implications for Turkey's future economic performance, for its ability to
converge to per capita income levels of developed countries, and for the
viability of its current bid for European Union membership.
JEL Classification: E60, N15, O40, O50 and O57
Keywords: determinants of growth, growth accounting and sectoral re-
allocation
Sumru G. Altuğ
College of Administrative Sciences &
Economics
Koç University
Rumelifeneri Yolu, Sariyer
34450 Istanbul
TURKEY
Email: saltug@ku.edu.tr
For further Discussion Papers by this author see:
www.cepr.org/pubs/new-dps/dplist.asp?authorid=132739
Alpay Filiztekin
Faculty of Arts and Social Sciences
Sabanci University
Orhanli 34956 Tuzla
Istanbul
TURKEY
Email: alpayf@sabanciuniv.edu
For further Discussion Papers by this author see:
www.cepr.org/pubs/new-dps/dplist.asp?authorid=132421

Şevket Pamuk
Economics and Economic History
The Ataturk Institute for Modern
Turkish History
Bogaziçi University
34342 Bebek-Istanbul
TURKEY
Email: pamuks@ttnet.net.tr
For further Discussion Papers by this author see:
www.cepr.org/pubs/new-dps/dplist.asp?authorid=162631
*An earlier version of this paper was presented at the ICE-TEA International
Economics Conference, September 11-13, 2006, Ankara, Turkey; the
Conference on Unifying European Growth and Integration since the Mid-
Nineteenth Century, October 13-15, Lund, Sweden, and the ICCEES Regional
European Congress, August 2-4, 2007, Berlin, Germany.
Submitted 24 August 2007

2
1 Introduction
The determinants of growth and of the distribution of income across countries have
been the focus of much debate in the recent literature. The preferred method of
analysis has been cross-country regressions which use information on individual
countries over different time periods. See, for example, Barro (1991). However,
cross-country growth regressions have come under criticism for failing to account
for the diversity of experiences of the individual countries and also because there
is considerably uncertainty regarding the appropriate specification or the set of
variables that should be considered.
1
A different approach is to examine the record
of specific countries over longer periods in the light of historical episodes, political
eve nts, different policy regimes, and the like. Lains (2003) and Prados de la Escosura
and Roses (2005) conduct such studies that detail growth and catch-up for the case
of Portugal and Spain, respectively. There also exist recent studies that seek to
identify common causes behind the growth performance of a group of countries that
share similar geographies, historical developments, and policy experiences. See, for
example, Ze ttelmeyer (2006) or Cole et al (2004) on the Latin American countries.
In this paper, we examine the determinants of long-term economic growth for
Turkey over the period 1880-2005. The growth experience of Turkey takes place
against a backdrop of a wide-ranging set of historical and political events and en-
compasses different policy regimes. In contrast to the countries of the New World
which we re built by European settlers, the modern state of Turkey emerged from
the ashes of the former Ottoman Empire. Up until its disintegration, this was a
far-flung entity that had its own elites and a centralized system of government. The
period culminating in the creation of the Turkish state was characterized by regional
and international conflict, on the one hand, and attempts at reform, on the other.
Yet after its creation, Turkey was much like many other developing countries during
the post World War I and World War II periods. In many ways, Turkey’s growth ex-
perience reflects the role of international factors that also governed growth in other
1
See the discussion in Levine and Renelt (1992) or Brock, Durlauf and West (2003).

3
devel oping countries. During the post World War II period, Turkey grew rapidly
alongside other countries of its size and income and its adherence to state-led growth
or import-substituting industrialization was not without precedent. Similarly, the
era of globalization beginning in the 1980’s was witness to recurring episodes of
financial and macroeconomic instability and crises in a number of developing coun-
tries such as Mexico and Argentina in 1994-1995, the East Asian countries in 1998,
Russia in 1999 – as they were in Turkey in 1994, 1999 and, more severely, during
2000-2001. Despite the influence of such phenomena, some countries were able to
display very high rates of growth and to catch up to developed country levels while
other were not. To date, Turkey has not featured among the “miracle economies”
though this prognosis may change in the future. Our interest in studying the Turk-
ish experience is first, to understand Turkey’s absolute and relative position in the
cross-country distribution of growth and convergence. Second, we seek to identify
the role of factor accumulation, including human capital accumulation, versus a
more broadly defined measure of technological progress in accounting for Turkey’s
growth experience across different historical periods and alternative policy regimes.
Our analysis proceeds as follows. In Section 2, we discuss Turkey’s growth perfor-
mance in absolute and comparative terms in the period from 1880 using purchasing-
power adjusted per capita GDP data compiled by Maddison (2001, 2003). We
consider Turkey’s performance relative to developed countries and also develop-
ing countries, excluding China. Our results indicate that the income gap between
Turkey and the present-day developed countries widened considerably during the
nineteenth century, and that this gap has stayed roughly unchanged since World
War I. In comparison to the developing countries as a whole, Turkey grew faster than
the developing country averages from the nineteenth century until the 1970’s. How-
eve r, it has been lagging behind the developing country averages since the 1980’s.
Another way of describing this phenomenon is to note that the distribution of per
capita GDP across countries has evolved towards a twin-peaked representation dur-
ing this period, and Turkey has only managed to maintain its relative position in

4
these distributions irrespec tive of initial conditions.
Next, we employ a growth accounting approach to decompose output growth
into growth in the factors of production versus total factor productivity (TFP). This
discussion takes places in Section 3. There exist a number of studies that calculate
TFP growth for the Turkish economy for the post 1960’s period and examine its
evol ution for the aggregate economy and on a sectoral basis – see, for example,
Filiztekin (2000), Saygılı , Cihan, and Yurto˘glu (2001, 2005), Altu˘g and Filiztekin
(2006), or Ismihan and Metin-
¨
Ozcan (2006). In contrast to these papers, we also
model the agricultural sector using a production function approach with inputs of
land, labor and capital. Arguably one of the most important transformations that
took place in the modern Turkish era was the increase in arable lands together with
the mechanization of agriculture that began in the 1950’s. (See Pamuk, 2007b.) To
quantify this transformation and to examine its implications for Turkish productivity
growth, we consider a two-sector model with an agricultural and non-agricultural
sector that also allows for the impact of human capital. Another novel feature of
our analysis is that we make use of a new capital stock series dating back to 1923.
We are not aware of any studies that use such a long capital stock series or that
allow for an explicit role for agriculture. Finally, we conduct a labor productivity
decomposition, whereby we examine the impact on total productivity of increases
in productivity within a sector and also the re-alloc ation of factors across sectors.
These features are similar to the analysis in a recent paper by Bosworth, Collins and
Virmani (2007) and Bosworth and Collins (2007), who examine the determinants of
long-run growth for India and provide a comparison of China and India, respectively.
Sever al recent papers have argued that the growth experience of various Latin
American countries or Turkey can be understood in the context of specific micro-
founded explanations such as barriers to competition (Cole et al, 2004 in the case of
the Latin Americans) or tax differences across sectors in an otherwise standard neo-
classical framework (see Adamopoulos and Akyol, 2006 for the case of Turkey). By
contrast, Zettelmeyer (2006) has examined the role of reforms that support positive

5
macroeconomic policy outcomes for Latin America. Taking a wider perspective, the
growth literature in recent years has concentrated on such factors as the role of insti-
tutions (see Hall and Jones, 1999 or Acemo˘glu Johnson, and Robinson, 2001) or the
impact of human capital (see Glaeser et al, 2004). Yet as Rajan and Zingales (2006)
argue, there may be a third factor, which is the proximate cause of both. In their
framework, the distribution of initial factor endowments leads to self-interested con-
stituencies who perpetuate the status quo. Much of the debate surrounding the role
of macro-policy-making has revolved around the notion that poor policy outcomes
are a major source of instability and low growth for developing countries. Yet one
could also argue that weak institutions (or self-interested constituencies) lead to po-
litical instability, which also leads to a poor macroeconomic outlook. However, even
after controlling for the impact of institutions, Sirimaneetham and Temple (2006)
have shown that macroeconomic policy-making may matter for the distribution of
growth rates.
In this paper, we examine separately the role of institutions, human capital,
and macroeconomic policy-making in determining Turkey’s growth experience. The
institutional environment matters for growth. If existing institutions favor diver-
sionary activi ties, then individuals will not prefer to accumulate human capital and
education and firms will not prefer to invest in productive capital. Both factors
will deter growth. In the era of globalization, lower human capital will impede the
diffusion of technology and hinder a society’s ability to catch up to income levels in
other rapidly developing parts of the world. The evolution of economic institutions
in Turkey and their consequences for economic growth and distribution of income
have not been closely studied. Nevertheless, one could argue that economic and
political power have taken a long time to reach broad segments of the p opulation.
Too often during the last half century, Turkey’s political system has produced fragile
coalitions and weak governments which have sought to satisfy the short-term de-
mands of various groups by resorting to budget deficits, borrowing and inflationary
finance. There has been a weak democratic regime that has alternated with military

6
regimes in Turkey since 1960. These facts point to the importance of examining the
institutional framework underlying Turkey’s growth experience.
Human capital considerations also appear critical at key junctures of Turkey’s
growth experience. At the inception of the Turkish Republic in 1923, the percentage
of the literate population was exceeding low. While official estimates are not avail-
able for adult literacy in the early years of the Republic, it can be safely assumed
that the rate did not exceed 10% before World War I and in the 1920s. This fact
itself may also help to explain why, despite a broad-reaching set of military, political
and economic reforms, the Ottoman Empire was not able to transform itself into a
modern state. Educational policies in Turkey in the last 80+ years have succeeded
in bringing the literacy rate to less than 89% by 2005. We can also provide a com-
parative perspective on this issue. Since 1913 and 1950 education levels i n Turkey
as measured by literacy, years of schooling and school enrollment have been lagging
significantly behind education levels in developing countries with similar levels of
GDP. Turkey lags behind Latin America and well behind East Asia in the educa-
tion variables, holding level of GDP constant. Along with other Muslim majority
countries, Turkey also lags behind developing countries with comparable levels of
per capita income in indices aiming to measure gender equality and socio-economic
devel opment of women.
The role of macroeconomic policy-making features in most analyses involving
countries’ growth experiences. In their sensitivity analysis, Levine and Renelt (1992)
are unable to discern a statistical relationship between long-run growth and any sin-
gle macroeconomic variable. Nevertheless they argue for the importance of national
policies for determining countries’ growth experiences. Turkey’s recent macroe co-
nomic history has been far from exemplary. Average inflation in the 1980’s ranged
around 50%, increasing to over 70% in the 1990’s. Throughout much of this period,
real interest rates in Turkey have traversed in the 20-25% range. Turkish GDP dur-
ing the 1990’s also shows a highly volatile path, declining by 6% in 1994 and by 5%
again in 1999 as a result of the Marmara earthquake. Turkey’s GDP registered one

7
of its largest declines in Republican history of 10% during the banking and finan-
cial crisis of 2000-2001 that erupted in the midst of an IMF-sponsored stabilization
plan. Turkey has been experiencing a remarkable recovery in per capita GDP levels
since the latest economic crisis in 2001. Per capita GDP levels have increased by
about 20% during these five years. Nevertheless, among the developing countries,
Turkey has the highest current account deficit to GDP ratio (7.5%) and the second
highest inflation (around 7%). It suffers from the phenomenon of “jobless growth”
and remains among the most v ulnerable to changes in international capital market
conditions.
With this introduction, we describe how the remainder of this paper is orga-
nized. In Section 2 below, we present an overview of Turkey’s growth record in
both absolute and comparative terms. In Section 3, we conduct a growth account-
ing exercise over the period 1880-2005 including a discussion of the contribution of
sectoral re-allocation from agriculture to the urban sector. Section 4 presents a set
of puzzles regarding Turkey’s growth experience and seeks to resolve them before
the conclusion in Section 5.
2 Turkey’s Economic Performance Since 1880
In this section we present an overv iew of Turkey’s growth record since the nineteenth
century in both absolute and comparative terms. Population of the area within
present day borders of Turkey increased from a little over 9 million in 1820 and
13 million in 1880 to 73 million in the year 2005. Per capita GDP increased from
about 680 purchasing power parity adjusted 1990 international dollars in 1820 to
880 dollars in 1880 and to 7500 dollars in 2005, an 11 fold increase for 1820-2005
and a 8 fold increase since 1880.
2
Since per capita incomes increased in most if
2
Per capita GDP in constant US dollars is the basic indicator for examining long term increases
in average incomes. The GDP series that we use in this section are calculated with a purchasing
p ower parity adjustment to take into account the fact that price levels tend to be lower and the
same dollar income purchases more in lower income countries.

8
1880-1913: Ottoman er a to World War I
1913-1950: interwar era
- 1913-1929: post WWI recovery ends and a new policy era begins
- 1930-1950: early Republic including the Great Depression and WWII
1950-1980: post-WWII era under import substituting industrialization
1980-2005: era of globalization since 1980
Table 1: Periodization
not all areas of the world during the last two centuries, it is necessary to examine
Turkey’s growth record relative to other countries. Thanks to the efforts of many
economic historians, most notably the work of Angus Maddison in recent decades,
we are able to compare below the growth record of Turkey with the population
weighted averages for the high income countries of western Europe and the United
States as wel l as the developing country averages as defined by Asia excluding Japan,
Africa and Latin America. We will also compare Turkey’s record with those of some
of the individual countries whose case histories are better known. Following the
established pattern in such long-term comparisons, we will divide this period into
four sub-periods, nineteenth century until World War I, the interwar era until 1950,
the post-World War II era until 1973 or 1980 and the current era of globalization
since. (See Table 1.)
The first period covers Turkey during the Ottoman era. During the decades and
the century before World War I, the areas comprising modern Turkey experienced
positive but modest levels economic growth at less than 1% per annum. These
increases in GDP per capita took place within the global context of open trade and
financial regimes and were led by the agricultural sector, more specifically by export-
oriented agriculture. Nonetheless, the gap in per capita incomes between Turkey and
the Ottoman Empire, on the one hand, western Europe and the United States, on
the other, widened considerably during the century before World War I, due to the
rapid rates of industrialization in the latter. The GDP per capita series constructed

9
by Maddison (2001, 2003) make clear that the gap between the developing countries
and the developed countries widened even more during the nineteenth century. GDP
per capita in the area within present-day borders of Turkey as a percent of the GDP
per capita in the high income countries of Western Europe and the United States,
calculated on a population-weighted basis declined from about 55% in 1820 and
about 41% in 1870 to 29% in 1913. (See Table 2.) GDP per capita in Turkey stood
at approximately 1200 purchasing power parity adjusted 1990 US dollars in 1913.
In the same year, GDP per capita of the areas comprising modern Turkey stood at
168% of the GDP per capita income in the developing countries of Asia, Africa and
Latin America, calculated on a population-weighted basis. (Maddison, 2001, 2003
and Pamuk, 2006.)
Our second period covers the interwar era until the end of World War II. This
period witnessed two worl d wars and a great depression. The Ottoman Empire
disintegrated at the end of World War I and modern Turkey was established within
a much smaller territory. In the areas covered by modern Turkey, World War I and
the War of Independence in 1919-1922 led to large population losses, approximately
18%, and large declines in GDP per capita, as much as 40% or more, followed by
a rapid recovery in the 1920s. To follow the large swings in per capita and total
GDP during this difficult period, we have decided to divide the second sub-period
into two, 1913 to 1929 and 1930 to 1950. We decided to choose 1929 for the end
of this second period as this was the year when pre-War levels of per capita GDP
was attained for the first time after WWI. By contrast 1930 marks the beginning
of a new policy era as the Great Depression ushered in new economic policies in
Turkey, protectionism and inward-oriented industrialization led by the state sector.
Turkey experienced high rates of growth during the 1930’s but the favorable trends
were reversed by the outbreak of World War II. We chose 1950 to end this second
sub-period and begin the next as this i s the benchmark year in many studies and
also because the Turkish post-World War II recovery was completed by this date.
The period after World War II was a period of high rates of growth around the

10
world. After a brief experiment with agriculture-led growth in the 1950s, Turkey
settled once again on import-substituting industrialization (ISI), this time led by
the private sector. Exports of manufactures remained low during this period and
the reasonably high rates of economic growth, exceeding 3% per annum for GDP per
capita, were led by domestic market-oriented industrialization during these decades
(see Table 2). The era of globalization arrived i n Turkey after a severe and prolonged
foreign exchange crisis at the end of the 1970s and with the adoption of a new and
liberal policy package in 1980. Rapid growth of exports of manufactures, on the one
hand, and macroeconomic instability, on the other, have been the most important
features of this fourth and last sub-period.
Per capita income in Turkey in 1950 had been at 1620 purchasing power parity
adjusted, 1990 US dollars. This was equal to 24% of the per capita income of the
high income countries and 188% of the per capita income in the developing countries.
By 2005, GDP per capita in Turkey had reached 7500 dollars, an increase of more
than five-fold since 1913. This figure corresponded to about 30% of the level of GDP
per capita in the high income countries of Western Europe and the United States,
and approximately 225% of the GDP per capita of the developing countries for the
same year. In other words, average incomes in Turkey have increased at about the
same rate as those in high income countries since 1913 and somewhat higher than
the rates experienced by these countries since 1950. As a consequence, Turkey has
not been able to close during the twentieth century the large gap that opened up
between it and the developed countries during the nineteenth century.
In Figure 1 we provide per c apita GDP series for Turkey and a number of other
regions and countries as percentages of the average for Western Europe and the
United States for the period since 1913. This figure allows further insights into
Turkey’s comparative economic record in the twentieth century. While Turkey’s
growth record is better than the averages for the developing countries as a whole
(see Table 2), since 1950 Turkey has lagged well behind the countries with well-
known episodes of “economic miracle” in Southern Europe and East Asia. This

11
0 20 40 60 80
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
Turkey South Europe
Korea Latin America
GDP Per Capita as percent of (US+W.Europe)
Figure 1: GDP Per Capita Relative to W. Europe and US
finding is also illustrated in Figure 2, which shows average annual per capita real
GDP growth over the period 1960-2000 as a function of initial income per capita
in 1960. Turkey’s average GDP growth clearly is slower than the GDP growth of
countries such as S. Korea, Thailand or Malaysia which had comparable pe r capita
levels of income in 1960. A similar observation can be made for the “late-starters”
in Europe. This phenomenon is also noted by Adamopoulos and Akyol (2006). We
return to some of their findings at a later point.
In Figure 3, we summarize the information about convergence in terms of the
distribution of per capita GDP levels for the developing countries for 1960 and 2000,
respectively, using measures of chained weighted GDP per worker and GDP per
capita obtained from the Penn World Tables. These distributions begin to display
a marked twin-peaked character by 2000, a phenomenon which has been noted in
the empirical growth literature.
3
While there is a single distribution describing per
capita GDP levels in 1960 (albeit one with a long upper tail), a large probability
3
See, for example, Quah (1996).

12
IND
IDN
THA
TUR
BRA
MEX
ARG
MYS
KOR
−.02 0 .02 .04 .06
Average Annual Growth Rate
6 7 8 9
Log of Per capita Real GDP in 1960
Per Capita Real GDP Growth
Figure 2: Average Annual Growth Rates as a Function of Per Capita Income in
1960
mass of this distribution has shifted to the right by 2000, corresponding to the group
of countries who have managed to forge ahead in their growth experiences during the
forty year period between 1960 and 2000. To get more insight into Turkey’s relative
performance, we note that Turkey has only managed to maintain its relative position
in these distributions over time irrespective of its initial conditions.
It is also possible to discuss long-term trends in structural change and the shift
of labor from agric ultural to the urban sector. Urbanization and structural change
did not gain momentum in Turkey until after World War II. Statistics in this area
are not precise but it appears that share of urban in total population increased
slowly during the nineteenth century and was about 28% on the eve of World War
I. In 1950 the same share was lower at around 25% or even less. Share of agriculture
in total employment was about 80% in 1880 and as late as in 1950.
4
Eve n after
4
The inclusion of women working in the f amily farm in the labor force but the exclusion of urban
women working at home from the labor force tends to overstate the share of agriculture in Turkish
employment statistics.

13
Turkey
Turkey
0 .1 .2 .3 .4 .5
7 8 9 10 11
Note: Solid line indicates 1960 distribution and broken line indicates 2000 distribution.
Per Worker
Logarithms of Real GDP
Turkey
Turkey
0 .1 .2 .3 .4 .5
6 7 8 9 10
Note: Solid line indicates 1960 distribution and broken line indicates 2000 distribution.
Per Capita (Chain Index)
Logarithms of Real GDP
Figure 3: Distribution of Per Worker and Per Capita GDP, 1960 and 2000. Source:
Penn World Tables
World War II, the shift to the urban sector was not very rapid. Numbers of people
employed in the agricultural sector continued to increase, albeit gradually, until the
1980s and began to decline in absolute terms only in the 1990s. Share of agriculture
in the labor force stood at 35% in the year 2000. Share of agriculture in GDP has
declined faster, from about 64% in 1880 to 53% in 1950 to 13% in 2000. Share of
industry has increased from about 13% in 1913 to 26% in 2005 while the share of
services has increased from 34 to 64% during the same period.
In Table 3, we present further information on the growth rates of output and
the factor inputs across the different per iods in our study. One of the significant
findings is the sharp break between the pre- and post-1950 perio ds. The growth
rates of output and of the factor inputs such as labor, capital, land and human
capital are all significantly greater in the post-1950 period. We can explain this
pattern with two observations. First, rates of per capita growth accelerated sharply
after World War II in all regions of the world. Second, in the Turkish case, there
was a very large decline in population (approximately 20%) during and after World
War I and even larger dec lines in GDP and GDP per capita (more than 40%) until
the early 1920’s. Even though by 1929 per capita GDP had returned to pre-World
War I levels, growth rates for output, labor and most other variables were negative

14
for the period 1913-1929 as a whole. As a consequence, GDP per capita growth
rates for Turkey for sub-periods until WWI I or until 1950 were all below 1% per
annum.
Rates of growth of labor and especially physical and human capital were sig-
nificantly higher since 1950. Indeed one of the distinguishing features separating
the pre- and post-1950 periods is the more than doubling of the growth rates of
output and capital. There are also appears to a significant increase in the rate of
human capital during the 1930-1950 periods. However, both the rates of physical
and human capital accumulation show a marked decline after 1980. Considering the
different sectors separately, we find that agricultural output increased most rapidly
in the period 1950-1980. The reason is that the cultivation of arable lands expanded
rapidly and reached its peak during this per iod. The land under c ultivation starts
to decline after 1980 as industry, tourism and residential housing begin to claim
some of the agricultural land. Likewise, accumulation of physical capital in agricul-
ture was the greatest during the 1950-1980 period. The non-agricultural sector also
shows the greatest growth during the 1950-1980 period: output averages 6.15 % and
capital accumulation proceeds at even a faster rate of 6.61%. As Turkey’s economy
is opened up to the rest of the world beginning in 1980, we see lower growth in
non-agricultural activities as well as lower rates of capital accumulation.
These observations already provide important clues regarding the sources of
growth for the Turkish economy since 1880. However, they do not isolate the impact
of technological progress or total factor productivity from the remaining sources of
growth. This is critical since most new growth theories, whether they be based
on endogenous growth or emphasize the role of institutions, assign a key role to
productivity or TFP growth. We now turn to this issue.

15
3 Growth Accounting
Growth accounting provides a convenient approach for analyzing the sources of
output growth in an economy. It has featured widely in the growth literature as a
way of examining the contribution to growth of the factor inputs versus total factor
productivity. In Solow’s (1958) original contribution, only the inputs of labor and
capital were considered. Subsequently, the role of human capital accumulation on
output growth was recognized by a variety of authors. (See, for example, Mankiw,
Romer, and Weil, 1992 or Collins and Bosworth, 1996.) In this study, we consider
a two-sector economy and allow for the impact of human capital growth on output
growth.
Suppose that the economy has two sectors, an agricultural sector (sector a) and
a non-agricultural sector (sector n). Let Y
it
denote the output produced in sector
i = a, n at date t. We differentiate between the production technologies in the
agricultural versus non-agricultural sectors by making the former depend on the
factor of production land in addition to the factors of labor and physical and human
capital. The production function in the agricultural sector and non-agricultural
sectors ar e given by:
Y
at
= A
at
K
α
a
at
N
β
a
at
L
γ
at
H
1−α
a
−β
a
−γ
at
, (1)
and
Y
nt
= A
nt
K
α
n
nt
N
β
n
nt
H
1−α
n
−β
n
nt
, (2)
where 0 ≤ α
i
≤ 1, 0 ≤ β
i
≤ 1 for i = a, n, α
a
+ β
a
+ γ < 1 and α
n
+ β
n
< 1. In
these expressions, K
it
denotes services from physical capital, L
at
denotes the services
from land, N
it
denotes the number of worke rs, H
it
denotes human capital, and A
it
denotes the possibly sector-specific technology shock. The form of the production
for the non-agricultural sector is similar to the one assumed by Mankiw, Romer,
and Weil (1992). For the agricultural sector, we allow for lands under cultivation to
enter as another factor. We assume that both pro duction functions display constant
returns to scale.

16
Totally differentiating the production function and assuming that technological
progress is Hicks neutral, the expression for the Solow residual or Total Factor
Productivity (TFP) in each sector is given by
dA
it
A
it
=
dY
it
Y
it
−
X
J=K,N,H(andL)
ξ
J
dJ
it
J
it
, i = a, n, (3)
where ξ
J
denotes the output elasticities with respect to the relevant factors.
5
One of the main issues has to do with determining the elasticities of output
with respect to the factors in the relevant production functions. According to one
approach, it is possible to estimate them from the data based on a simple Cobb-
Douglas specification for the production function with multiple inputs. Ismihan and
Metin-
¨
Ozcan (2006) employ a c ointegration approach to deriving such estimates in
a model with human capital augmented-labor and capital as inputs for the period
1960-2004.
6
They find that the capital share is estimated between 0.58-0.65. They
also calculate the share of capital income from the national income accounts and
report large variations in this quantity due to distributional shifts during the late
1970’s and the late 1980’s.
In contrast to this approach, we consider a much longer period and also include
human capital explicitly as an additional factor in the production function. In such
a long sample period, finding reliable instruments for the production function esti-
mation is difficult.
7
Hence, our choice of the input shares follows standard estimates.
5
Alternatively, we can assume that the product and factor markets are perfectly competitive in
each sector and derive an expression for the Solow residual as the difference between the rate of
output growth and the share-weighted growth rates of the factors, where the shares are defined as
revenue shares of the factors. It is well-known that the Solow residual overstates technical progress
if there is imperfect competition or endogenous changes in efficiency due to increasing returns to
scale. See, for example, Altug and Filiztekin (2002).
6
Sp ecifically, they define human capital-augmented labor as H
t
= h
t
N
t
, where h
t
is educational
attainment p er worker. A similar specification is adopted by Hall and Jones (1999).
7
In the production function literature, standard instruments are oil shocks, the money supply,
government spending, and the political party in power. See Hall (1988), for example. Finding the
equivalent of these instruments for the Turkish case over our sample perio d is not easily accom-
plished.

17
0 1 2 3 4 5 6
1920 1930 1940 1950 1960 1970 1980 1990 2000 2010
K/Y K/Y Agr. K/Y Non−Agr.
Capital Output Ratio
Figure 4: Capital-Output Ratio
(See also Lains, 2003.) The specifications that we adopt for the production functions
in each sector are given in Tables 4 and 5. Thus, in the one-sector model without
human capital, we take the share of capital to be 0.40. This is also the case for
the non-agricultural sector in the two-sector model. For the agricultural sector, the
share of capital is taken to be 0.25 for the period up to 1950 and 0.30 for 1950-
2005, reflecting the mechanization of agric ulture that began in the 1950’s. In the
models with human capital, we assign zero weight to human capital for the period
1880-1929 in the non-agricultural sector, reflecting the very low levels of literacy and
educational attainment for this era. (For discussion on this point, see Section 4.1.)
After 1929, we take the human capital share to be 0.15. For the agricultural sector,
we assign even lower weights to human capital, assuming that the contribution of
human capital to production in this sector was essentially zero until 1950 and 0.10
thereafter.
In Figures 4 and 5, we display the evolution of capital-output ratio and our
education variable over the sample period. Be ginning from 1950 onwards, we observe

18
0 2 4 6 8
Years of schooling
1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000
Total Pop. Males Females
Human Capital
.2 .3 .4 .5 .6 .7 .8
1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000
Gender Ratio in Human Capital
Figure 5: Human Capital and the Gender Ratio
that for the aggregate economy the capital stock has increased from a little less than
twice the l evel of output to over three times. The capital-output ratio in the non-
agricultural sector displays a similar pattern as the aggregate capital-output ratio.
However, we observe that the growth rate of the capital-output ratio has fallen
since 1980. In terms of the education variable, the period especially after 1960
sees an increase in the trend growth of this variable. de la Fuente and Domenech
(2000) have observed that various data-related problems such as sharp breaks due
to classification problems or the increase in the trend growth of educational or
human capital variables (especi ally for developing countries) in periods when there
has been decline in productivity may be a reason behind the negative effect of
human capital variables in some cross-country regressions. The construction of our
education variable is intended to minimize such problems.
8
In Figure 5, we also
graphed average female educational attainment as a fraction of the overall average
attainment. This figure shows that there are differences between the educational
attainments of females and males, although these differences appear to be leveling
off over time.
The difference between the rate of growth of output and the contribution of
input growth represents the rate of growth of TFP. That is, it represents the various
8
See the Appendix.

19
types of incre ases in the productivity of the factors including technological and
organizational change. How much output growth is attributed to improvements
in TFP and how much to the growth of inputs also depends on the ways in which
inputs are measured. Dale Jorgenson helped construct quality-adjusted input series.
These improvements have substantially reduced the measured contribution of TFP
to output growth. In our case, we were unable to use such quality-adjusted input
series. On the other hand, if labor is measured in hours, unadjusted for education
and experience, for example, increases in labor and total productivity will be higher.
In our case, we measure labor in person years, admittedly a crude measure, due to
the absence of other more detailed or disaggregated series. On the other hand, we
are unable to take into account changes in hours worked per year. In the absence of
the related data, more hours worked per person shows up as an increase in TFP in
our calculations.
9
By adding the human capital into equation, however, we attempt
to account for at least part of the improvement in the quality of labor. Similarly,
since our capital stock series does not take into account quality improvements, the
contribution of increases in the quality of capital will be attributed to TFP growth.
In the developed countries TFP growth has been a major source of growth
eve n in cases where inputs are adjusted for quality improvements. Studies show
that any where from 20% to as much as 50% of the total increase in output is
attributable to TFP in the developed countries. (See Denison, 1985.) Moreover,
recent research has also shown that there are very large differences in total factor
productivity levels and rates of growth of TFP between countries even after taking
into account differences in education levels. In fact, variation in the productivity of
inputs amongst countries is more important than the variation in the quantity of
inputs as physical and human capital in explaining levels of output or output per
worker. (Islam, 1995 and Helpman, 2004, pp. 28-31)
9
Adamop oulos and Akyol (2004) make use of an hours worked series for Turkey from data from
the Groningen Growth and Development C entre (GGRC). H owever, they admit that the GGRC
approximates the hours for Turkey by those for Greece.

20
3.1 Main Findings
The results of the growth accounting exercise are displayed in Tables 6 and 7. These
results are instructive in that they show strong evidence for the effects of alternative
historical, institutional, and policy factors. Considering the Ottoman period, the
contribution of TFP growth to GDP growth is negligible. While there is some
evidence of higher TFP growth during the 1880-1913 period for the non-agricultural
sector, TFP growth calculated under the different models for the aggregate economy
is less than 0.5% for this period. Recall that the share of human capital has been
set at zero over the 1880-1929 period, reflecting the low incidence of skills or formal
education in the overall Ottoman populace. Hence, the models with or without
human capital yield the same results regarding the contri bution of TFP over this
period. Undoubtedly the massive decline in output and the labor force in the period
corresponding to WWI and the War of Independence play a role in this result.
However, even if we were to look at the pre-WWI period between 1880-1913, TFP
growth implied by some of the models is only slightly higher.
TFP growth essentially begins to emerge after 1930, beginning with the agricul-
tural sector. This date corre sponds to the adoption of protectionist measures and
state-led industrialization following 1929 and the onset of autarkic policies world-
wide. However, it also signals an era of new polici es that were implemented in a
variety of political and economic spheres as part of the creation of the new Repub-
lic. One of the reforms is undoubtedly the liberalization of laws and practices left
over from the Ottoman era that governed peasant life such as the abolition of the
tithe (or ¨o¸s¨ur) in 1925.
10
Considering the 1930-1950 period, we see the effect of
these changes in the tripling of TFP growth relative to the 1880-1929 peri od for
the aggregate economy (from 0.25% to 0.92%). If we allow for human capital, this
increase is smaller but still significant (from 0.25% to 0.63%.) An examination of
the two-sector model shows that it is the increase i n TFP growth in the agricultural
10
For a further description of economic and social conditions in Ottoman society, see Inalcık and
Quataert 1996).

21
sector that is responsible for the overall increase in TFP.
11
Whereas TFP growth is
negative in the non-agricultural sector during 1930-1950 (see the columns marked
2-N and 2H-N in Table 6), there are significant increases in agricultural TFP growth
for this period.
Table 6 shows that there is a trend increase i n TFP growth for the Turkish econ-
omy. Comparing the broad periods 1930-2005 with 1950-2005, we observe that the
contribution of TFP growth rises from around 1.3% to nearly 1.5% for the one-sector
models acr oss these two periods. Taken as a whole, the results for 1950-2005 are,
in spirit, very similar to the results for 1930-2005, namely, that TFP improvements
observed over the consecutive overlapping periods are mainly due to improvements
in the agriculture sector, and that the contribution of non-agricultural TFP growth
to aggregate TFP improvements is not important. However, this picture begins to
change when we consider the post-1980 era. TFP growth begins to rise above 1.5%
for the first time for the one-sector models. Furthermore, an examination of the re-
sults for the two-sector models shows that it is TFP increases in the non-agricultural
sector that are at the source of TFP increases for the economy as a whole.
These results yield some noteworthy conclusions about the role of TFP growth
in overall output growth for the Turkish economy. The contribution of TFP growth
arising from the agricultural sector over the fifty-year p eriod stretching from 1930
to 1980 reflects the agrarian basis for the Turkish economy which had its origins
in the Ottoman era, and the role that liberalization and modernization of practices
in agriculture had on improving agricultural productivity beginning from the early
Republican era and onwards. Second, we observe a trend in TFP improvements
over the period 1930 to 2005. Third, there is a qualitative change in the pattern of
productivity growth after 1980, with TFP improvements originating from the non-
agricultural sector including manufacturing and services becoming more important.
As we discuss more fully later, the main factor underlying this change lies in the
11
The results for the two-sector model are obtained by weighting the results for the agricultural
versus non-agricultural sectors with their weights in total output.

22
trade and financial liberalization measures that took place after 1980 and that had
the effect of opening up the Turkish economy to the rest of the world. Despite these
positive facts, however, an inescapable fact that arises from our analysis is that
TFP growth is, in general, low for Turkey. During the entire 1950-2005 period, the
two-sector model without human capital implies that TFP growth is 0.83%.
These findings are in line with the results of other studies that have conducted
growth accounting exercises for Turkey. At the level of aggregate economy, Saygılı ,
Cihan, and Yurto˘glu (2001) find that TFP growth is equal to -0.29% for 1972-1979
and 0.44% for 1980-2000. Altu˘g and Filiztekin (2006) examine the behavior of the
manufacturing sector - the so-called “engine of growth” - for the period 1970-2000,
and find that the contribution of TFP growth to output growth becomes positive
only after 1980. The contribution of the current paper is to demonstrate that this
result holds over much longer horizons and after taking into account the role of
human capital and differences between the agricultural and non-agricultural sec-
tors. In terms of the fraction of GDP growth explained by TFP growth, we find
that TFP growth explains around 30% of output growth in all of the three main
periods – 1880-1929, 1930-2005, and 1950-2005 – under study. Taken at face value,
this finding seems to suggest that, in percentage terms, the contribution of TFP
growth to output growth remained unchanged over the 125 years considered in our
study. However, the results for the pre-World War I most likely re flect problems in
overestimating output growth or underestimating the growth of some inputs. Like-
wise, the contribution of TFP growth to agricultural output (in percentage terms)
appears high during the period 1980-2005 but this is due the declining share of the
labor force employed in agriculture. (See Table 3.) Summarizing, we conclude that
TFP growth becomes important as a source of Turkish long-term growth, reach-
ing nearly 40% of total output growth in the absence of human capital or around
30% when human capital is accounted for, following the reforms aimed at trade
and financial liberalization after 1980. Regardless of the ultimate causes of these
results, whether they be a one-time increase in productivity, the diffusion of technol-

23
ogy and learning-by-doing effects, increased efficiency through greater competition,
or a variety of other effects that have been named in the growth literature, they
are consistent with the findings of Edwards (1998) who shows in a sample of 90-
odd countries over the period 1960-1990 that greater openness leads to higher TFP
growth.
12
Our second main finding is that for the post-1930 era, the growth rate of output
is essentially due to capital accumulation. In the earlier period 1880-1929, the role
of capital accumulation is negligible. Considering some sub-periods, we observe
that capital accumulation was the primary source of output growth for the “import-
substituting” era corresponding to 1950-1980. During this period, the growth rate
of physical capital is around 2.5%, with capital accumulation in both agricultural
and non-agricultural sectors occuring at their highest observed historical values.
The rate of capital accumulation is 2.14% between 1950-2005. By contrast, out
of total output growth of 4.55% for this period, TFP growth is 1.41% and the
growth of the labor input is 1.16%. Capital accumulation continues to remain as
the primary source of growth even after allowing for human capital or different
production technologies in the agricultural versus non-agricultural sectors. Indeed
the main factor that distinguishes the pre- and post-1950 periods is the increased rate
of physical capital accumulation in both agric ultural and non-agricultural sectors.
After 1980, we observe that the rate of physical capital accumulation is nearly on
par with TFP growth.
Are these results surprising? In our opinion, no. Young (1995) was among the
first to argue that TFP growth played little role in East Asian growth. Likewise,
Collins and Bosworth (1996) find that TFP growth makes a surprisingly small con-
tribution to East Asia’s growth performance. Instead, it has been the ability of
these countries to achieve high rates of saving and investment that has led to rapid
12
There are a number of papers that examine the impact of openness on TFP growth for Turkey
directly. See Filiztekin (2000); see also Altu˘g and Filiztekin (2006) for a further review and discus-
sion of the impact of openness on growth.

24
growth in output. Lains (2003) also demonstrates that Portuguese growth during
1934-1947 was strikingly due to capital accumulation, which continued to remain
as the most important source of growth for the periods 1947-1973 and 1973-1990.
Kumar and Russell (2002) and Maudos, Pastor, and Serrano (2000) construct world
production frontiers using output and inputs of labor and capital under the assump-
tion of CRTS. Their results indicate that technical change is non-neutral, leading
to productivity growth at higher capital-labor ratios, and that capital deepening
is the primary cause of convergence in the distribution of labor productivity for
the peri od 1960-1990. Thus, in common with many developing contrives including
some high-performers such as the East Asian or South European countries, we find
that output growth for Turkey derives primarily from capital growth. However, the
rate of capital accumulation is typically lower for Turkey than it is for Portugal,
for example. (Compare 5.35% for Turkey over the period 1950-2005 with 6.47% for
Portugal over the period 1947-1990.) Furthermore, we find that the rate of capital
accumulation slows after 1980, falling to 4.21%.
3.2 Labor Productivity Decomposition
Another way of examining a country’s growth performance is in terms of the behavior
of labor productivity. Total labor productivity can increase for two reasons. The
first reason is due to productivity increases within each sector. The second has to
due with factor re-allocation across sectors. In this sectio n, we examine the impact
of sectoral re-allocation out of agriculture as another source of overall growth for
the Turkish economy.
Total labor productivity can be expressed as:
p
t
=
X
i=a,n
w
it
Y
it
L
it
, (4)
where w
it
shows the weight of sector i in total employment at time t. Taking first

25
differences one reaches:
∆p
t
=
X
i=a,n
w
i,t−1
∆p
it
|
{z }
P GE
+
X
i=a,n
p
i,t−1
∆w
it
|
{z }
SE
, (5)
where p shows labor productivity, P GE denotes within sectoral productivity growth,
and SE denotes the sectoral re-allocation effe ct.
13
Table 8 shows the impact of sectoral re-allocation out of agriculture across the
different pe riods for Turkey. We find that the contribution from sectoral allocation
arises primarily during the 1950-1980 period. Indeed, for this period, 57% of overall
growth is due to resources moving out of low productivity uses in agriculture into
higher productivity uses elsewhere, primarily manufacturing. For the period 1980-
2005, the contribution of this channel falls to 33%. Our results indicate that a major
source of growth has involved shifting resources out of agriculture and into more high
productivity uses elsewhere. However, the proce ss of structural transformation has
proceeded at a relativel y slow pace in Turkey. Furthermore, productivity growth
in the non-agricultural sector precisely during the period when this sectoral re-
allocation was occuring has been either negative (-10.3% during 1930-1950) or low
(13.4% during 1950-1980.) Adamopoulos and Akyol (2006) focus on a similar set
of findings to explain Turkey’s relative pe rformance or “relative stagnation”, in
their parlance. They fo cus on the behavior of real GDP per working age person
and conduct decompositions into labor productivity and labor force participation
rates (or employment rates). They conclude that Turkey’s inability to catch up to
devel oped country levels rests on the moderate growth rate of labor productivity
and the decline in labor force participation (or employment) rates. They then seek
to explain the decline in the labor market participation (or employment) rates in a
neoclassical model with home production and differing rates of return to activities
in the market versus the household sector. We return to the findings of this paper
at a later point. Returning to Table 8, we note that it is only after 1980 when the
13
It is also possible to illustrate the sectoral shift effect using TFP as a measure of productivity
instead of labor productivity.

26
process of structural transformation had begun to slow do we see significant intra-
sectoral productivity increases in non-agricultural activities. In this period, 67%
of total labor productivity growth is due to the productivity growth component
(PGE), with 52.6% growth being attributed to the non-agricultural sector.
The role of sectoral allocation in contributing to overall growth has been stressed
by a number of authors. Broadberry (1998) argues that the US and Germany over-
took the UK not by increasing productivity levels in manufacturing - as commonly
supposed - but by shifting resources out of agriculture and by improving their rela-
tive productivity position in services. Using a shift-share analysis that accounts for
the effect of changes in employment in each sector as in our analysis, he finds that
structural change is sufficient to account for the bulk of the growth differential be-
tween Germany and the UK during the period 1871-1990 and to more than account
for the US-UK growth differential over a comparable period. Peter Temin (2002)
has described how Western Europe utilized sectoral re-allocation out of agriculture
after WWII to achieve high rates of overall growth. Ventura (1997) has argued that
the process of structural transformation may be also linked to capital deepe ning
or capital accumulation when the interdependence between different economies is
taken into account. In this vein, the East Asian countries achieved a rapid struc-
tural transformation through a conscious effort aimed at high saving and investment.
The growth in the capital-output ratios for a variety of East Asian countries ranged
around 3-3.5% during the period 1966-1990. By contrast, Turkey’s investment per-
formance has been peripatetic. Turkey’s capital-output ratio grew at a rate of 1.2%
during 1960-2005. For the period 1960-1980, this growth rate is 3.14%, which is
comparable to some of the East Asian countries. However, the growth rate is only
0.4% over the per iod 1980-2005. Taken as a whole, these results suggest that the
relatively slow process of structural transformation and the inability of the Turkish
economy to achieve sustained increases in its growth rate may be inextricably linked
to the low rates of saving and capital accumulation.

27
4 “Puzzles”
These main findings raise a number of questions or puzzles regarding the Turkish
experience.
• Why is TFP growth so apparently low in Turkey? For example, average TFP
growth is around 2% during 1966-1990 for the East Asian countries.
• Another puzzle is the low rate of capital accumulation itself - not why growth
in Turkey is primarily due to capital ac cumulation.
• A third puzzle for Turkey has to do with relatively low rate of transition from
agriculture to non-agricultural activities. As an indicator of this slow pace, we
note that the share of population remaining in agriculture in Turkey by 2005
is near ly 34%, one of the highest in Europe.
4.1 Resolving the puzzles
In this section, we provide both a narrative discussion of the factors that we have
identified and also seek to quantify the impact of these factors over a sample stretch-
ing back to 1880. Our goal is not to identify one out of a potentiall y competing
number of explanations of growth but to examine the Turkish experience in the light
of some existing theories. Our discussion is organized around three main factors.
• Institutions
• Human capital
• Macroeconomic policy-making
These factors typically figure among the co-variates of cross-country growth regres-
sions. They have also been used in comparative exercises of growth experiences.
As we discuss below, the recent literature has come to view the interrelationships
among these factors as providing key insights into long-term growth.

28
Institutions
The idea that institutions are a key determinant of growth has witnessed a revival.
One may define as an institution as a formal or informal set of rules in the sphere
of economic, political or social interactions. This revival has taken place against
a backdrop of scholarship that assigns a crucial role to institutions in economic
devel opment; see North (1981). Early studies that recognize the importance of in-
stitutions include Knack and Keefer (1995) and Mauro (1995). More recently, Hall
and Jones (1999) argue that the disparity in observed income worker or in different
words, productivity, may only be explained by differences in social infrastructure,
by which they mean “the institutions and government policies that determine the
economic environment within which individuals accumulate skills, and firms accu-
mulate capital and produce output.”
14
They identify the quality of institutions in
terms of diversionary activities. Acemo˘glu, Johnson, and Robinson (2001) provide
an ingenious test of the importance of institutions by arguing that the incidence
of institutions that favor diversion or expropriation in place of investment or pro-
duction in a set of developing countries with a former colonial past correlate with
settler mortality. That is, locations that were favorable to settling by Europeans
in the early stages of colonialism also feature, given the persistence of institutions,
good institutions today. Both studies use a measure of expropriation risk developed
by Political Risk Services to proxy for the quality of social infrastructure or insti-
tutions. While they find a significant effect from their measure of institutions to
differences in incomes per capita, Acemo˘glu, Johnson and Robinson (2001) admit
that their treatment of institutions can be likened to a “black box” in that insti-
tutional features such as expropriation risk, the enforcement of property rights, or
eve n the rule of law are typically equilibrium outcomes.
A contrarian argument is put forward by Glaeser al (2004), who argue that coun-
tries that manage to accumulate sufficient levels of human capital are also able to
devel op sound institutions. Thus, their argument stresses the fact that there may be
14
See also Rodrick, Subramanian, and Trebbi (2004).

29
reverse causality from growth to good institutions. In making their case for the pri-
macy of human capital over institutions, they also take issue with various measures
used in the literature to measure the quality of institutions. In their parlance, it is
“constraints on government” which help to secure property rights that are a defining
feature of institutions. (See also North, 1981.) Another characteristic of institutions
is their “durability” or permanence. By both counts, they argue that measures such
as expropriation risk, government effectiveness, and constraints on the executive are
poor measures of institutions. The first two represent to a greater or lesser degree
the outcomes from past choices made by governments while the third is typically
ver y volatile in developing countries and reflects the results of the most recent elec-
tions. Following the second line of argument, Persson and Tabellini (2006) create an
index of democracy in terms of a democratic capital variable by cumulating periods
of democratic rule as in a simple capital accumulation equation. They argue that
the influence of democracy must occur through investment decisions and expecta-
tions. In their framework, the level of TFP differs depending on whether the current
regime is a democracy or an autocracy. In their empirical results, they assume that
TFP is higher in the democratic regime. After controlling for the stability of the
regime, they then find that democracie s - on average - grow faster than autocracies.
Rajan and Zingales (2006) present a third line of attack that suggests neither insti-
tutions nor the level of human capital may be sufficie nt to explain long-term growth
experiences if one abstracts from political economy considerations. Their point is
that economic institutions also determine the distribution of income and wealth. In
other words, they determine not only the size of the aggregate pie, but also how it is
divided amongst different groups in society. If those with the greatest income favor
institutions that allow for rent-seeking or expropriation, then weak institutions may
persist through time due to their self-fulfilling effect on the income distribution.
For long-term growth, economic institutions should not offer incentives to narrow
groups, but instead open up opportunities to broader sections of society.
As our previous arguments indicate, an institution is defined by its durability

30
as much as by characteristics such as constraints on rule. Following Persson and
Tabellini (2006), we created a measure of democratic capital by cumulating periods
of democratic rule as in a simple capital accumulation equation. If there is a change
of regime, say, a military take-over, then the existing stock of political capital starts
depreciating at some given rate. We use data on regime changes and transitions from
the Polity IV database which are avaliable for all countries with populations greater
than 0.5 million since 1800. Specifically, this database provides separate informa-
tion on quantitative indicators for the state of democracy and autocracy in a given
country as well as on the nature of various interruptions such as foreign occupations
or the general disruption of central authority. It also provides a measure denoted
polity2, which considers the difference between the demo cracy and autocracy indi-
cators. Following Persson and Tabellini (2006), we suppose that the current re gime
equals one if the polity2 is strictly positive, and zero otherwise.
15
This measure can
be constructed for Turkey since the 1880’s because there is potentially very rich
information on the political regime since that period.
The process of political and social reform and transformation in Turkey has a
long history, although this process was interrupted by such major events as World
War I. Turkey’s process of institutional change also reflects the impact of global
factors. Turkey’s democratic experience achieves an early and if imperfect start in
1876. This year corresponds to the announcement of the first Constitution in a Mus-
lim country. In the Polity IV data this short-lived experiment registers as a small
positive increase in their democracy variable. However, there is no net increase in
democracy because the continuing autocratic practices during this pe riod outweigh
any positive effects of the constitution. Following the abolition of the First Consti-
tutional Assembly, the thirty years that follow are characterized by the absolutist
rule of Ab dulhamit in an era when the Ottoman Empire is witness to large terri -
15
Let s
t
denote the current regime, where s
t
= 0, 1, and d
t
denote the index of democratic capital.
Then d
t
= s
t
+ (1 − δ)d
t−1
, where 0 < δ < 1 We chose a depreciation rate of δ = 5%. This is
consistent with the approach in Persson and Tabellini (2006), who consider depreciation rates of
1% and 5%.

31
5 10 15 20 25 30
1950 1960 1970 1980 1990 2000 2005
GDP per Worker Democratic Capital
GDP per Worker and Democratic Capital
Figure 6: Democratic Capital versus GDP Per Worker. Source: Polity IV
torial losses. This period typically achieves the worst rating on the autocracy scale
in Polity IV. The next experiment in representative rule is the Second Constitu-
tional Assembly formed after the deposition of Abdulhamit by the emerging Young
Turk movement in 1908. This registers as a positive event in the democracy indica-
tor which is not quite high enough to outweigh the impact of remaining autocratic
practices.
16
The aftermath of World War I is an interregnum period, ending with a
collapse of central authority in 1922. The period beginning with 1923 correspond-
ing to the establishment and early years of the Turkish Republic does not attain a
positive rating on the polity2 scale because this period is characterized essentially
by one-party rule led by a single leader, Mustafa Kemal Atat¨urk. Nevertheless the
creation of a formal law-making body together with parliamentary elections sees a
lightening of the autocracy index for this period. The period after 1934 and during
World War II period constitute a slight worsening in these circumstances.
Turkey’s process of accumulating democratic capital essentially obtains a start
16
The polity2 variable is -1 for the perio d 1909 to 1917.

32
in the post World War II era with multi-party elections in 1946. Figure 6 displays
our measure of democratic capital together with a measure of GDP per worker. In
1950 a newly-formed party, the Democratic Party, ascends to power. The period
after 1950 also deserves spe cial mention. As we noted earlier, this corresponds to
an era of rising incomes globally, as it does in Turkey. There is also a shift in power
from the cadres who led the transition from Ottoman rule to a secular nation state.
In the economic sphere, these cadres promoted the creation of national industrial
bourgeoisie based on a state-led model of development. In contrast to the earlier
period, the 1950’s is witness to the rise in influence of landed intere sts and private
sector industrialists. From Figure 6, we observe a steady increase in demo cr atic
capital during the 1950’s.
For better or worse, however, this period is also witness to many of the populist
policies that have shaped political life in Turkey for the past fifty years. In a so-
ciety where the majority of the population has been agricultural but political and
economic power has rested with a narrowly defined elite and its outgrowth, political
parties in Turkey have seen a b enefit to following gradualist and populist policies.
On the one hand, in contrast to the rapid structural transformation and the growth
in the utilization of labor and capital that has been observed in the East Asian coun-
tries, agricultural subsidies in Turkey prevented those employed in agriculture from
moving to more high productivity sectors. Likewise, an open economy orientation
to Turkish industry came only in 1980. Even after trade liberalization had taken
place, Turkey failed to adopt policies that would promote greater international c om-
petitiveness by concentrating on exports of technology-intensive goods, for example.
Instead, the distribution of export subsidies took a more prominent place.
17
17
In contrast to Turkey’s experience, it appears that China is repeat ing the performance of the
East Asian countries. In their comparison of China and India, Bosworth and Collins (2007) note
that China moved more rapidly than India in lowering trade barriers and in attracting foreign direct
investment. Furthermore, China has experienced “explosive growth” in its industrial sector whereas
Indian growth has been driven by the services. Finally, both a high rate of capital accumulation
and TFP growth appear important in China.

33
Indeed it is precisely to this type of scenario that Rajan and Zingales (2006)
refer, in which self-interested constituencies determine the economic and political
agenda. In their paper, they consider three types of constituencies, oligopolists, the
educated (or the middle class), and the uneducated (or the poor), and two types
of reforms, those promoting greater competitiveness and those leading to greater
access to education (or human capital). They show that there is a wide array of
circumstances in which no reforms will take place. If there is a small educated
constituency that enjoys substantial rents, then this constituency may side with the
oligopolists in opposing educational reforms. Likewise, the uneducated will oppose
pro-market reforms unless these are accompanied by human-capital or endowment-
enhancing measures. They also show that there exists a narrow set of circumstances
in which reform will take, namely if there is a substantial middle-class and the
oligopolist is relatively efficient.
It is also worth commenting on the approach followed by Adamopoulos and Akyol
(2006) at this point. In their analysis, they link the relatively slow pace of structural
transformation i n Turkey to the high burden of taxation on market activities in the
non-agricultural sector. It is our contention, however, that such policy variables
as government spending and the tax rates needed to sustain them are determined
as part of a political equilibrium which has been slow to change in Turkey. These
authors also argue that Turkey’s stagnation relative to the US, say, is not essentiall y
attributable to physical or human capital accumulation because these variables have
tended to display catch-up relative to the US. Our argument, by contrast, is that
pro-market reforms or reforms aimed at increasing the human capital endowment
were late in coming and when they did come, they were implemented in partial
ways that did not lead to take-off as they did the East Asian countries or the
late-starters in Southern Europe. In the absence of such reforms, investment in
the non-agricultural sector needed to effect a rapid structural transformation did
not take place, leading to an outcome of slowly rising productivity in that sector
combined with low employment rates.

34
Returning to the historical record, the end of the 1950’s and the beginning
of the 1960’s is witness to political and economic difficulties for Turkey. On the
one hand, a foreign e xchange crisis and serious foreign exchange shortages were
causing economic difficulties. On the other hand, politically repressive measures
were leading to widespread opposition, espe ciall y among the educated classes. These
eve nts culminate with the military coup of 1960. Suprisingly, the coup itself and the
period after the coup rate relatively highly according to the polity2 sc ale. One reason
may resi de in the new liberal Constitution, encompassing as it does a broad-reaching
set of rights of expression, assembly, and organization. The 1960’s and early 1970’s
see rapid growth in Turkey, with capital accumulation in heavy industries being
undertaken by state enterprises and the private sector producing consumer goods
under an import-substituting regime. The year 1971 corresponds to another military
takeover, this time in response to growing student and worker demonstrations. The
political and economic power of the workers, as well as their share in the total
pie, had been on the rise after World War II, esp eci ally during the ISI era after
1960. Student and union militancy continued after the return to a civil government
in 1974 but by this time, the first oil shock had o c cured. Many have attributed
the serious economic and political troubles that erupted after the mid-1970’s to
Turkey’s inadequate policy response to the first oil shock.
18
If we examine the
course of democratic capital accumulation during this period, we will observe an
increasingly erratic path. One way to view the military interventions that occured
in 1960 and 1971 are as interventions to curb the economic and political power of the
various constituencies. Indeed the 1960 intervention had the hues of a “middle-class”
revolution in so far as average incomes in Turkey were on the rise, and educational
attainment kept pace with this growth. By the late-seventies, however, the Turkish
economy was in serious difficulties, with a very severe foreign exchange cr isis and
the end of the import-substituting era in sight. The military coup of 1980 and its
18
See Dervi¸s, De Melo, and Robinson (1982) for an analysis of the various factors that contributed
to the buildup of the crisis.

35
−.1 −.05 0 .05 .1 .15 .2
% Growth in GDP per Worker
−.05 0 .05 .1 .15 .2
Democratic Capital
Growth of GDP per Worker vs. Change in Democratic Capital
Figure 7: Growth of Per Worker GDP versus Changes in Democratic Capital
aftermath registers as the largest loss of democratic capital for Turkey in the post
1950’s period. It is only by 1987 or thereabouts that the level of democratic capital
is able to reach its former level in 1970. The immediate effect of the 1980 coup
is to reduce the power of the worker s and trade unions. A second outcome is the
opening up of the Turkish economy to the rest of the world through a series of trade
and financial liberalization measures, many advocated as part of the Washington
Consensus. However, as we have argued above, these changes or reforms did not
go the full measure as the existing constituencies who had gained in the import-
substituting era now became principal beneficiaries of the new economic re gime.
In Figure 7 we examine the relationship between changes in our measure of
democratic capital and a measure of per worker GDP growth for Turkey. Interest-
ingly, we find that there is a positive association between per capita GDP growth
and increases in democratic capital for Turkey. This relationship essentially obtains
for the post World War II era because prior to that, Turkey’s democratic capital
is zero. Persson and Tabellini (2006) present a fully-specified model in which it is

36
the probability of a regime shift or the likelihood of exiting from democracy that
matters for growth.
19
Thus, their argument hinges on the fact that the relationship
between democracy and growth can be partly attributed to the stability of the cur-
rent regime. If market participants attach a positive probability of a shift to a more
unfavorable regime – as they also do in the analysis of Altu˘g, Demers, and Demers
(2007) – then they will refrain from undertaking productive investments, which will
lead to lower growth. If the unfavorable regime such as a military intervention is
actually re alize d, then there is also the impact through the transition to a state
which features a lower level of TFP and the further loss of democratic capital.
These results indicate that the process of democratization has led to growth
in Turkey. However, there are significant periods when there has be en a loss of
democratic capital. Furthermore, as we argued above, as in the case of other de-
vel oping countries such as the Latin American countries, the status quo in Turkey
has favored the pursuit of privileges from local and national governments for the
producers rather than the pursuit of productivity improvements or competition in
international markets. The political fragmentation induced after the 1980 military
intervention has produced fragile coalitions and weak governments. Furthermore,
Turkey’s political system has features which appear conducive to large government
and rent-seeking. Persson and Tabellini (2006) study the implications of political
institutions such as presidential versus parliamentary regimes, majoritarian versus
proportional voting systems, and accountability of politicians for public spending,
rent extraction, and structural policy. They show that proportional voting systems
(such as those in Turkey) lead to a larger size of government and higher govern-
ment spending. Similarly, parliamentarian regimes have larger government spend-
ing, though their economic performance rates better. In Turkey, these features have
been accompanied by political and macroeconomic instability, which has also led to
19
A similar argument is put forward by Altu˘g, Demers, and Demers (2007), who consider a model
with irreversible investment and regime shifts. In their case, the regime shift involves an unfavorable
state, namely, the state of separation, that has never occurred in the data.

37
corruption and the deterioration in the rule of law.
Human capital
Does human capital matter? In our previous discussion, we uncovered various
strands of the recent literature which state that human capital is as indispensable
for the functioning of democratic societies as it is for increasing the overall pro-
ductivity of the labor force. In his comparison, Zettelmeyer (2006) notes that the
Latin American and Asian countries’ p e rformance on simple education and health
indicators such as adult mortality track each other very well. Whereas Turkey’s
experience with democracy may not be any worse than that of these countries or
eve n the countries in Southern Europe, its record of human development is signif-
icantly different than the one observed for these countries. In their microfounded
account of Turkey’s growth experience, Adamopoulos and Akyol (2006) have argued
that Turkey’s record of human accumulation is not necessarily a factor underlying
its growth performance. They measure educational attainment by average years of
schooling in the Barro-Lee (2000) data set and argue that whereas Turkey had the
lowest level of educational attainment among Greece, Portugal, Spain and the US
in 1950 (or 1960) and still has the lowest level today, it has exhibited the largest
catch-up relative to the US. However, this view does not take into account that
there may be threshold or distributional effects of education.
20
Examining the Turkish experience in this light unveils some interesting facts.
As we ar gued earlier, the li teracy rate at the inception of the Republic stood at no
more than 10%. In 1935, the literacy rate for individuals over age 15 was 19%: 31%
for men and only 8% for women. By 1950, the adult literacy rate had increased to
28%; 47% for men and 13% for women. Educational policies in Turkey in the last
80+ years have succeeded in bringing the literacy rate to less than 89% by 2005;
20
In our analysis, we assumed that the impact of human capital is captured through a constant
returns to scale production function. Yet there also may exist externalities to human capital
accumulation as in Lucas (1988).

38
95% for men and 82% for women. The average level of schooling for the workforce
remains at 5.3 years. This is 2 or 3 years less than a country average based on the
experience of 50 countries in 2000. (See Saygılı , Cihan, and Yavan, 2006). The
role of educational attainment is also stressed by Bosworth and Collins (2007) and
Bosworth, Collins and Virmani (2007) for Chinese and Indian growth.
In cross-country regressions, Barro (2001) examines the q uantity and quality of
education along different dimensions after controlling for a set of variables that are
found to be significantly related to growth. As an example, his results indicate that
scores on science tests have a particularly strong positive effect on growth. These
results have important ramifications in the context of Turkey. The PISA study by
the OECD shows that the performance of Turkish school children on examinations
measuring mathematical and analytical ability is the lowest in the 27 countries in
the OECD following Mexico. On the other hand, Turkish school children display
the greatest (relative) success in terms of the percent who score in the top percentile
on the examinations. Duygan and G¨uner (2006) argue that a growing inequality
in access to education may be as worrying as observed consumption and income
inequality. (See also Saygılı , Cihan, and Yavan, 2006).
21
Are these worries allayed
by current polic ies and practices? The share of government spending accruing to
education i s 3.82% of GDP in 2002 whereas the OECD average is 5.73%.
Thus, we find that there are significant differences in the levels of educational at-
tainment between Turkey and countries with comparable income per capita. There
are also distributional issues in educational outcomes across performance and income
levels as well as large regional differences in Turkey’s human development indica-
tors between the mostly Kurdish southeast regions and the rest of the country. All
of these factors may impede overall growth through a variety of mechanisms. In
evaluating Turkey’s growth performance in the light of political events and alterna-
21
Similar worries are expressed for the Indian educational system, where the low quality of el-
ementary schools is leading to a move from public to private schools. See Bosworth, Collins, and
Virmani (2007).

39
−.4 −.2 0 .2 .4 .6 .8
1920 1930 1940 1950 1960 1970 1980 1990 2000 2010
Inflation
Figure 8: Inflation
tive policy regimes, we believe that a further focus on education and human capital
remains i mportant.
Macroeconomic policy-making
The role of prudent and effective macroeconomic policy-making obtains short shrift
when one begins to contemplate the impact of such factors as institutions or human
capital. Yet there is evidence that macroeconomic policy-making matters. Sirima-
neetham and Temple (2006) show that bad policies can offset by other factors, but
the best-performing countries are all characterized by high-quality macroeconomic
management. This is also confirmed by Collins and Bosworth (1996) in their study
of the determinants of long-growth for the East Asian c ountries. In Figures 8 and
9, we illustrate the behavior of two key macroeconomic series over a long horizon,
inflation and government budget deficits.
22
We note the persistently high inflation
rates dating from the mid-1970’s until the 2001 crisis. We also observe high and
22
See the Data Appendix for how these variables are measured.

40
0 5 10 15 20
1920 1930 1940 1950 1960 1970 1980 1990 2000 2010
Ex ante Bud. Def. Ex post Bud. Def.
Budget Deficit as a Share of GNP
Figure 9: Ex Ante and Ex Post Measures of the Budget Deficit
growing fiscal deficits which date back to the mid-1970’s. The government’s lax fis-
cal stance was thus supported by various inflationary measures. In order to finance
a growing fiscal deficit, public investment, including expenditures on education, de-
clined sharply and continues to remain low. A concomitant result of high inflation
and loose fiscal policy were the high real interest rates observed in Turkey during
the 1980’s and 1990’s. Undoubtedly, these factors led to the low and volatile growth
observed during this period, especially the 1990’s.
23
The culmination of these factors was the 2000-2001 financial crisis. This crisis
ushered in a new era. On the one hand, significant institutional changes were im-
plemented such as central bank i ndependence, the creation of new regulatory and
supervisory bodies for the banking sector, the i nstitution of fiscal discipline, to name
a few. Interestingly these changes were implemented by new e c onomic and political
groups c orresponding to the industrialists from the Anatolian region - as opposed
23
In a related analysis, Adrogue, Cerisola, and Gelos (2006) use a dynamic panel regression
analysis for Brazil to document the role of increasing government consumption and the real interest
rate in reducing Brazil’s per capita growth rate since the mid-1990’s.

41
to the Istanbul-based industrial elites that flourished under the import-substituting
regime and remained dominant in the 1980’s and 1990’s. The AKP (or the Justice
and Welfare Party) government of the recent years has been supported by these
emerging elites from the provinces.
24
5 Conclusions
In this paper, we have conducted a growth accounting exercise across broad historical
periods and policy regimes for the Turkish economy. We have also studied the
process of sectoral re-allocation for Turkey. Our results indicate that the Turkish
economy underwent significant changes across broadly defined historical and policy
regimes. It succeeded in changing its institutional environment, increasing the level
of educational attainment in the population, and ac cumulating significant amounts
of physical capital, all of which contributed to overall growth. Yet at various points
it departed from the company of the high-performing countries. Its experience of
democracy may be no worse than many other countries of its income level but its
economic performance in the 1990’s and even earlier has been disappointing.
In this paper, we have sought to understand the reasons behind this performance,
and to relate it in a comparative way to the performance of countries with better
known histories. We believe that the Turkish experience is interesting because it il-
lustrates many of the recent debates in the literature on the determinants of growth.
For one, Turkey’s experience with democracy bears witness to the fact that it is the
stability of the favorable regime that matters for growth. However, Turkey’s ex-
perience also points to the relevance of various political economy arguments that
have been forward to account for the persistence of poor outcomes observed in de-
vel oping economy contexts. These factors appear to be compounded and to interact
with problems arising from the relatively low level of human capital and, in the
24
In this paper, we have not provided a detailed examination of the period since the Justice and
Welfare Party took power.
¨
Oni¸s (2005) provides a careful look at these new political players, and
the transformation of Turkish politics that is occuring under their stewardship.

42
short-run, the lack of effective macroeconomic policy-making. We believe that none
of these factors can, by themselves, account for the Turkish experience. We also
believe that understanding the role of these factors is indispensable for predicting
Turkey’s performance in the future.
Appendix: Data and Sources
For most of the data employed in this study, we relied on the official population,
GDP, land under cultivation series as published by the State Institute of Statistics
for the period since 1923. For the labor force series we used the Bulutay (1995) study
which was published jointly by the State Institute of Statistics and the International
Labor Office at Ankara. For the capital stock series we began with the Saygılı
, Cihan, and Yurto˘glu (2005) study which goes back to 1972. We extrapolated
this series back to 1923 using the investment series as included in the national
income accounts and the national income study for the period 1923-48 undertaken
by Bulutay, Tezel and Yildirim (1974) and a depreciation rate of 4.2% per annum.
For the Ottoman era 1880-1913, we be gan with the data compiled and the na-
tional income estimates prepared by Vedat Eldem (1970) for the years before World
War I. These are mostly based on the official Ottoman statistics. We extended these
back to 1880 utilizing population growth rates, rural-urban breakdown of the pop-
ulation, Ottoman agricultural censuses and other relevant data. Estimates for the
capital stock for 1880 are derived by assuming rates of investment for the decades
before World War I as given by Eldem (1970).
Our education variable is defined as the average years of schooling of the la-
bor force, ages 15-64. Educational attainment of the population by gender and
age groups are available at 5-year intervals since 1935 through General Population
Censuses. The data for years in between the censuses are imputed using number
of diplomas awarded by gender and school level (primary, secondary - junior and
senior le vels -, vocational, and university degrees) in each year. The survival rates

43
by gender and five-year age groups are used to the depreciate educational stock.
The human capital series is then constructed by multiplying the number of per sons
that are alive and finished a particular school with the years of education required
for that degree.
Inflation is measured based on the GDP deflator. Due to lack of data, we
measured the budget deficit using both ex ante and ex post measures. The ex ante
or planned budget deficit figures are take from the official publication Maliye ve
G¨umr¨uk Bakanlı˘gı Yayın No: 1991/320. Konsolide B¨ut¸ce Kaynak: B¨ut¸ce Ba¸slangı¸c
¨
Odenekleri ve Gelir Tahminleri (1930-1991), (Der leyenler: Iclal Demir ve Mukadder
¨
Oner). The ex post or realized budget deficit figures beginning from 1975 are based
on the consolidated government budget figures obtained from the State Planning
Organization.
References
Acemo˘glu, D., S. Johnson and J. Robinson (2001), “The Colonial Origins of Com-
parative Development: An Empirical Investigation,’ American Economic Re-
view 91, pp. 1369-1401.
Adamopoulos, T. and A. Akyol (2006). “Relative Stagnation Alla Turca,” Manuscript,
York University, Canada.
Adrogue, R., M. Cerisola, and G. Gelos (2006). “Brazil’s Long-term Growth Per-
formance – Trying to Explain the Puzzle,” International Monetary Fund.
Altu˘g, S., F. Demers, and M. Demers (2007). “Political Risk and Irreversible
Investment,” CES-ifo Economic Studies, forthcoming.
Altu˘g, S. and A. Filiztekin (2002). “Scale Effects, Time-varying Markups, and
the Cyclical Behavior of Primal and Dual Productivity,” Applied Economics
34(13), pp. 1687-1702.

44
Altu˘g, S. and A. Filiztekin (2006) “Productivity and Growth, 1923-2003.” In
S. Altu˘g and A. Filiztekin, (eds.) The Turkish Economy: The Real Econ-
omy, Corporate Governance, and Reform, (2005) Routledge Studies in Middle
Eastern Economics.
Barro, R. (1991). “Economic Growth in a Cross-Section of Countries,” Quarterly
Journal of Economics, 106, pp. 407-443.
Barro, R. (2001). “Human Capital and Growth,” American Economic Review, 91,
pp. 12-17.
Barro, R. and J.-W. Lee (2000). “International Data on Educational Attainment:
Updates and Implications,” CID Working Paper 42.
Bosworth, B. and S. Collins (2007). “Accounting for Growth: Comparing China
and I ndia,” NBER Working Paper 12943.
Bosworth, B., S. Collins and A. Virmani (2007). “Sources of Growth in the Indian
Economy,” Indian Policy Forum 3, forthcoming.
Broadberry, S. (1998). “How Did the United States and Germany Overtake Britain?
A Sectoral Analysis of Comparative Productivity Levels, 1870-1990,” Journal
of Economic History, 58, pp. 375-407.
Brock, W., S. Durlauf and K. West (2003). “Policy Evaluation in Uncertain Eco-
nomic Environments,” Brookings Papers on Economic Activity 1.
Bulutay, T. (1995). Employment, Unemployment, and Wages in Turk ey, State
Institute of Statistics, Ankara.
Bulutay, T., Y. Tezel and N. Yildirim (1974). Turkey’s National Income (1924-
1974), 2 vols. Ankara University Publications.
Cole, H., L. Ohanian, A. Riascos, and J. Schmitz (2004). “Latin America in the
Rearview Mirror,” Federal Reserve Bank of Minneapolis Staff Report 351.

45
Collins, S. and B. Bosworth (1996). “Economic Growth in East Asia: Accumula-
tion or Assimilation,” Brookings Papers on Economic Activity, Fall.
Denison, E. (1985), Trends in American Economic Growth, 1929-1982, Washington
DC, B rookings.
Dervi¸s, K., J. De Melo, and S. Robinson (1982). General Equilibrium Models for
Development Poli cy, Cambridge: Cambridge University Press.
Duygan, B. and N. G¨uner (2006). “Consumption and Income Inequality,” In S.
Altu˘g and A. Filiztekin (eds.), The Turki sh Economy: The Real Economy,
Corporate Governance, and Reform, Routledge Studies in Middle Eastern Eco-
nomics.
Eldem, V. (1970). Osmanlı Imparatorlu˘gunun Iktisadi S¸artları Hakkında bir Tetkik,
Istanbul, I¸s Bankası Yayınları .
Edwards, S. (1998). “Openness, P roductivity, and Growth: What Do We Really
Know?” Economic Journal 108, pp. 383-98.
Filiztekin, A. (2000). “Openness and Productivity Growth in Turkish Manufac-
turing,” Sabanci University Discussion Paper Series No. 0104.
de la Fuente, A.and R. Domenech (2006). “Human Capital in Growth Regressions:
How Much Difference Does Data Quality Make?” Journal of the European
Economic Associa tion 4, pp. 1-36.
Glaeser, E., R. La Porta, F. Lopez-de-Silanes, and A. Schleifer (2004). “Do Insti-
tutions Cause Growth?” Journal of Economic Growth, 9, pp. 271–303.
Hall, R. (1988). “The Relation Between Price and Marginal Cost in U.S. Industry.”
Journal of Political Economy, 96, 921–947.
Hall, R. and C. Jones (1999). ”Why Do Some Countries Produce So Much More
Output Pe r Worker than Others?” Quarterly Journal of Economics, 114, pp.
83-116.

46
Helpman, E. (2004) The Mystery of Economic Growth, Cambridge and London:
Harvard University Press.
Inalcık, H. and D. Quataert (1996). An Economic and Social History of the Ot-
toman Empire, 1300-1914. Cambridge: Cambridge University Press.
Islam, N. (1995). “Growth Empirics: A Panel Data Approach,” Quarterly Journal
of Economics 110, pp. 1127-170.
Ismihan, M. and K. Metin-
¨
Ozcan (2006). “The Growth Performance of the Turkish
Economy, 1960-2004,” Paper presented at the ICE-TEA International Eco-
nomics Conference, September 10-13, 2006, Ankara.
Knack, S. and P. Keefer (1995). “Institutions and Economic Performance: Cross-
Country Tests Using Alternative Institutional Measures, ” Economics and
Politics, 7, pp. 207-227.
Kumar, S. and R. Russell (2002). “Technological Change, Technological Catch-up,
and Capital Deepening: Relative Contributions to Growth and Convergence,”
American Economic Review 92, pp. 527-548.
Lains, P. (2003). “Catching Up to the European Core: Portuguese Economic
Growth, 1910-1990,” Explorations in Economic History 40, pp. 369-386.
Levine, R. and D. Renelt (1992). “A Sensitivity Analysis of Cross-country Growth
Regressions,” American Economic Review 82, pp. 942-63.
Lucas, E. E (1988). “On the Mechanisms of Economic Development,” Journal of
Monetary Economics 22, pp. 3-32.
Maddison, A. (2001). The World Economy: A Millennial Perspective, Paris and
Washington, D.C.: Organisation for Economic Co-operation and Develop-
ment.
Maddison (2003). The World Economy, Historical Statistics, Paris and Washing-
ton, D.C.: Organisation for Economic Co-operation and Development.

47
Mankiw, N.G., D. Romer, and D. Weil (1992). “A Contribution to the Empirics
of Economic Growth,” Quarterly Journal of Economics 107, pp. 407–437.
Maudos, J., J.M. Pastor, and L. Serrano (2000). “Convergence in OECD Countries:
Technical Change, Efficiency, and Productivity,” Applied Economics 32, pp.
757-65.
Mauro, P. (1995). “Corruption and Growth,” Quarterly Journal of Economics 110,
pp 681-712.
North, D. (1981). Structure and Change in Economic History. New York: W.W.
Norton and Co.
¨
Oni¸s, Z. (2005). “The Political Economy of Turkey’s Justice and Development
Party.” In H. Yavuz, ed., The Transformation of Turkish Politics: The Justice
and Dev elo pment Party of Turkey. Salt Lake City: University of Utah Press.
¨
Ozel, I. ve S¸. Pamuk (1998), “Osmanlıdan Cumhuriyete Ki¸si Ba¸sına
¨
Uretim
ve Milli Gelir,” Mustafa S¨onmez (derl.), 75 Yılda Paranın Ser¨uveni, Tarih
VakfıYayınları, Istanbul, pp. 83-90.
Pamuk, S¸. (2006), “Estimating Economic Growth in the Middle East since 1820,”
The Journal of Economic History, 66, pp. 809-28.
Pamuk, S¸. (2007a). “Is the Glass Half Full? Economic Change in Twentieth
Century Turkey,” In R. Kasaba (ed.) Cambridge History of Modern Turkey,
Cambridge: Cambridge University Press.
Pamuk, S¸. (2007b). “Agricultural Output and Productivity Growth in Turkey
since 1880,” P. Lains and V. Piniella (eds.), Agriculture and Economic Devel-
opment in Europe since 1870, Routledge Publishers.
Per sson, T. and G. Tabellini (2003). The Economic Effects of Constitutions. Cam-
bridge, MA: The MIT Press.

48
Per sson, T. and G. Tabellini (2006). “Democratic Capital: The Nexus of Political
and Ec onomic Change,” CEPR Discussion Paper 5654.
Prados de la Escosura, Leandro and J. Roses (2005). “The Sources of Long-run
Growth in Spain,” Paper presented at the Sixth European Historical Eco-
nomics Society Conference, September 8-9, 2005, Istanbul Turkey.
Quah, D. (1996). “Empirics for Economic Growth and Convergence,”European
Economic Review, 40, pp. 1353-75.
Rajan, R. and L. Zingales (2006). “The Persistence of Underdevelopment: Insti-
tutions, Human Capital, or Constituencies,” CEPR Discussion Paper 5867.
Rodrik, D., A. Subramanian and F. Trebbi (2004). “Institutions Rule: The Pri-
macy of Institutions over Geography and Integration in Economic Develop-
ment,” Journal of Economic Growth, 9, pp. 131-65.
Saygılı , S¸., C. Cihan, and H. Yurto˘glu (2001), “Productivity and Growth in OECD
Countries: An Assessment of the Determinants of Productivity,” Yapi Kredi
Economic Review 12, pp. 49-64.
Saygılı , S¸., C. Cihan, and H. Yurto˘glu (2005). “T¨urkiye Ekonomisinde Sermaye
Birikimi, B¨uy¨ume ve Verimlilik, 1972-2003,” Devlet Planlama Te¸skilati (State
Planning Organization) Publication Number 2686, Ankara.
Saygılı , S¸., C. Cihan, and Z. Yavan (2006), “E˘gitim ve S¨urd ¨ur¨ulebilir B¨uy¨ume:
T¨urkiye Deneyimi, Riskler ve Fırsatlar,” T
¨
USIAD.
Sirimaneetham, V. and J. Temple (2006). “Macroeconomic Policy and the Distri-
bution of Growth Rates,” CEPR Discussion Paper 5642.
Solow, R. (1958). “Technical Change and the Aggregate Production Function”
Review of Economics and Statistics, 39, pp. 312-320.
Temin, P. (2002). “The Golden Age of European Growth Reconsidered”, Review
of European Economic History, 6, pp. 3-22

49
Ventura, J. (1997). “Growth and Interdependence,” Quarterly Journal of Eco-
nomics, 112, pp. 57-84.
Young, A. (1995). “The Tyranny of Numbers: Confronting the Statistical Realities
of the East Asian Growth Experience,” Quarterly Journal of Economics 110,
pp. 641-680.
Zettelmeyer, J. (2006). “Growth and Reforms in Latin America: A Survey of Facts
and Ar guments,” IMF Working Paper WP/06/210.

50
1880 1913 1929 1950 1980 2005
Population (mill.) 13 17 13 21 45 72
Share of urban pop. (%) 26 28 24 25 44 68
Share of agric. in labor force (%) 80 80 85 84 51 34
Share of agric. in GDP (%) 54 55 42 54 26 11
Share of industry in GDP (%) ? 13 11 13 21 26
GDP per capita
†
850 1200 710 1620 4020 7500
- as % of W. Europe+US 37 29 16 24 25 30
- as % of dev. countries 147 168 n.a. 188 219 225
- as % of world 81 79 n.a. 77 89 117
Annual Growth Rates (%) 1880-1913 1913-50 1929-50 1950-1980 1980-2005 1913-2005 1929-2005
Population 0.8 0.6 1.8 2.6 1.9 1.6 2.1
GDP per capita 0.8 0.8 3.1 3.1 2.5 2.0 2.9
Total agricultural output 1.3 1.0 2.7 2.9 1.2 1.7 2,3
Total industrial output ? 3.1 5.3 7.7 5.8 5.3 6.5
†
PPP adj. in 1990 US dollars
Sources: For Turkish data except per capita GDP: State Institute of Statistics, Statistical Indicators, 1923-2002;
for GDP per capita series: Maddison (2001, 2003); Eldem (1970),
¨
Ozel and Pamuk (1998), and Pamuk (2006).
Table 2: Economic Indicators for Turkey

51
Annual Growth Rates (in %)
Aggregate Ec onomy
Output Labor Capital Human
Capital
1880-1913 1.48 0.73 1.76 2.78
1913-1929 -0.72 -1.31 -0.03 2.10
1930-1950 2.80 1.93 1.82 3.81
1950-1980 4.95 1.93 6.31 2.74
1980-2005 4.07 1.35 4.21 2.44
1880-1929 0.76 0.06 1.18 2.56
1930-2005 4.08 1.74 4.41 2.93
1950-2005 4.55 1.67 5.35 2.60
Agricultural Sector
Output Labor Capital Human Land
Capital
1880-1913 1.24 0.75 1.35 2.78 1.08
1913-1929 -0.87 -1.12 -1.31 2.10 -1.36
1930-1950 3.06 1.74 1.31 3.81 2.02
1950-1980 2.89 0.41 3.78 1.10 1.83
1980-2005 1.13 -0.96 1.83 -0.34 -0.34
1880-1929 0.55 0.14 0.48 2.56 0.28
1930-2005 2.35 0.31 2.47 1.34 1.16
1950-2005 2.09 -0.22 2.90 0.44 0.84
Non-Agricultural Sector
Output Labor Capital Human
Capital
1880-1913 1.73 0.61 1.88 2.78
1913-1929 -0.57 -2.47 0.26 2.10
1930-1950 2.57 3.09 1.92 3.81
1950-1980 6.15 5.56 6.61 3.73
1980-2005 4.74 2.98 4.35 3.18
1880-1929 1.53 0.62 1.52 3.07
1930-2005 4.73 4.04 4.60 3.57
1950-2005 5.51 4.39 5.58 3.48
Table 3: Annual Growth Rates (in %)

52
Factor Shares
1 Sector 2 Sector 2 Sector-Agric 2 Sector-Non-Agri.
Labor
1880-1929 0.60 0.45 0.60
1930-1950 0.60 0.45 0.60
1950-1980-2005 0.60 M 0.40 0.60
Capital i
1880-1929 0.40 x 0.25 0.40
1930-1950 0.40 t 0.25 0.40
1950-1980-2005 0.40 u 0.30 0.40
Land r
1880-1929 e 0.30
1930-1950 0.30
1950-1980-2005 0.30
Table 4: No Human Capital
Factor Shares
1 Sector 2 Sector 2 Sector-Agric 2 Sector-Non-Agri.
Labor
1880-1929 0.60 0.45 0.60
1930-1950 0.50 0.45 0.50
1950-1980-2005 0.50 0.30 0.50
Capital M
1880-1929 0.40 i 0.25 0.40
1930-1950 0.35 x 0.25 0.35
1950-1980-2005 0.35 t 0.30 0.35
Human Capital u
1880-1929 0.00 r 0.00 0.00
1930-1950 0.15 e 0.00 0.15
1950-1980-2005 0.15 0.10 0.15
Land
1880-1929 0.30
1930-1950 0.30
1950-1980-2005 0.30
Table 5: Models with Human Capital

53
Model 1 1-H 2 2-A 2-N 2-H 2H-A 2H-N
Contribution of TFP - Annual Growth Rates (in %)
1880-1913 0.33 0.33 0.42 0.25 0.61 0.42 0.25 0.61
1913-1929 0.08 0.08 0.59 0.37 0.81 0.59 0.37 0.81
1930-1950 0.92 0.63 0.86 1.17 -0.05 0.76 1.17 -0.21
1950-1980 1.27 1.37 0.58 1.04 0.17 0.78 0.97 0.50
1980-2005 1.57 1.55 1.14 1.07 1.21 1.20 1.00 1.25
1880-1929 0.25 0.25 0.48 0.29 0.68 0.48 0.29 0.68
1930-2005 1.28 1.23 0.84 1.08 0.46 0.91 1.08 0.56
1950-2005 1.41 1.45 0.83 1.05 0.64 0.97 0.99 0.84
Contribution of TFP Growth to Growth (in %)
1880-1913 22.7 22.7 28.5 19.7 35.4 28.5 19.7 35.4
1913-1929 -11.4 -11.4 -81.8 -42.5 -143.1 -81.8 -42.5 -143.1
1930-1950 32.7 22.4 30.5 38.1 -1.8 27.2 38.1 -8.3
1950-1980 25.6 27.6 11.6 36.0 2.7 15.8 33.6 8.1
1980-2005 38.7 38.2 28.0 94.5 25.6 29.4 89.0 26.4
1880-1929 33.1 33.1 62.4 51.7 69.0 62.4 51.7 69.0
1930-2005 31.2 30.2 20.5 46.1 9.7 22.4 46.1 11.8
1950-2005 30.9 31.9 18.3 50.4 11.7 21.3 47.2 15.2
Table 6: TFP

54
Annual Growth Rates (in %)
Contribution of Capital
Model 1 1-H 2 2-A 2-N 2-H 2H-A 2H-N
1880-1913 0.71 0.71 0.54 0.34 0.75 0.54 0.34 0.75
1913-1929 -0.01 -0.01 -0.11 -0.33 0.11 -0.11 -0.33 0.11
1930-1950 0.73 0.64 0.58 0.33 0.77 0.52 0.33 0.67
1950-1980 2.52 2.21 2.13 1.14 2.64 1.91 1.14 2.31
1980-2005 1.68 1.47 1.52 0.55 1.74 1.34 0.55 1.52
1880-1929 0.47 0.41 0.32 0.12 0.54 0.29 0.12 0.53
1930-2005 1.76 1.54 1.54 0.74 1.84 1.37 0.74 1.61
1950-2005 2.14 1.87 1.88 0.87 2.23 1.67 0.87 1.95
Contribution of Labor
1880-1913 0.44 0.44 0.35 0.34 0.36 0.35 0.34 0.36
1913-1929 -0.79 -0.79 -0.98 -0.51 -1.48 -0.98 -0.51 -1.48
1930-1950 1.16 0.97 1.39 0.78 1.85 1.21 0.78 1.54
1950-1980 1.16 0.97 2.26 0.16 3.34 1.88 0.12 2.78
1980-2005 0.81 0.68 1.39 -0.39 1.79 1.17 -0.29 1.49
1880-1929 0.37 0.31 0.33 0.28 0.37 0.30 0.28 -0.20
1930-2005 1.04 0.87 1.78 0.12 2.42 1.48 0.12 2.02
1950-2005 1.00 0.83 1.92 -0.09 2.63 1.61 -0.06 2.19
Contribution of Human Capital
1880-1913 0.00 0.00 0.00
1913-1929 0.00 0.00 0.00
1930-1950 0.57 0.00 0.57
1950-1980 0.41 0.11 0.56
1980-2005 0.37 -0.03 0.48
1880-1929 0.38 0.00 0.38
1930-2005 0.44 0.13 0.53
1950-2005 0.39 0.04 0.52
Contribution of Land
1880-1913 0.32 0.32
1913-1929 -0.41 -0.41
1930-1950 0.61 0.61
1950-1980 0.55 0.55
1980-2005 -0.10 -0.10
1880-1929 0.09 0.09
1930-2005 0.35 0.35
1950-2005 0.25 0.25
Table 7: Growth Accounting

55
Productivity Growth Component (PGE) Sectoral Shift Component (SE)
Agr Non-Agr Total Agr. Non-Agr. Total
1880-1913 34.2 72.1 106.3 1.5 -7.9 -6.3
1913-1929 21.0 163.4 184.4 15.5 -99.9 -84.4
1930-1950 56.1 -10.3 45.8 -10.0 64.2 54.2
1950-1980 29.5 13.4 43.0 -18.7 75.7 57.0
1980-2005 14.4 52.6 67.0 -15.8 48.8 33.0
Table 8: Labor Productivity Decomposition