CHERRY Discussion Paper Series CHERRY DP 13/1

The emergence of a European region: Business cycles in South-East Europe from political independence to World War II

By

Matthias Morys Martin Ivanov The emergence of a European region: Business cycles in South-East Europe from political independence to World War II

Matthias Morys Martin Ivanov Department of Economics Institute of History University of York Bulgarian Academy of Sciences

October, 2013

Abstract

Relying on dynamic factor business cycle indices for five South-East European countries (Austria(-Hungary), Bulgaria, Greece, Romania, Serbia/Yugoslavia), we document steadily increasing synchronisation as part of a pan-European business cycle before 1913 and the emergence of a regional business cycle (including and radiating from Germany) in the interwar period. These dynamics were largely driven by trade, involving initially England, France and Germany but increasingly centred on Germany. Our results also show that the Balkan countries travelled a long way from an economic backwater of Europe in the 1870s to a much more integrated part of the European economy six decades later.

Keywords: South-East European business cycle, national historical accounts, common dynamic factor analysis

JEL classification: N13, N14, C43, E32

______Earlier versions of this paper were presented at the EHA, EHES, EHS, IEHA and WEAI meetings and to the 5th Conference of the South-East European Monetary History Network as well as to seminars and workshops in Belgrade, Cologne, Groningen, Leuven, Lund, Münster, Oxford, Strassbourg, Tübingen and York. We are grateful to the participants for their spirited discussion and helpful suggestions. We owe a special thanks to the following people in helping us collect the data and/or sharing their data with us: Rumen Avramov, Olga Christodoulaki, Giovanni Federico, Georgios Kostelenos, Georgios Mitrofanis, Max-Stephan Schulze, Coskun Tuncer, Jeff Williamson, all members of the data collection task force of the South-East European Monetary History Network (SEEMHN), and in particular Alina Blejan, Brandusa Costache, Ljiljana Durdevic, Dragana Gnjatovic, Sofia Lazaretou and Thomas Scheiber. We are grateful to Mario Jovanovic, Rainer Metz and Martin Übele for discussion on the econometrics of this paper. The usual disclaimer applies. Financial support from the British Academy is gratefully acknowledged. - 2 -

1. Introduction South-Eastern Europe has largely remained terra incognita for new economic history. While quantitative economic history has made important contributions for Central and Eastern Europe in recent years – even if most of the research remains confined either to the large countries of Russia and Poland (as exemplified by the work of Andrei Markevich and Nikolaus Wolf) or to relatively advanced economies such as Czechoslovakia (Klein&Broadberry 2011) – , little comparable work is available for the Balkan countries. The only exception to this is arguably Greece, where the country’s Western allegiance during the Cold War allowed the US-inspired cliometric revolution to shape economic history research to a certain extent, as evidenced by the work of Sofia Lazaretou, Olga Christodoulaki and George Kostelenos. There is a considerable amount of qualitative economic history research on South-Eastern Europe (SEE in the following), but this body of literature is mostly confined to individual countries (Lampe 1986 for Bulgaria; Mazower 1991 for Greece; Lupu 1968, Chirot 1976, Turnock 1986 and Hitchins 1994 for Romania; Singleton 1876 and Singleton&Carter 1982 for Yugoslavia). The historiographical traditions in the region, which go back to the late 19th century and are to a considerable extent rooted in the political needs of the nascent Balkan nation states to portray their political, economic and cultural development as unique, apparently acted as an obstacle to cross-country research the way it became fashionable in Western academia after World War II, though there are some notable exceptions (Lampe&Jackson 1982, Palairet 1997). Our paper, therefore, breaks new ground along two dimensions. First, it introduces the powerful concepts of new economic history to SEE. Second, it attempts to overcome the tradition of country-specific historiographies by treating the Balkan peninsula as one unit; or, to be more precise, it asks, from the perspective of business cycle research, whether the Balkans should be seen as an economic region in the sense of exhibiting a self-contained or “regional” business cycle.

At the same time, our work also seeks to contribute to a recent shift in business cycle research. For a long time, it was conventional wisdom that past levels of cross-country synchronisation had been very low. For the period 1870-1913, for instance, several studies found virtual zero correlation even between the so-called core economies of - 3 -

England, France and Germany (Backus&Kehoe 1992, Bordo&Helbling 2011, Artis et al. 2011). These results always sat oddly with our general understanding of the period which (often also referred to as the First Age of Globalisation) was characterised by highly integrated factor and product markets as well as a quasi-universal system of fixed exchange-rates in the form of the gold standard (Daudin&Morys&O’Rourke 2010); all factors potentially conducive to high levels of business cycle synchronisation. A solution to this so-called “business cycle paradox” has recently been proposed by a group of researchers centred on Albrecht Ritschl and Martin Übele. Based on the idea that historical GDP data are not suited for the analysis of cyclical activity, they derive business cycle indices based on dynamic factor methodology and show that cross-country synchronisation levels between the European core countries were high for the period 1870-1913 (Übele 2011). The aim of our paper, then, is to extend this new line of historical business cycle research into a peripheral region of Europe. None of the European regions is better suited for such a study than SEE: we have continuous data for five neighbouring countries (Austria(-Hungary), Bulgaria, Greece, Romania, Serbia/Yugoslavia) since the mid-1870s and early 1880s. This compares favourably with potential competitors such as the four Nordic countries or Iberia (two countries), let alone Central and Eastern Europe where statehood of present-day countries materialised only after the break-up of the Habsburg and Romanov empires at the end of World War I. On the Balkan peninsula, the decline of the Ottoman Empire and the concomitant rise of Balkan nationalism resulted in early independence for Greece (1832), followed by Romania (1859) and Serbia at the Congress of Berlin in 1878. In the same year, Bulgaria obtained far-reaching autonomy within the Ottoman Empire, releasing the country into independence in all but name (Mazower 2001).

In order to appreciate our choice of countries, a proper definition of “the Balkans” and “South-East Europe” seems in place. The Balkans are conventionally defined as the South-Eastern European part of continental Europe demarcated by the Danube, Sava and Kupa rivers to the North and the West (Todorova 2009). The Ottoman legacy (or even presence), the predominance of the Orthodox faith and the high levels of multi- ethnicity all created a sense that the Balkan peninsula was a region very different not only from Western Europe but also from Central and Eastern Europe. According to - 4 - this geographic definition, Bulgaria and Greece are certainly Balkan countries, but some clarifications are appropriate for Romania, Serbia/Yugoslavia and Austria(- Hungary). While most of the Romanian territory lies north of the Danube, the country is conventionally considered part of the Balkans, partly because of the Dobrudja region south of the Danube, but mainly because of the Ottoman legacy which it shares with the other Balkan countries but does not have in common with its other neighbours (Todorova 2009). As for Serbia, the country was fully located on the Balkan peninsula before World War I, but the Kingdom of Serbs, Croats and Slovenes – renamed as Yugoslavia in 1929 – involved parts that are not considered part of the Balkans (Slovenia, Vojvodina) or at least not in their entirety (Croatia north of the Sava river). As the Yugoslav data we rely on does not distinguish between different parts of the country, we decided to include Yugoslavia for the interwar period rather than leave it out completely. The greatest concern arguably relates to Austria-Hungary and interwar Austria. Austria-Hungary had a considerable footprint in the Balkan peninsula through Dalmatia (Austria), the Balkan parts of Croatia (Hungary) and Bosnia and Hercegovina (jointly administered since 1878); it also shared land borders with Serbia, Romania and the Ottoman Empire. Crucial in our context, our results show that understanding the business cycle dynamics of the other four countries is impossible without taking Austria-Hungary into account; which suggests including the dual monarchy irrespective of the country’s exact relationship to the Balkan peninsula. As for the interwar period, we include Austria for reasons of consistency with the earlier period, but it is worth emphasizing that our interpretation for the other four “genuine” Balkan countries remains unaffected if excluding Austria. We were left with no other choice than excluding the remaining three European provinces of the Ottoman Empire (1878-1912//13), Montenegro (1878- 1922) and Albania (since 1913) for the complete absence of or the lack of reliable data. We do not include Hungary for the interwar period, as the country lies in its entirety outside of the Balkan peninsula and has neither seen itself nor has it ever been seen by outsiders as part of the Balkans (Todorova 2009). Last but not least, we use the terminology “South-East Europe” and “the Balkans” interchangeably, but we note that some authors see SEE as potentially a slightly wider geographical notion. - 5 -

The remainder of the paper is structured as follows. In the second section, we will outline the problems inherent to a business cycle reconstruction based on national historical accounts and how recent studies involving dynamic factor methodology have addressed them. In the third section, we will explain the specific dynamic factor model and estimation procedures we follow; we will also outline the time series used for the purpose of constructing business cycle indices. In the fourth section we will then address the three main questions of this paper: (1) To what extent was there a common SEE business cycle? (2) Was there synchronisation of business cycles with England, France and Germany? (3) What drove the business cycle and how do we account for different levels of business cycle synchronisation across countries and over time? An answer to all three questions will allow us to locate the Balkans’ position as part of the wider European economy and how this changed over the six decades from political independence to World War II. The fifth section summarises and concludes.

2. Pitfalls of a business cycle reconstruction based on historical national accounts In a perfect world, we would study business cycles by analysing GDP data on annual or quarterly frequency. In this section, we will explain why historical national accounts are not as helpful for this purpose as they initially appear. Our concerns partly stem from the idiosyncrasies of SEE GDP data, partly from general considerations as to why historical national accounts are unlikely to reflect the true but unknown GDP series. The most obvious limitation of SEE GDP data refers to the period 1870 – 1913. GDP estimates on an annual basis are available only for Austria-Hungary (Schulze 2000), Bulgaria (Ivanov 2012), Greece (Kostelenos et al. 2007) and Romania (Axenciuc 2012), of which, however, only the data for the dual monarchy has made it into the Maddison (2009) data set. There are no similar data available for Serbia. The pre-WW I SEE GDP data reported by Maddison (2009) are on a decadal basis only (except for Austria-Hungary); moreover, the data do not constitute genuine GDP data but the results of proxy estimates by Good&Ma (1999).1 For the interwar period, Maddison (2003) reports GDP data for all five SEE countries. If the detailed critique

1 The Greek case is somewhat different; for details cf. Morys (2006). - 6 - of the Maddison data for Bulgaria by Ivanov&Tooze (2007) has implications for other countries (as is likely), then we have good reason to be equally sceptical towards the interwar data reported for Greece, Romania and Yugoslavia. But even discounting the idiosyncrasies of SEE GDP data, an argument can be made for relying on DFA rather than historical national accounts. These considerations have led to the use of DFA even for countries such as the US (Ritschl et al. 2008) and Germany (Sarferaz&Übele 2009) for which much more reliable GDP are available. First, national historical accounts are normally constructed with an eye for the level rather than the volatility; this (understandable) preference determines interpolation techniques which can lead to serious differences in volatility between the reconstruction and the true but unknown GDP series. Second, disaggregate series are often abundant for historical periods, but in many cases do not match national accounting categories very well; DFA allows us to exploit the business cycle characteristics of these series. Third, DFA deals better with structural breaks in sub- series than GDP, as DFA is more flexible in excluding disaggregate time series with serious faults.2 All three issues raised are likely to be of more concern the further we go back in time. Table 1 provides an overview of business cycle synchronization before WW I according to previous research, differentiating between (1) intra-core, (2) intra- periphery and (3) core-vis-à-vis-periphery. In the case of the latter two categories, we report synchronization for peripheral countries emanating from the same region provided such information could be extracted from the publication. As Scandinavian countries are overrepresented in the studies (as a result of better GDP), we report results for Sweden, Denmark, Norway and (occasionally) Finland to provide an idea of synchronization levels for a peripheral region before WW I.

[Insert Table 1 about here]

The column to the right in table 1 gives the statistical method and the underlying data. As for the statistical method, we see that most of the research relies on GDP. The only exceptions are an early study by Morgenstern (1959) following the NBER methodology and a recent paper by Übele (2011).

2 For a more detailed comparison of both techniques cf. Ritschl et al. (2008) and Aiolfi et al. (2006). - 7 -

Focusing on the GDP-reliant studies for the moment, we find that none of the average correlations reported exceeds 30% and many of them are below 10%; particularly astonishing are the low results for intra-core correlations none of which surpasses 10%. These results contradict our expectations: the period 1870 – 1913 was characterised by highly integrated factor and product markets as well as a quasi- universal system of fixed exchange-rates in the form of the gold standard (Daudin&Morys&O’Rourke 2010); factors which should all be conducive to a high level of business cycle synchronization. This so-called business cycle paradox (Bordo&Helbling 2011; Artis et al. 2011) can be solved in two different ways: either the correlation between business cycle synchronization and market integration is not necessarily positive. Krugman (1993), for instance, noted that stronger trade integration may lead to greater regional specialisation, which can lead to less output synchronization due to industry-specific shocks. Alternatively, the underlying data – i.e. GDP – poses the problem. Table 1 also includes Morgenstern’s (1959) work which follows the NBER tradition in not relying on GDP but, instead, on reference cycles established by drawing on a multitude of time series. Using NBER reference cycles, Morgenstern showed that the UK, France and Germany were in phase in fully 83 per cent between 1870 and 1913. He also found the US cycle to be in phase with the European cycle in 54 per cent of months. In earlier work, the idea of a closely integrated pan-European business cycle had already found statistical support by Mitchell (1927), and Kuznets (1958) came to a similar result for a sample of Atlantic economies. In sum, the earlier work by Mitchell (1927), Kuznets (1958) and Morgenstern (1959) is indirectly supportive of our concerns over using GDP data for the purpose of business cycle studies. And yet there is another parallel between the early work on business cycles pioneered by the NBER (Burns&Mitchell 1946) and our approach. The NBER approach relied on a multitude of time series which were condensed into a (country-specific) “reference cycle”. Given the computational limitations at the time, such an approach was very cumbersome and involved more discretion on behalf of the researcher than was deemed appropriate; it was effectively abandoned, when modern (post-WW II) macroeconomics began building models centered around a small - 8 - number of well-defined national accounting variables such as Y, I, G etc.3 As a result of these very fundamental changes in the discipline of economics, business cycle research became increasingly focused on GDP as the most important (or even the only) business cycle indicator. As computational restrictions gradually eased, interest in calculating business cycles à la Burns&Mitchell (1946) resurfaced and led Geweke (1977), among others, to pioneer DFA.

3. Explaining and applying common dynamic factor analysis

The model The Common Dynamic Factor Model is best understood behind the background of its parent model, the static factor model, which takes the following form:

(1) yt = λ0 + λ ft + εt (n x 1) (n x 1) (n x o) (o x 1) (n x 1)

yt is a (n x 1) vector of variables yit with = 1 to n (with n = 20 in our case) and t =

1875 to 1941. Factor models posit that the different yi’s (i.e., the different time series) are explained partly by a common component and partly by a variable-specific (or idiosyncratic) component. This dual structure is captured in the second and the third summand (the first summand simply being a vector of variable-specific intercepts). ft is a (o x 1) vector of unobserved latent factors (where o < < n) which are common to every yi though different factors might be of different importance for different yi’s (as factor loadings – i.e., the entries in the (n x o) matrix λ – may differ in different rows).

The third summand – i.e., the different εi’s – capture the idiosyncratic component. Static factor models are turned into dynamic factor models by allowing for dynamic properties typical of macroeconomic variables. As we are dealing with a common and an idiosyncratic component, the necessary extensions relate to the second and the third summand of (1) which we re-write below for individual yi’s.

3 It is not entirely clear how causality runs: one of the reasons why modern macroeconomics was embraced so readily in the 1950s was arguably because it reduced computing needs by agreeing on a small number of variables. We thank Paul David for drawing our attention to what the fundamental changes in economics at the time meant for the sub-field of business cycle studies. - 9 -

(2) yit = λ0i + λi ft + εit (1 x 1) (1 x 1) (1 x o) (o x 1) (1 x 1)

Different dynamic properties can be assumed but there is little reason to deviate from standard assumptions according to which ft follows a VAR(p) process while εit follows a AR(q) process. To emphasize the different nature of the autoregressive process (vector versus scalar), we choose Φ (o x o) and φ (scalar), respectively.

(3) ft = Φ1 ft-1 + Φ2 ft-2 + … + Φp ft-p + ζt (o x 1) (o x o) (o x 1) (o x o) (o x 1) (o x o) (o x 1) (o x 1)

(4) εit = φi1 εi,t-1 + φi2 εi,t-2 + ... + φiq εi,t-q + ηit (all variables are scalars)

The model requires several identifying assumptions, the most important of which is that we assume ζt to be i.i.d. N(O, In). For more details cf. Koop&Korobilis (2010: 52- 53).

We interpret the first factor (i.e., the first element of the (o x 1) vector ft) as the business cycle for the specific country under investigation; “first factors” in DFA is the equivalent to the first principal component in principal component analysis, i.e., it explains more of the covariance structure than the second factor, which, in turn, explains more than the third etc. DFA studies differ as to whether they extract only one factor (Sarferaz&Übele 2009) or more than one factor (Aiolfi et al. 2011). The standard procedure for testing for the number of factors by Bai&Ng (2002) suggests in our case extracting only one factor, thereby avoiding the often difficult question what interpretation to give to the second factor (as conceded by Aiolfi et al. 2011: 215).

Estimating the model Classical as well as Bayesian procedures are available for DFA (Aiolfi et al. 2011 versus Sarferaz&Uebele 2009). We have opted for the Bayesian approach, as it tends to deliver superior results when dealing with time series of limited length; due to poor - 10 - data reporting in World War I, estimation needed to be carried out separately for the pre-WW I and interwar periods. Calculations were carried out with Matlab relying on a code developed by Koop and Korobilis.4 p and q were set at 8 and 1, respectively; different assumptions were tried out but our findings hardly changed. Raw data series were transformed into logarithms (except for domestic interest rate, terms of trade and real effective exchange rate where levels were used) and subjected to the Hodrick-Prescott filter (with a smoothing parameter λ = 6.25). The resulting 20 cyclical series (for each country) were then standardised by adjusting the mean to naught and the standard deviation to unity; this step is crucial in ensuring that each series yi is given equal weight in establishing the business cycle.

Robustness checks The main concern from an econometric perspective does not relate to the exact specification of the CDFA model but to the filtering techniques used to transform the raw data into yi. More specifically, econometricians have long argued that filters might generate cycles not present in the underlying data, potentially resulting in spurious correlations between time series subjected to the same filtering technique (inter alia Metz 2009). To address this issue, we therefore carried out the same calculations with three different filters: Hodrick-Prescott (both with λ = 6.25 and λ = 100), Baxter-King and simple first differences (all with respect to log data), but results hardly changed. In the following, we will report results based on the HP-filter (with λ = 6.25) which probably remains the most widely used filter in macroeconomics, thereby allowing better comparability with other studies.

Data We include 20 time series for each country, ranging from sectoral output indicators over fiscal and financial variables to trade data (table 2). There is considerable variety among DFA studies on the number and the characteristics of time series to be included (Sarferaz&Übele 2009, Ritschl et al. 2008, Aiolfi et al. 2011); we opted for a “conservative” approach in that (a) time series needed to be identical across countries

4 The code can be downloaded on Koop’s webpage under “MATLAB Code for Factor Models” under http://personal.strath.ac.uk/gary.koop/bayes_matlab_code_by_koop_and_korobilis.html - 11 - and (b) economic theory would need to suggest some connection to the business cycle (ruling out theoretically loosely underpinned time series such as “purchases of lipstick” that have figured prominently in applied work for very recent periods, see The Economist, 29th July 2011). This approach left us with a medium size list of 20 variables which could all, with very minor exceptions (tables 5, 11), be located for the 5 SEE countries and 3 core countries for the entire period of 1875 to 1941.

[Insert Table 2 about here]

Our data collection efforts show that “new economic history” can easily be written for this part of Europe, largely drawing on Statistical Yearbooks (publication of which started shortly after political independence; except for Greece, where Yearbooks only appear in the interwar period but then provide data for earlier periods as well). For England, France and Germany we were able to rely on readily available data. A full summary of the sources, including the exact estimation period for each country, can be found in the data appendix. Last but not least, when several data series for the sectoral indicators were available, we opted for time series given in units such as weight, volume, kilometres or simple count data (for instance, number of letters for communication) rather than time series denominated in prices. We avoided the latter category wherever possible, as the business cycle should reflect real economic activity rather than be driven by commonality of price movements across different types of economic activity. - 12 -

4. Business Cycles in South East Europe, 1875 – 1941

Our measure of business cycle synchronization Business cycle synchronisation can be measured in different ways. One methodology involves spectral analysis; based on Fourier transformation, a cycle A is dissected into a multitude of cycles of different periodicity and then compared to a similarly decomposed cycle B. Spectral analysis can, for example, establish that two time series are highly correlated at one periodicity but less so at another. This technique, then, is applied especially in cases (e.g., A’Hearn&Woitek 2001) in which the research tries to distinguish between cycles of different length (either suggested by economic theory or empirical work) such as the Juglar (1889) cycles (fluctuations of 7-10 years) and the Kitchin (1923) cycles (fluctuations of 3-4 years). More widely used is the Harding&Pagan (2002) concordance index. Emanating from the NBER tradition of business cycle research, national business cycles are said to be synchronized if turning points in the corresponding reference cycles are at the same time (or at least close to each other); in other words, synchronisation means that national business cycles are in the same phase – expansion or recession – at the same time. The most widely used indicator for business cycle synchronization is Pearson’s correlation coefficient. To our knowledge, most if not all of the recent research into historical business cycles has relied on this indicator (Bordo&Helbling 2011, Artis et al. 2011, Sarferaz&Übele 2009, Übele 2011); which is why we will employ it in the following to allow for direct comparison with earlier work. More specifically, our measure of business cycle synchronization is the bilateral correlation of the (country- specific) business cycle index as calculated by CDFA. - 13 -

4.1 Pre-World War I: 1875 - 1913

Documenting business cycle synchronisation Table 3 shows business cycle synchronization among the eight countries under investigation, i.e. the SEE-5 and England, France and Germany. For each country pair we provide correlations for three different periods: the full period (which is given by pair-wise intersection of the estimation period, cf. data appendix); 1893-1913; 1903- 1913. While shortening the period makes it more difficult to ascertain statistical significance, it will become clear in the following why reducing the full period is warranted. Table 4 provides summary statistics of the (8 * 7 ) / 2 = 28 bilateral correlations (and sub-groups thereof) of table 3.

[Insert Table 3 about here]

[Insert Table 4 about here]

Tables 3 and 4 reveal an increase in business cycle synchronization the closer we come to WW I. Comparing the full period with the last decade, the median and the average increase from 0.27 to 0.56 and from 0.34 to 0.44, respectively, and the number of statistically significant correlations increases from 12 to 16 (notwithstanding the fact that shorter periods require higher correlations to establish statistical significance). The results are novel both in terms of synchronization levels as well as changes thereof over time. First, business cycle synchronisation in South-East Europe, one of the poorest parts of the European peripheries, was by no means negligible. Our average value of 0.34 compares “favourably” to the 0.03 and 0.02 reported in Bordo&Helbling (2011: 212) and Artis et al. (2011: 186) (both of whom rely on larger samples for the 1880- 1913 period but use GDP instead). Following earlier work that helped solve the business cycle paradox for core countries (Übele 2011), our study is the first of its kind to document similar patterns for the European periphery. As for the increase in synchronization over time, our results confirm a proposition advanced by Bordo&Helbling (2011) but add nuance to it. Bordo&Helbling (2011) find a secular trend towards increased synchronization from 1880 to the present day. - 14 -

While they document this increase from one period to the next (gold standard; interwar period; Bretton Woods; modern floating era)5, our results indicate that increases also occurred within periods. If we expect the correlation between market integration and business cycle synchronization to be positive, this finding should not come as a surprise; studies on market integration during the First Age of Globalization have shown markets as increasingly integrated the closer we come to World War I (O’Rourke&Williamson 1999, Daudin&Morys&O’Rourke 2010).

Intra-core correlations We begin by analyzing business cycle synchronization among the core countries to allow comparison with earlier research (table 1). For the period 1879-1913, England, France and Germany exhibit an average correlation of 0.80. France and Germany are marginally more correlated with each other than with England which supports the concept of a more closely integrated continental European economy often encountered in business cycle research (Artis et al. 20116). These results are broadly in line with Übele (2011), who finds an average value of 0.61 for the three core countries based on a similar methodology but a slightly longer period (1862-1913).7 Taken together, they lend credence to the new view that the business cycle paradox is the result of overreliance on GDP, and hence best addressed by drawing on disaggregate data and estimation techniques such as dynamic factor methodology. Figure 1 shows that peak and troughs were mostly not further apart than one year (and often identical), and several turning points are reminiscent of key events such as the business cycle upswing of the late 1880s (culminating in and reversed by the Baring crisis of 1890) and the American banking crisis of 1907 (which was preceded by a long upswing) (Kindleberger 2005, Übele 2011).

[Insert Figure 1 about here]

5 The only exception they find relates to the transition from the interwar to the Bretton Woods period where synchronization levels hardly change. 6 While this is one of the main findings of Artis et al. (2011), it is worth pointing out that they can establish a specifically continental European business cycle only for the post-WW II period. 7 Private communication of the authors with Prof. Übele reveals that his correlation values increase if analysing the period 1879-1913 only. This finding for the core countries is supportive of our argument in the main text that business cycle synchronisation increased over time. - 15 -

Intra-SEE-5 Having provided benchmark values for the more advanced core economies, we now turn to the 10 bilateral correlations of the SEE-5. A pattern of increasing business cycle synchronization is confirmed, with the average correlation rising from 0.24 (full period) to 0.26 (1893-1913) and 0.38 (1903-1913); Austria-Hungary, Greece, Serbia and Romania all increase their bilateral correlations over time. Such a trend is easily detected in figure 2 which focuses on Austria-Hungary, Greece and Romania as the three countries with the longest time-series due to earlier political independence. For some of the earlier periods, business cycles nicely track each other (inter alia, Austria- Hungary and Greece between 1875 and 1886), but for others such a pattern is less clear-cut or lags between the different cycles are longer. Patterns of synchronisation increase towards the turn of the century and reach their highest level in the decade preceding World War I (figure 3), a period in which peaks and troughs for all countries either coincide or are no more apart than a single year. For the latter period we recognise upswings and downswings which are well-documented for other countries (including England, France and Germany, figure 1): (a) the long upswing in the early 1900s, a period of global boom which ended in the bust of the American banking crisis of 1907; (b) the impact of the American banking crisis on the rest of the world; (c) another upswing starting in 1909/10 which lasted until WW I.

[Insert Figure 2 about here]

[Insert Figure 3 about here]

It is worth highlighting two issues: first, the impact of the American banking crisis (failure of the Knickerbocker trust in October 1907) was both instantaneous and prolonged. While the impact of the American banking crisis on some countries on the European periphery has been studied in some detail (in particular regarding Italy, cf. Bonelli 1971), little was known for SEE except for their needs to protect their currencies through discount rate increases and foreign exchange policy at the height of the crisis (Morys 2013). We document that all countries in the region reached their peak in 1907 and their troughs two years (Greece, Romania) or three years (Austria- Hungary, Bulgaria, Serbia) later. - 16 -

Second, the global upswing after 1909/10 – which lasted for most countries until World War I or even into the war (for instance in the case of the UK and many neutral countries, cf. Ritschl&Straumann 2010) – was cut short for all SEE-5 by the Balkan Wars (1912-13), as revealed by a common peak in 1912.8 Our results also indicate differences among the SEE-5. Austria-Hungary and Serbia – neighbouring countries at the Western edge of the Balkans – not only exhibit the strongest correlation with each other but are also more strongly synchronized with the rest of the region, followed by Romania and Greece. Bulgaria appears least well synchronised, though results seem to be driven by the bilateral correlation with Greece with which it does not share a common border until the Balkan Wars; the bilateral correlations with Austria-Hungary, Romania and Serbia remain positive throughout. We will explain the differences between the individual countries below but we will first analyse the 15 bilateral correlations between the SEE-5 and the three European core economies: do we find the same two characteristics just described, i.e., increasing synchronisation over time and sizeable differences between Austria- Hungary, Bulgaria, Greece, Romania and Serbia?

Synchronization of SEE-5 vis-à-vis England, France and Germany Our results indicate increasing levels of synchronisation of SEE vis-à-vis the core countries, with the average of all 15 bilateral correlations increasing from 0.32 for the full period to 0.40 for the restricted one. Similar to the correlations intra-SEE, there is a difference between a more synchronized Austria-Hungary and Serbia and the other three countries. Austria-Hungary constitutes a special case, as it is the only SEE country better integrated towards the West than towards the (South-)East. Interestingly, synchronisation between Bulgaria, Greece, Romania and Serbia and their synchronisation vis-à-vis the core economies are comparable: 0.24 vs. 0.32 for the full period and 0.26 vs. 0.28 and 0.38 vs. 0.40 for the two shorter periods. This suggests that there was no regional SEE business cycle at the time (in the sense of a group of contiguous countries enjoying synchronization levels substantially elevated to what they exhibit vis-à-vis countries outside of this group). Rather, the “Janus- faced” position of the dual monarchy appears pivotal. Austria-Hungary’s

8 This finding is consistent with Lampe&Jackson (1982) who see 1912 rather than 1914 as the fundamental caesura in SEE economic history. - 17 - synchronisation with the core countries does not change much over time, remaining at levels of approximately 80% and hence comparable to what the core countries have among each other. On its South-Eastern borders, the dual monarchy sees its synchronisation levels increase over time, reaching an average value of 0.50 after the turn of the century; the closer the country to the dual monarchy, the higher the value. The picture emerging is one of SEE increasingly participating in a pan-European business cycle which emanated from the core countries and for the diffusion of which into SEE Austria-Hungary became increasingly important. Last but not least, we conducted Wilcoxon rank sum tests for equality of correlations to bolster confidence in our results. They show that the increases of the correlations of the SEE countries among each other and vis-à-vis the core countries are statistically significant but that the synchronisation changes for the core countries are not.9 This supports our interpretation of a stable pan-European business cycle with which SEE increasingly synchronised.

Explaining business cycle synchronisation How can we account best for cross-country and inter-temporal variation of synchronisation patterns in SEE for the period 1875 – 1913? In the following, we will draw on the 20 time series underlying each business cycle to shed light on the structure of the SEE economies and analyse what drove their cycles. We will argue that (cross-country) synchronisation was largely driven by trade, with countries geographically positioned better to Western Europe and better equipped with infrastructure enjoying higher levels than others. Table 5 shows correlation patterns between each (national) business cycle and the underlying 20 time series. The left column for each country provides the value of the contemporaneous correlation. The middle column gives the value of the maximum correlation comparing the two time series at five points in time, i.e., -2, -1, 0, +1 and +2. This aims at detecting lead and lag patterns between the business cycle and the time series underlying it; the window length of four years was chosen to capture at least half of a typical cycle of eight years length. By “maximum correlation” we mean

9 The null-hypothesis of equality of correlations across periods is rejected at the 5%-level for the SEE countries while it cannot be rejected at any conventional level for the core countries (p-value of 0.66). - 18 - correlation in absolute terms, so as to capture the strength of association rather than confining our search to positive values only. The right column for each country then provides the lag at which the correlation is maximised (in absolute terms). A positive number indicates a leading factor, whereas a negative number means that the cycle leads the underlying time series (which is then said to be a lag-factor).

[Insert Table 5 about here]

Sectoral indicators Agriculture is not strongly correlated with the business cycle in any SEE country. This might surprise given the importance of the primary sector in SEE which even in Austria-Hungary accounted for 31.3% of overall GDP as late as 1913 (Schulze 2000, pp. 337-338). Our finding is in line, however, with much of the business cycle literature which has emphasised that business cycles are a phenomenon specifically related to the secondary and tertiary sector.10 As for industrial production, all contemporaneous correlations are safely in positive territory and with an average value of 0.46 it constitutes the third most closely correlated indicator (after transportation and narrow money, in that order). Cross-country differences are revealing. The elevated level of 0.63 for Austria- Hungary might reflect that it was the only SEE country boasting a sizeable secondary sector. Schulze (2000, pp. 337-338) estimates that industry accounted for 22.2% and 31.8% of overall GDP in 1870 and 1913, respectively, with a more industrialised Austria (1870: 27.9%; 1913: 37.4%) compared to a more agricultural Hungary (1870: 10.7%; 1913: 21.9%); by contrast, Greece’s relatively low value of 0.23 mirrors an economy devoid of a meaningful industrial sector which accounted for only 7.7% and 17.2% of overall GDP in 1870 and 1913, respectively (Kostelenos 2007, table 6-2). Transportation – proxied by the volume of freight traffic – exhibits a particularly high correlation and constitutes one of the best four indicators for all countries bar Greece (for which, incidentally, no freight related proxy was available). It tracks the business cycle more closely than industry in the cases of Austria- Hungary, Bulgaria, Romania and Serbia. Only marginally inferior as indicators are fixed investment and construction, much of which went into the built-up of transport

10 See already the classical definition of Burns&Mitchell (1946: 3) on this. - 19 - infrastructure given the economic needs of Balkan countries at the time. The overarching importance of transport-related indicators is suggestive of an economic structure different from the industrialised economies of North-Western Europe. Given the pre-dominance of the primary sector in the SEE economies, freight traffic often meant the transport of agricultural goods many of which were valuable export commodities. The strong correlation of transportation, though evidently an economic activity of the tertiary sector, indirectly demonstrates the importance of the primary sector: bringing agricultural produce to domestic and international markets was important for the upswings and downswings of the economy. The other two sectoral indicators – communication and mining – are mostly positively correlated to the business cycle, with average correlations only slightly lower than industry, transportation, fixed investment and construction. Values are lower presumably because mining was only tangential to overall economic activity in SEE (with the exception of the Romanian oil industry, cf. Jordan 1955 and Pearton 1971) and because communication – which we proxy by letters sent – also captures a great deal of private correspondence and other commercially not relevant communication.

Fiscal indicators Fiscal indicators show mixed results, with expenditure largely independent of the cycle and revenue mildly pro-cyclical. These inconclusive results probably reflect the small share of government in aggregate economic activity before World War I (ca. 8 – 12%); this interpretation is supported by a comparison with the interwar period where both indicators are strongly pro-cyclical and the government share is much higher.

Monetary indicators Narrow and broad money exhibit some of the highest contemporaneous correlations; in the cases of Greece (0.88 and 0.87, respectively) and Romania (0.94 for both), they constitute the two indicators most closely correlated with the business cycle. While some of this might be attributable to data quality (especially in the case of M0, which was taken from accurately reported central bank balance sheets), it is also testimony to the successful efforts of the Balkan countries to establish banks of note issue immediately after obtaining political independence and to gradually move from transactions based on barter and coin to monetized economies (Morys 2009). Price - 20 - level changes are positively correlated with the cycle on a contemporaneous basis; this association strengthens further one period later, suggesting that CPI is a lag indicator of overall economic activity. The short term interest rate appears neutral to the cycle, presumably reflecting infrequent discount rate changes on the European periphery (Morys 2013); this interpretation is supported by the fact that the one country in SEE with a more frequent use of the discount rate – that is, Austria- Hungary – shows the expected positive correlation (0.55).

Trade indicators Exports and imports are positively correlated with the business cycle for all five countries, resulting in high average correlation values of 0.33 and 0.44, respectively. In the case of imports, it should be kept in mind that many of them were financed with capital inflows (as an analysis of the trade balance – not reported here –demonstrates; cf. also Morys 2006 for Austria-Hungary and Feis 1930 for the other SEE countries)); in other words, in establishing a high correlation value for imports for all countries under investigation, we have indirect evidence that their business cycles were also closely tied to the flow and ebb of foreign capital exports from the European core economies. The importance of trade for the business cycle is further underlined by the positive correlation exhibited by the terms of trade (export prices over import prices) for Austria-Hungary, Bulgaria, Greece and Romania.

What drove synchronisation between countries? Comparing financial and trade indicators provides an answer as to what drove increasing levels of business cycle synchronisation. Interest rate linkages are the main channel through which financial integration can result in business cycle synchronisation; a process more likely to happen under systems of fixed exchange- rates such as the gold standard as a result of increased correlations between interest rates (Obstfeld et al. 2005). Synchronisation through trade can be thought of as an expansionary phase in one economy stimulating demand for foreign goods, thereby translating into an export boom for other economies. Prima facie, the (average) zero correlation between the short term interest rate and the business cycle suggests that interest rate linkages cannot have played an important role for cross-country synchronisation, as interest rates failed to drive even - 21 - the domestic cycle. By contrast, synchronisation through trade looks more plausible, with exports, imports and terms of trade all positively correlated with the cycle. There is another way of testing the two competing views. As for interest rate linkages, the correlation between the interest rate cycle of SEE countries with the cycle of England and Germany, the two European interest rate setters (Morys 2013), can be established; as for trade linkages, the correlation between the export cycle of SEE countries with the import cycle of England, France and Germany can be measured. Both tests support the view that business cycle synchronisation in SEE was the result of trade and not of financial integration. As for interest rate linkages, only in the Austro-Hungarian case there is a meaningful positive correlation with the England and the German cycle; a correlation which increases slightly after joining the gold standard in 1896 (average value of 0.64 for 1875-1895 compared to 0.83 for 1896- 1913). Romania also witnessed an increase in correlation after joining the gold standard in 1892, but the value remains low (average value of 0.25) and is close to zero if the entire period 1881 – 1913 is considered (average value of 0.09). In the case of the three other countries – all of which joined the gold standard only after the turn of the century (Bulgaria: 1906; Serbia: 1909; Greece: 1910, cf. Morys 2009) – , the correlation is either negative or close to zero, with little difference if considering the entire period or the period before joining the gold standard only.11 Overall, the low and often negative correlations with the core countries’ interest rate cycle indicate that synchronisation did not happen through this route; if at all, the low correlations point to the opposite conclusion and can help explain why the SEE joined the gold standard so late. By contrast, correlating the SEE-export cycle with the core countries’ import cycle leads to positive results in all cases (Austria-Hungary: 0.48; Greece: 0.54; Romania: 0.11; Serbia: 0.18) except for Bulgaria (-0.07), which, incidentally, also appeared as the least synchronised SEE-country in our findings above. We conclude this section with some observations on trade openness, trade patterns and infrastructure in SEE which will help understand differences between the 5 SEE countries. This also aims at putting SEE into the wider perspective of the European peripheries in the First Age of Globalisation. That trade enabled but also

11 Average values for the entire period and the pre-gold standard period are: Bulgaria: -0.27 and -0.14; Serbia: -0.37 and -0.36; Greece: 0.05 and 0.06. - 22 - limited synchronisation becomes clear on inspection of trade data. The most aggregate figure is trade openness, which we have calculated for the SEE countries and compare to readily available data for other European peripheries, namely Southern Europe, Eastern Europe and the Nordic countries (table 6). The general pattern emerging is that large peripheral countries (Russia, Italy, Spain) are less open (as a result of a large domestic market) than small peripheral countries; and that countries become more open over time. The SEE countries buck this trend on both accounts: they are more closed than any other small peripheral country and trade growth does not outstrip GDP growth. Furthermore, focusing on the “small” countries, trade openness and synchronisation of the SEE export cycles with the core countries’ import cycles lead to the same ranking: Greece, Serbia and Romania, followed by Bulgaria.

[Insert table 6 about here]

Disaggregating trade data by countries sheds more light on the difficulties facing SEE: To begin with, Bulgaria, Greece, Romania and Serbia traded remarkably little with each other. In any given benchmark year, each of the four countries had with the other three countries combined a trade share of no more than 7%12; in some cases, trade was so negligible that statistical agencies did not report the data at all, as was the case for Greece which began including Serbia only after the Balkan Wars (when both countries began having a common border). The low trade volumes do not reflect any specific protectionist measures for agricultural commodities (which account for the bulk of trade among the four countries); it rather is the result of a similar economic structure which did not lend itself to trade among each other but to trade with economically more advanced Austria-Hungary and the industrialised core economies of England, France and Germany. The trade the four countries enjoyed with Austria- Hungary stands, averaged over the benchmark years, at 58.5% for Serbia, 23.1% for Romania, 18.3% for Bulgaria and 12.2% for Greece (average for SEE-4: 28.0%). Yet these numbers appear small again (with the exception of Serbia) when compared to the trade that the four countries conducted with England, France and Germany taken

12 The only exception to this is the 1882 trade share of Bulgaria with Romania (11.3%): The Principality of Bulgaria, as it emerged from the 1878 Congress of Berlin, did not include Eastern Rumelia (to the South of the Principality) which became part of Bulgaria only in 1885. As a result, trade patterns of Buglaria initially were more strongly focused on the Danubian principalities to the North of it, i.e., Romania. After the inclusion of Eastern Rumelia trade patterns change and follow the general pattern outlined above. - 23 - together. Averaging over the 4 SEE countries and all benchmark years, they conducted 36.3% of their trade with the three European core economies. The same broad pattern is evident for Austria-Hungary: in any given benchmark year, it traded less than 10% with the SEE-4, but consistently more than 60% with England, France and Germany. Trade shares among the SEE countries compared to trade with England, France and Germany then also explain why we do not find a regional business cycle but rather a pan-European business cycle in which the SEE countries increasingly participated.

[Insert table 7 about here]

At the same time, substantial trading links to the “East” remained in the cases of Bulgaria, Serbia and Greece which had a common border with the Ottoman Empire until the Balkan Wars. This was especially true for Bulgaria for which the Ottoman Empire remained the largest trading partner in many single years and comes a close second to Austria-Hungary even when looking at the period from political independence to World War I as a whole. We conclude this section with some reflections on geography and infrastructure in the Balkans. Bulgaria – which we found to be the least synchronised country and the country least open to trade – faced the most isolated geographical position regarding market access to Western Europe. Access to the Aegean Sea was not only impeded by the natural barriers of the Rodopy and Pirin mountains, but also complicated by politics. The Ottoman Empire, in its attempt to hold on to Macedonia and Thrace, deliberately blocked Bulgarian attempts to develop its infrastructure towards the South (Deyanov 2005). To the North, the Danube – which forms the border with the Romania – acted as a natural barrier, with the first bridge built only in 1954 and situated much to the East, connecting Ruse and Giurgiu (Nikova 2007). As a consequence, the new Bulgarian state was left with no choice other than developing its trade ties with Western Europe along the 1st century AD via diagonalis connecting Belgrade, Nis, Sofia, Plovdiv, Edirne and Istanbul. Roads were cumbersome, and the rail link from Sofia to Belgrade and Vienna (as part of the Orient Express connecting Vienna with Istanbul) was completed only in 1888 (Stanev 2011, pp. 27-30). Serbia and Romania (let alone Greece given its sea-bound trade) found themselves in an easier environment to foster trade ties with the West. In the late - 24 -

1880s, Serbia not only possessed a rail line connecting Belgrade and Nis (its two most important cities), which the Ottoman authorities allowed to connect to Skopje and Salonica, giving it access to the Aegean Sea which Bulgaria eluded. Crucial for its access to the West was that Serbia enjoyed access to the railway network covering Croatia, Slovenia and the Vojvodina (all of which were part of Austria-Hungary) which had been rapidly expanding since the 1860s (Stanev 2011, pp. 71-76). Romania, for its part, was also in a more favourable position than Bulgaria. Supported by Western Europe’s long-standing desire to import cheap grain from the fertile Danubian plains, it had connected to the Austro-Hungarian railway network in 1879 through Brasov in Transylvania; only four years later, Romania boasted a complex railway network of 1359 km length compared to Bulgaria with a single line of 223 km.

4.2 Interwar period: 1919 – 194113

Documenting business cycle synchronisation On first inspection, business cycle synchronization appears to have continued in the interwar period hardly unchanged from the immediate period before World War I with an average bilateral correlation of 0.42 (tables 8 and 9). This average number conceals, however, substantial differences (a) between core and peripheral countries and (b) between the periods before and after the onset of the Great Depression in 1929. While there has been more academic interest in the interwar business cycle – often as part of research on the Great Depression (Ritschl 2002) – , there is less of a conventional view compared to the period 1870 – 1913. Two opposing forces are thought to have been at work: on the one hand, if de-globalization – starting with WW I but amplified during the Great Depression – meant an implosion of trade and a reduction of capital flows and migration, this would lead to less closely integrated national business cycles. Similarly, the currency instability of the 1920s (a poorly synchronized process of re-establishing the gold link) and the 1930s (the dismemberment of the gold standard into a multitude of currency blocs) was less

13 For Bulgaria, Romania and Yugoslavia the estimation period stretches beyond 1939 (cf. data appendix) due to hostilities starting only in the 1940s. - 25 - conducive to business cycle synchronization than the period of quasi-universal currency stability preceding WW I. On the other hand, countries worldwide were affected by the Great Depression which can be interpreted as a global shock for the purpose of business cycle studies (Basu&Taylor 1999). As common shocks are one of the factors driving synchronisation, the Great Depression might well have resulted in more closely correlated business cycles. Moreover, the disintegration of the world economy in the interwar period was often coupled with increased integration on a regional level, as exemplified by the currency blocs and preferential trade agreements of the 1930s (Feinstein et al. 2008, pp. 135-159). It is therefore no surprise that scholars have not been able to agree on a business chronology for the interwar period, let alone on a specific interpretation (Ritschl&Straumann 2010). Most recent studies find synchronization levels higher than before World War I but they also point to major regional differences (Backhus&Kehoe 1992, Ritschl&Straumann 2010). Synthesizing a large and occasionally bewildering literature, Ritschl&Straumann (2010) refrain from providing a uniform chronology but point to three major parameters explaining patterns of synchronisation and dis-synchronisation. (1) did countries experience a recession during World War I (the standard case) or a boom? (2) the impact on the business cycle from tying to (in the 1920s) and untying from (in the 1930s) the gold standard; (3) the development of bilateral trade patterns. We will argue that these three factors – and the development of trade patterns in particular – explain well the SEE interwar experience, but we will first present the numbers.

Intra-core correlations Compared to an average value of 0.80 before WW I, the correlation among England, France and Germany is substantially reduced and stands at only 0.47, with 0.29 for the period before 1929 and 0.67 thereafter (tables 8, 9).

[Insert Table 8 about here]

[Insert Table 9 about here]

[Insert Table 10 about here] - 26 -

Differences between correlations before and after 1929 are large and statistically significant (table 10), suggesting that the two periods should be treated separately. Visualizing the English and the German business cycle (the two countries are more strongly correlated with each other than they are with France in both sub-periods) reveal country-specific idiosyncracies for the 1920s compared to the global impact of the Great Depression for the 1930s (figure 4). Troughs (1932) and peaks (1929 and 1937) are identical after 1929 but different before. Some well-studied shocks of the 1920s can easily be identified, for instance the pronounced trough for Germany in 1923 (Ruhr occupation, general strike and hyperinflation) and the 1925/26 UK recession after returning to gold at an overvalued parity (Ritschl 2002, Morys forthcoming 2014).

[Insert Figure 4 about here]

Core-periphery correlations Differentiating between the 1920s and the 1930s is also justified when analysing synchronisation levels between core and periphery; the increase from 0.11 to 0.63 is similar to the intra-core findings and also statistically significant. In the latter period, the synchronisation levels with Germany in particular are very high, reaching values of 0.89 and 0.93 for Yugoslavia and Romania, respectively.

Intra-SEE correlations The only exception to the concept of two distinct periods before and after 1929 is the SEE countries themselves: they initially exhibit an average correlation of 0.57 which declines marginally after the onset of the Great Depression to 0.53, with the difference not being statistically significant. Closer inspection reveals a particularly strong integration between Austria, Bulgaria, Romania and Yugoslavia with an average correlation of 0.72 (again with little difference before and after 1929). The correlations between the four countries are all statistically significant at the 1%-level and they remain so (mostly at the same level of statistical significance) even if reduced to the two shorter periods before and after 1929.

[Insert Figure 5 about here] - 27 -

[Insert Figure 6 about here]

These very high numbers – both compared to our findings for the 1875–1913 period and the other bilateral relationships in the interwar period – suggest that the SEE countries formed a regional business cycle: a group of contiguous countries enjoying synchronization levels substantially elevated to what they exhibit vis-à-vis countries outside of this group. Greece was part of this regional business cycle but remained less closely integrated. The only other country participating in this cycle – though in the 1930s only – was Germany which an average bilateral correlation of 0.73.

Explaining business cycle synchronisation Sectoral indicators In explaining business cycle dynamics in the interwar period, it is again warranted to compare the business cycle and the 20 underlying time series. The association of agriculture increases slightly but remains low, in line with our results for the earlier period. By contrast, the average correlation of each of the six sectoral indicators driving the business cycle before World War I increases further in the interwar period (with the exception of communication). Transportation remains the sectoral indicator most closely correlated with the business cycle in the interwar period (average value of 0.65), followed by construction, fixed investment and industry. Industry now tracks closely the business cycle in the cases of Austria (+0.89), Romania (+0.78) and Greece (+0.71), less so but still positive for Yugoslavia (+0.37) and Bulgaria (+0.13). Transport and fixed investment, two of the closely related indicators already before World War I, increase their correlations on average by a further 10.4% and 19.0%, respectively. Mining and construction also became more relevant to the cycle, with only communication exhibiting a slightly smaller correlation on average compared to pre-World War I (-0.07).14 In sum, the increase in correlation between the seven sectoral series and the business cycle suggests that the interwar business cycles in all SEE-5 countries became more firmly entrenched and bore more relevance for the different sectors and different types of economic activity than had been the case

14 The reduced importance of communication incidentally reinforces the interpretation advanced above, according to which the correlation of communication – which also captures a great deal of non- business activity – is reduced as soon as data more specifically connected to commercial activity become available. - 28 - before World War I. This is likely to reflect the successful industrialisation efforts of the interwar period which are well documented for Bulgaria (Lampe 1986: 68-74) and Greece (Mazower 1991, Kostelenos 1995, 2007, Christodoulaki 2001) but also took place in Romania and Yugoslavia, albeit on a smaller scale (Lampe&Jackson 1982).

[Insert Table 11 about here]

Fiscal, monetary and trade indicators Expenditure and revenue are now strongly positively correlated to the business cycle, most likely reflecting the bigger share of government in aggregate economic activity compared to pre-World War I. With an average value of 0.69, government revenue constitutes the most closely associated indicator of all 20 time series. Narrow and broad money, strongly correlated already before World War I, strengthen their indicative power further. The CPI, as for the earlier period, is positively correlated to the cycle either on a contemporaneous basis or as a lag-factor. The short term interest rates delivers again mixed results (mildly anti-cyclical with an average value of - 0.12), presumably reflecting the econometric problems associated with infrequent discount rate changes (cf. above). The positive correlation for exports strengthens further in the interwar period (from 0.33 to 0.46), while the correlation for imports hardly changes (from 0.44 to 0.42). The terms of trade also remain positively correlated but they exhibit more strongly than before World War I the characteristics of a lead-indicator.

What drove synchronisation? Regional trade and the rise of Germany in SEE Comparing the indicator for the short-term interest rate with the trade indicators is suggestive of cross-country synchronisation through trade and not by means of interest rate linkage (as was the case before World War I). To support this view and in line with the approach pursued for the earlier period, we establish the correlation between the short-term interest rates of England, France and Germany with those of the SEE-5, resulting in an average value of 0.0 for the full period, with little difference for the period before or after 1929 or if differentiating between England, France and Germany (not reported). By contrast, the average correlation of the SEE export cycles with the core countries’ import cycle stands at 0.35, with Austria and Yugoslavia on the Western Balkans exhibiting above-average values (0.64 and 0.52, - 29 - respectively), Greece – which we showed to be less synchronised in our main results (tables 8, 9) – showing a value close to zero (0.07) and Bulgaria and Romania in between (0.30 and 0.21, respectively). While trade was an important factor for business cycle synchronisation, a comprehensive explanation for the emergence of a regional SEE business cycle needs to take wartime developments and monetary events – the other two parameters discussed by Ritschl&Straumann (2010) – into account as well. Table 12, in providing a summary of peaks and troughs for the SEE-5, shows that turning points of Austria, Bulgaria, Romania and Yugoslavia were never further apart than one year for any country-pair. Greece, by contrast, exhibits a first peak two or three later than the other four countries (in 1923), followed by a period of synchronicity for the 1st trough and the 2nd peak (in 1926 and 1929, respectively), but then shows dis-synchronisation again with a Great Depression trough as early as 1931 and a peak in 1934 (in both cases two or three years earlier than its SEE counterparts).

[Insert Table 12 about here]

An answer to all three questions suggests increased synchronization for the SEE economies in the interwar period but can also explain why Greece remained less closely integrated. To begin chronologically, the SEE countries seem to have followed the standard (continental) European pattern of a war time recession; while our calculations only begin in 1919, some of the 20 time series were available for 1914- 1918 as well, documenting a steep decline in economic activity for all countries, some of which were drawn into the war immediately (Austria-Hungary and Serbia) and some later (Bulgaria and Greece in 1915 and Romania in 1916). War time recessions were – as is well documented for the German case, for instance (Ritschl&Straumann 2010) – usually followed by a peak some two years later; which is exactly what we find for Austria, Bulgaria and Romania (Yugoslavia peaks one year later in 1921). Greece’s first peak in 1923 does not violate the two-year-rule by much either, taking into account that Greece, due to the Greco-Turkish war of 1919-1922, was longer tied up in military conflict than the other Balkan countries. Troughs and peaks were subsequently well synchronized in the second half of the 1920s. A first peace-time trough occurred in 1926 in the Eastern Balkans (Bulgaria, Greece, Romania) and the year thereafter in Austria and Yugoslavia; this - 30 - was followed by a peak in 1929 for all countries except for Austria which peaked in the following year. The business cycle upswing was used by Bulgaria, Greece, Romania and Yugoslavia to join the interwar gold standard, a process which happened almost simultaneously in the region, with Greece, Bulgaria and Romania re- establishing convertibility in May 1928, December 1928 and February 1929, respectively, and Yugoslavia stabilising its exchange-rate vis-à-vis gold standard countries in 1928 (Lazaretou 2006, Nenovksy&Dimitrova 2007, Stoenescu et al. 2008).15 That 1920s synchronisation was underpinned by trade not only emerges from the econometric results presented above but also from developments in trade openness and trade patterns at the time (tables 13, 14). Comparing 1925 and 1929 with 1913, the average openness ratio increased by approximately 10% (1913: 27.7%; 1925: 38.5%; 1929: 36.3%). Trade policies in the Balkan countries became more protectionist in the 1920s (Liepmann 1938, Berov 1989), but their effect was more than compensated by the vast improvements in transport infrastructure at the time (Stanev 2011). The increase in shared borders as a result of the Balkan Wars also helped the 1920s become the heyday of intra-Balkan trade. Greece, which then for the first time had common borders with Bulgaria and Yugoslavia, tripled its trade with the two neighbouring countries and Romania compared to pre-World War I. Bulgaria, Romania and Yugoslavia all approximately doubled their intra-Balkan trade share, and Austria traded with the 4 Balkan countries on a year-by-year basis more than (much larger) Austria-Hungary had ever done before World War I.

[Insert Table 13 about here]

[Insert Table 14 about here]

At the same time, most of the trade continued to involve the Western European core economies. Austria, Bulgaria, Greece and Romania conducted throughout the interwar period more trade with England, France and Germany than with the SEE-5 (this is true even when including trade with Hungary and Czechoslovakia in order to allow comparison with our results for pre-WW I trade patterns). Only Yugoslavia

15 Only the Austrian case is different, where the hyperinflation of 1924 forced an earlier stabilisation, somewhat similar to events one year earlier in Germany. - 31 - consistently traded more with its SEE neighbours than with England, France and Germany. Beneath the aggregate numbers for the European core countries, the importance of Germany as the main trading partner for the Balkan countries becomes clearer. While magnitude and reorientation of German trade with SEE in the 1930s is well-documented (Basch 1943, Momtchiloff 1944, Kaiser 1981, Feinstein et al. 2008, pp. 152-154) and has given rise to a controversial discussion on the exact nature of the trading relationship (Neal 1979, Ritschl 2001, Ivanov&Tooze 2011), the trade data provided in tables 7 and 13 from the Congress of Berlin (1878) to the onset of World War II show the growing dominance of Germany in SEE to be a longer drawn out process (cf. Ritschl 2001, pp. 338-339 for a similar observation). In some cases, this goes back to the early 1900s and the process gathered considerable pace in all SEE countries in the 1920. As early as 1910, Germany captured a larger trade volume of Bulgaria, Serbia and Romania than the UK did. In the 1920s, Germany traded three times as much with Bulgaria as the UK does, twice as much in the Romanian case and still 50% more in the Yugoslav case. Even in Greece, traditionally a stronghold of UK trade, Germany was catching up quickly, roughly playing on par with the UK in the 1920s. The even more dominant trading position of Germany in the 1930s then appears as an extension of a pre-existing trend. The bilateral clearing agreements and the same choice of currency regime after 1931 (imposition of capital controls but no formal abandonment of the gold standard by all SEE countries with the notable exception of Greece, cf. below) certainly help explain the extraordinary levels at which German trade shares stood at the outbreak of World War II, reaching 66.7% in the case of Bulgaria. But clearing agreements and common currency regime were reinforcing factors rather than causal, as entering into the trade agreements with Germany and the choice of currency regime themselves had been motivated by the economic position Germany had acquired in SEE (Ellis 1941, Momtchiloff 1944, Wolf&Ritschl 2011). German trade dominance as early as the 1920s in all of SEE except for Greece is important for another reason: the choice of the exchange rate regime after the breakdown of the interwar gold standard and the implications of this choice for the business cycle. All 5 SEE countries became embroiled in the European financial crisis of 1931 but they took different decisions which were, at least partly, motivated by the - 32 - choice of their main trading partner: Austria, Bulgaria, Romania and Yugoslavia chose, in line with Germany’s approach of July 1931, deflationary policies cum exchange controls without formally abandoning the gold standard (Tooze&Ivanov 2011, Nenovsky&Dimitrova 2006). Greece, by contrast, was the only SEE country to follow the UK in leaving the gold standard, allowing the drachma to float in April 1932. While this decision was motivated by a number of economic factors and political considerations, the long-standing trade relationship with the UK was clearly one of them (Mazower 1991, Ivanov&Tooze 2011, Bank of Greece 2011, Christodoulakis forthcoming 2013). Greece’s decision to devalue in April 1932 implied business business cycle dynamics different from the other four SEE countries. In line with Eichengreen&Sachs (1985)’s finding that an early exit from the gold standard led ceteris paribus to a quicker recovery, the Greek economy began to recover in the same year (trough in 193116); by contrast, the exchange-control countries needed two (Austria, Romania, Yugoslavia) and three (Bulgaria) years more to reach their respective trough. And yet, while this different policy choice explains why Greece remained less well integrated into the SEE business cycle in the 1930s, the country could not escape the dominance of the German economy in the Balkans either. Germany became Greece’s most important trading partner in the 1930s and the cyclical correlation between the two countries increased to 0.54 for the period after 1929, a value lower than the average value of 0.78 for the other four SEE economies but clear evidence of synchronisation.

5. Conclusion This paper constructs dynamic factor business cycle indices for five SEE countries from the 1870s to World War II and analyses patterns of synchronisation across countries and over time. While all emanating from the same political and economic unit – the European provinces of the late Ottoman Empire - , synchronisation levels on the Balkan peninsula were initially low but steadily increased towards World War

16 This finding of a quick recovery after leaving the gold standard appears robust to the choice of methodology. Relying on more conventional techniques, the industrial production indices of the Supreme Economic Council (a contemporary estimate), Mazower (1991) and Christodoulaki (2001) all detect 1932 as trough. For an overview cf. Christodoulaki (2001). - 33 -

I, reaching levels not yet documented for any peripheral region of Europe for this period. Analysing the driving forces of national cycles as well as of cross-country synchronisation patterns revealed that the SEE economies increasingly participated in a pan-European business cycle radiating from England, France and Germany. The formation of a genuine SEE regional business cycle only happened in the interwar period. Driven by growing intra-regional trade, the 1920s were the first period in Balkan history where countries were more strongly synchronised among each other than they were with any of the European core countries. This finding not only contradicts the conventional wisdom according to which such regional business cycles begin to appear only after World War II; crucially, it is testimony to the economic integration which the Balkan countries had achieved since political independence half a century earlier. Improved infrastructure, not overly protectionist policies and a substantial increase in common borders since the Balkan Wars all played a role in this success story, but this was a process of regional integration through economics not politics. While the political attempts at regional integration all failed (or, alternatively, deliberately excluded specific countries from the onset) (Kaiser 1981), economic integration progressed slowly but surely.17 Yet there was another gradual force at work: the increasing economic penetration of SEE by Germany. While the trade dominance of Nazi Germany in the Balkans has generated considerable scholarly interest, our findings show this to be a long drawn-out process in which Germany, starting at the turn of the century, became step-by-step the largest trading partner for each and every Balkan country; which, in turn, explains the extraordinarily high levels of business cycle synchronisation between Germany and the Balkan countries. With the notable exception of the 1920s where the UK’s initially strong economic position in the Balkans had already faded and Germany was temporarily weak following World War I, for most of the time, then, the Balkan countries had economic ties to countries outside of the region stronger than to each other. This is not without irony: 19th century Balkan nationalism was partly driven by the desire to shackle off “outside forces” which, at the time, presented themselves in the form of the Ottoman and the Habsburg Empires. While Austria and Turkey played a marginal - 34 - role in the Balkans already in the interwar period, their once dominant role was taken up by the European core economies and, increasingly, by Germany in particular.

Data Appendix Austria(-Hungary), Bulgaria, Greece, Romania, Serbia/Yugoslavia The main source for all five SEE countries is the respective official Statistical Yearbook. In the following, we provide information for those time series where other sources were used. Numbers for individual time series follow the numbers as given in table 2.

Austria-Hungary (1875-1913) 1, 3, 4, 5 kindly communicated by Schulze 7 Schulze (2008) 10 kindly communicated by Austrian National Bank 11 Komlos (1987) 12 Muehlpeck et al. (1979) 14, 15 own calculations based on variety of sources (detailed description upon request) 18 own calculations based on Bunzl (1914) 20 kindly communicated by Schulze

Austria (1919-1937) 1, 3, 5, 6 Kausel et al. (1965) 8 – 13 Butschek 1999 14 own calculations based on variety of sources (detailed description upon request) 16, 17, 20 Butschek 1999

Bulgaria: 1887-1913, 1918-1940 1, 3, 5 GDP calculations by Ivanov 10, 11 kindly communicated by Bulgarian National Bank 12 own calculations based on variety of sources (detailed description upon request) 13 kindly communicated by Bulgarian National Bank 14, 15 own calculations based on variety of sources (detailed description upon request) - 35 -

Greece: 1875-1913, 1918-1941 3 – 6 Kostelenos et al. (2007) 8 kindly communicated by Kostelenos 9 kindly communicated by Bank of Greece 10, 11 Kostelenos et al. (2007) 12, 13 kindly communicated by Bank of Greece 14, 15 own calculations based on variety of sources (detailed description upon request) 16, 17 kindly communicated by Kostelenos 18 kindly communicated by Tuncer 20 Kostelenos et al. (2007)

Romania: 1881-1913, 1918-1941 12, 14, 15 own calculations based on variety of sources (detailed description upon request)

Serbia/Yugoslavia: 1886-1912, 1919-1940 8, 9 Sundhausen (1989) 10, 11 kindly communicated by National Bank of Serbia 12 own calculations based on variety of sources (detailed description upon request) 13 kindly communicated by National Bank of Serbia 14, 15 own calculations based on variety of sources (detailed description upon request) 16, 17 Sundhausen (1989)

England, France, Germany

The main reference for England, France and Germany is Mitchell (2007). In the following, we provide information for those time series where other sources were used.

Numbers for individual time series follow the numbers as given in table 2.

England 7 Feinstein (1972) 11 Capie&Webber (1985) 13 Hawtrey (1962) and League of Nations Statistical Yearbook 15 Solomou&Catao (2000)

France 13 Hawtrey (1962) and League of Nations Statistical Yearbook 15 Solomou&Catao (2000)

Germany 5, 7 Hoffmann (1965) 13 Reichsbank Annual Reports and League of Nations Statistical Yearbook 15 Solomou&Catao (2000) - 36 -

Bibliography

A'Hearn, B., and U. Woitek. "More international evidence on the historical properties of business cycles." Journal of Monetary Economics 47 (2001): 321-46. Aiolfi, M., L. Catao, and A. Timmermann. Common Factors in Latin America's Business Cycles, IMF Working Paper 06/49 2006. ———. "Common factors in Latin America's business cycles." Journal of Development Economics 95 (2011): 212-28. Artis, M., G. Chouliarakis, and P.K.G. Harischandra. "Business cycle synchronization since 1880." The Manchester School 79 (2011): 173-207. Axenciuc, V. Produsul Itern Brut a Romaniei 1862-2000: Serii Statistice Seculare si Argumente Metodologice. Bucharest: Adademia Romana, 2012. Backus, D.K., and P.J. Kehoe. "International evidence on the historical properties of business cycles." American Economic Review 82 (1992): 864-88. Bai, J., and S. Ng. "Determining the number of factors in approximate factor models." Econometrica 70 (2002): 191-221. Basch, A. The Danube Basin and the German Economic Sphere. New York: Columbia University Press, 1943. Basu, S., and A.M. Taylor. "Business Cycles in International Historical Perspective." Journal of Economic Perspectives 13 (1999): 45-68. Bergman, U., M.D. Bordo, and L. Jonung. "Historical evidence on business cycles: The international experience." In Beyond shocks: What causes business cycles?, edited by J.C. Fuhrer and Schuh, S. Boston/MA: Federal Reserve Bank of Boston, 1998. Bernanke, B., and H. James. "The gold standard, deflation, and financial crises in the Great Depression: An international comparison." In Financial markets and financial crises, 33-68. Chicago: Chicago University Press, 1991. Berov, L. Protectionism and Competition in the Balkans During the 20th Century. Sofia 1989. Bonelli, F. La crisis del 1907: Una tappa dello sviluppo industriale in Italia. Turin: L. Einaudi Foundation, 1971. Bordo, M.D., and T. Helbling. Have National Business Cycles Become More Synchronized?, NBER Working Paper Nr. 10130 2003. Bordo, M.D., and T.F. Helbling. "International business cycle synchronization in historial perspective." The Manchester School 79 (2011): 208-38. Broadberry, S., and A. Klein. "When and Why did Eastern European Economies begin to Fail?: lessons from a Czechoslovak/UK Productivity Comparison, 1921-1991." Explorations in Economic History 48 (2011): 37-52. Bunzl, O. Der Wiener Rentenmarkt. 1884 bis 1914. Vienna: Manzsche k.u.k. Hof- Verlags- und Universitätsbuchhandlung, 1914. Burns, A.F., and W.C. Mitchell. Measuring business cycles. New York: NBER, 1946. Butschek, F. Statistische Reihen zur österreichischen Wirtschaftsgeschichte. Die österreichische Wirtschaft seit der Industriellen Revolution. Vienna: Österreichisches Institut für Wirtschaftsforschung, 1999. Capie, F.H., and A. Webber. A Monetary History of the United Kingdom. 1870-1982. Volume 1. Data, sources, and methods. London, New York: Routledge, 1995. Chirot, D. Social change in a peripheral society: the creation of a Balkan colony. New York: Academic Press, 1976. - 37 -

Christodoulaki, O. "Industrial Growth in Greece between the Wars: A New Perspective." European Review of Economic History 5 (2001): 61-90. Craig, L., and D. Fisher. "Integration of the European business cycle: 1871-1910." Explorations in Economic History 29 (1992): 144-68. Daudin, G., M. Morys, and K. O'Rourke. "Globalization, 1870-1914." In The Cambridge Economic History of Modern Europe. Volume 2: 1870 to the present, edited by S. Broadberry and O'Rourke, K., 5-29, 2010. Deyanov, D. Bulgarian Railway Network 1866-1975 [in Bulgarian]. Sofia 1995. Eichengreen, B., and J. Sachs. "Exchange rates and economic recovery in the 1930s." Journal of Economic History 45 (1985): 925-46. Ellis, E. Exchange Control in Central Europe. Cambridge/MA: Harvard University Press, 1941. Feinstein, C.H., P. Temin, and G. Toniolo. The European economy between the wars. Oxford: Oxford University Press, 1997. ———. The World Economy between the World Wars. Oxford: Oxford University Press, 2008. Feis, H. Europe, the World's Banker, 1870-1914. An account of European foreign investment and the connection of world finance with diplomacy before the War. New Haven, CT: Council on Foreign Relations, 1930. Geweke, J. "The dynamic factor analysis of economic time series." In Latent variables in socio-economic models, edited by D. Aigner and Goldberger, A. Amsterdam: North Holland, 1977. Good, D., and T. Ma. "The Economic Growth of Central and Eastern Europe, 1870- 1989." European Review of Economic History 3 (1999): 103-38. Harding, D., and A. Pagan. "Dissecting the cycle: A methodological investigation." Journal of Monetary Economics 49 (2002): 365-81. Harvey, A., and A. Jaeger. "Detrending, stylized facts and the business cycle." Journal of Applied Econometrics 8 (1993): 231-47. Hawtrey, R.G. A Century of Bank Rate. 2nd ed. London: Longmans Green & Co., 1962. Hitchins, K. Rumania 1866-1947. Oxford: Oxford University Press, 1994. Ivanov, M. Bulgarian GDP in Historical Perspective, 1870-2000. Sofia 2009 (in Bulgarian). ———. Bulgarian National Income between 1892 and 1924, Bulgarian National Bank Discussion Paper Nr. 54 2006. ———. The Gross Domestic Product of Bulgaria, 1870-1945. Sofia: Ciela, 2012. Ivanov, M., and A. Tooze. "Convergence or Decline on Europe’s Southeastern Periphery? Agriculture, Population and GNP in Bulgaria 1892–1945." Journal of Economic History 67 (2007): 672–704. Jordan, C.N. Romanian oil industry 1955. Juglar, C. Des crises commerciales et de leur retour périodique en France, en Angleterre et aux Etats-Unis. 2nd ed. Paris: Guillaumin, 1889. Kaiser, D.E. Economic diplomacy and the origins of the Second World War: Germany, Britain, France, and Eastern Europe, 1930–1939. Princeton/NJ 1981. Kausel, A., N. Nemeth, and H. Seidel. Österreichs Volkseinkommen 1913 bis 1963 (Monatsberichte des Österreichischen Institutes für Wirtschaftsforschung. 14. Sonderheft). Vienna: Österreichisches Institut für Wirtschaftsforschung, 1965. Kindleberger, C.P. Manias, Panics and Crashes. A History of Financial Crises. 5th ed. New York: Palgrave Macmillan, 2005. - 38 -

Komlos, J. "Financial Innovation and the Demand for Money in Austria-Hungary, 1867-1913." Journal of European Economic History 16 (1987): 587-606. Koop, G., and D. Korobilis. Bayesian multivariate time series methods for empirical macroeconomics. Glasgow 2010 (manuscript). Kose, A., C. Otrok, and C. Whiteman. "Understanding the evolution of world business cycles." Journal of International Economics 75 (2008): 110-30. Kostelenos, G. Money and Output in Modern Greece: 1858-1938. Athens: KEPE, 1995. Kostelenos, G., D. Vasileiou, S. Petmezas, and M. Sfakianakis. Gross Domestic Product. Athens: KEPE, 2007 (in Greek). Krugman, P. "Lessons of Massachusetts for EMU." In Adjustment and growth in the European Monetary Union, edited by F. Torres and Giavazzi, F. Cambridge: Cambridge University Press, 1993. Kuznets, S. "Long swings in the growth of population and in related economic variables." Proceedings of the American Philosophical Society 102 (1958): 25- 57. Lampe, J.R. The Bulgarian Economy in the Twentieth Century. London: Croom Helm, 1986. ———. Yugoslavia as a history. Twice there was a country. Cambridge: Cambridge University Press, 2000. Lampe, J.R., and M.R. Jackson. Balkan economic history, 1550-1950. From imperial border lands to developing nations. Bloomington: Indiana University Press, 1982. Lazaretou, S. "Greek monetary economics in retrospect: History and data." In Monetary and Fiscal Policies in South-Eastern Europe: Historical and Comparative Perspectives (Conference proceedings of the 1st meeting of the South-Eastern European Monetary History Network), edited by R. Avramov and Pamuk, S., 77-97. Sofia: Bulgarian National Bank, 2006. Lupu, M.A. The Economy of Romania. Bucharest: Meridiane Publishing House, 1968. Maddison, A. The World Economy. Historical Statistics. Paris: OECD Development Centre Studies, 2003. Mazower, M. The Balkans. New York: Phoenix Press, 2001. ———. Greece and the inter-war economic crisis. Oxford: Oxford University Press, 1991. Metz, R. "Comment on "Stock markets and business cycle comovement in Germany before World War I: Evidence from spectral analysis"." Journal of Macroeconomics 31 (2009): 58-67. Mitchell, B.R. International Historical Statistics. 6th ed 2007. Mitchell, W.C. Business cycles: The problem and its setting. New York: NBER, 1927. Mitchin, J. "Cycles and trends in economic factors." Review of Economics and Statistics 5 (1923): 10-17. Momtchiloff, N. Ten years of controlled trade in south-eastern Europe. Cambridge 1944. Morgenstern, O. International financial transactions and business cycles. Princeton: Princeton University Press, 1959. Morys, M. "The Classical Gold Standard in the European Periphery: A Case Study of Austria-Hungary and Italy, 1870-1913." PhD thesis London School of Economics and Political Science, 2006. ———. "Discount rate policy under the Classical Gold Standard: Core versus periphery (1870-1914)." Explorations in Economic History 50 (2013): 205-26. - 39 -

———. "South-Eastern European Growth Experience in European Perspective, 19th and 20th Centuries." In Monetary and Fiscal Policies in South-Eastern Europe: Historical and Comparative Perspectives (Conference Proceedings of the 1st Meeting of the South-East European Monetary History Network), edited by R. Avramov and Pamuk, S., 23-44. Sofia: Bulgarian National Bank, 2006. ———. "South-Eastern European Monetary History in a Comparative Perspective., 1870s to 1914." Bank of Greece Working Paper 94/2009 (2009). Mühlpeck, V., R. Sandgruber, and H. Woitek. "Index der Verbraucherpreise 1800- 1914." In Geschichte und Ergebnisse der zentralen amtlichen Statistik in Österreich. 1829-1979, 649-88. Vienna: Österreichisches Statistisches Zentralamt, 1979. Neal, L.D. "The Economics and Finance of Bilateral Clearing Agreements: Germany, 1934-1938." Economic History Review 32 (1979): 391-404. Nenovsky, N., and K. Dimitrova. "Exchange rate and inflation: France and Bulgaria in the interwar period." In Monetary and Fiscal Policies in South-Eastern Europe: Historical and Comparative Perspectives (Conference Proceedings of the 1st Meeting of the South-East European Monetary History Network), edited by R. Avramov and Pamuk, S., 61-98. Sofia: Bulgarian National Bank, 2006. Nikova, E. "Roads Connecting, Roads: Infrastructure in South East Europe." Etudes Balkanique (2007): 3-16. O'Rourke, K.H., and J.G. Williamson. Globalization and History: The Evolution of a Nineteenth-Century Atlantic Economy. Cambridge: MIT Press, 1999. Obstfeld, M., J.C. Shambaugh, and A.M. Taylor. "The Trilemma in History: Tradeoffs among Exchange Rates, Monetary Policies, and Capital Mobility." Review of Economics and Statistics 87 (2005): 423-38. Palairet, M. The Balkan Economies ca. 1800-1914: Evolution without Development. Cambridge: Cambridge University Press, 1997. Pearton, M. Oil and the Romanian state. Oxford: Clarendon Press, 1971. Ritschl, A. Deutschlands Krise und Konjunktur: Binnenkonjunktur, Auslandsverschuldung und Reparationsproblem zwischen Dawes-Plan und Transfersperre 1924-1934. Berlin: Akademie-Verlag, 2002. ———. "Nazi economic imperialism and the exploitation of the small: evidence from Germany’s secret foreign exchange balances, 1938–1940." Economic History Review 54 (2001): 324-45. Ritschl, A., S. Sarferaz, and M. Uebele. The U.S. Business Cycle, 1867-1995: Dynamic Factor Analysis vs. Reconstructed National Accounts, DFG SFB 649 Discussion Paper Nr. 66 2008. Ritschl, A., and T. Straumann. "Business cycles and economic policy, 1914-1945." In The Cambridge Economic History of Modern Europe. Volume 2: 1870 to the present, edited by S. Broadberry and O'Rourke, K., 156-80, 2010. Romer, C. "Is the stabilization of the postwar economy a figment of the data?" American Economic Review 76 (1986): 314-34. ———. "New estimates of prewar gross national product and unemployment." Journal of Economic History 46 (1986): 341-52. ———. "The prewar business cycle reconsidered: New estimates of gross national product, 1869-1908." Journal of Political Economy 97 (1989). - 40 -

Sarferaz, S., and M. Uebele. "Tracking Down the Business Cycle: A Dynamic Factor Model for Germany, 1820-1913." Explorations in Economic History 46 (2009): 368-87. Schulze, M.-S. "The Growth and Composition of Imperial Austria's Fixed Capital Stock: New Annual Time Series for 1870-1913." Economies et Sociétés (Série "Histoire économique quantitative") 9 (2008): 1597-614. ———. "Patterns of Growth and Stagnation in the Late 19th Century Habsburg Economy." European Review of Economic History 4 (2000): 311-40. Singleton, F.B. Twentieth Century Yugoslavia. London: Macmillan, 1976. Singleton, F.B., and B. Carter. The Economy of Yugoslavia. London: Croom Helm, 1982. Solomou, S., and L. Catao. "Effective exchange rates 1879-1913." European Review of Economic History 4 (2000): 361-82. Sundhausen, H. Historische Statistik Serbiens, 1834-1914, mit europaeischen Vergleichsdaten. Munich: Oldenbourg, 1989. Todorova, M. Imagining the Balkans. Oxford: Oxford University Press, 2009. Tooze, A., and M. Ivanov. "Disciplining the 'black sheep of the Balkans': financial supervision and sovereignty in Bulgaria, 1902-1938." Economic History Review 64 (2011): 30-51. ———. "Tightening the Golden Fetters: Financial Supervision and Sovereignty in Bulgaria, 1902-1935." Economic History Review (2009 forthcoming). Turnock, R. The Romanian Economy in the Twentieth Century. London: Croom Helm, 1986. Uebele, M. "Die Identifikation internationaler Konjunkturzyklen in disaggregierten Daten: Deutschland, Frankreich und Grossbritannien, 1862-1913." Jahrbuch fuer Wirtschaftsgeschichte (2011). Uebele, M., and A. Ritschl. "Stock markets and business cycle comovement in Germany before World War I: Evidence from spectral analysis." Journal of Macroeconomics 31 (2009): 35-57. Wolf, N., and A. Ritschl. "Endogeneity of Currency Areas and Trade Blocs: Evidence from a Natural Experiment." Kyklos 64 (2011): 291-312. 5

4

3

2

1

0

-1

-2

-3

-4 1880 1885 1890 1895 1900 1905 1910

ENGLAND FRANCE GERMANY

Figure 1: Business cycles of England, France and Germany, 1879 - 1913.

Source: Own calculations based on data as described in the main text.

6

4

2

0

-2

-4

-6 1875 1880 1885 1890 1895 1900 1905 1910

AUSTRIA ROMANIA GREECE

Figure 2: Business cycles of Austria, Greece and Romania, 1881 - 1913.

Source: Own calculations based on data as described in the main text. 4

3

2

1

0

-1

-2

-3 03 04 05 06 07 08 09 10 11 12 13

AUSTRIA ROMANIA GREECE

Figure 3: Business cycles of Austria, Greece and Romania, 1903 - 1913.

Source: Own calculations based on data as described in the main text.

6

4

2

0

-2

-4

-6

-8 1922 1924 1926 1928 1930 1932 1934 1936 1938

ENGLAND GERMANY

Figure 4: Business cycles of England and Germany, 1921 - 1939.

Source: Own calculations based on data as described in the main text. 6

4

2

0

-2

-4

-6 20 22 24 26 28 30 32 34 36 38

BULGARIA ROMANIA YUGOSLAVIA

Figure 5: Business cycles of Bulgaria, Romania and Yugoslavia, 1919 - 1939.

Source: Own calculations based on data as described in the main text.

6

4

2

0

-2

-4

-6 20 22 24 26 28 30 32 34 36 38

BULGARIA ROMANIA YUGOSLAVIA GREECE

Figure 6: Business cycles of Bulgaria, Greece, Romania and Yugoslavia, 1919 - 1939.

Source: Own calculations based on data as described in the main text. Table 1 Business cycle synchronization during the First Age of Globalization (ca. 1870-1913) according to previous research

Correlation Source Average Countries1 Time Statistical among / correlation frame method: between (# of bilateral correlation of correlations) Core Morgenstern 0.83 (3) E, F, G 1870-1914 concordance countries (1959) index2 Backus&Kehoe 0.03 (1) E, G 1870-1913 de-trended GDP 1992 (table 4) Artis et al. 2011 0.09 (3) E, F, G 1880-1913 de-trended GDP (table 2) Bordo&Helbling 0.04 (15) E, F, G, 1880-1913 GDP growth 2011 (table 1) Netherlands, rates Switzerland, US Bordo&Helbling 0.09 (6) F, G, 1880-1913 GDP growth 2011 (table 1) Netherlands, CH rates Übele 2011 0.61 (3) E, F, G 1862-1913 CDFA business (table 2) cycle indices Peripheral Backus&Kehoe 0.29 (3) Denm., Norway, 1865-1914 de-trended GDP countries 1992 (table 4) Sweden 3 Artis et al. 2011 0.11 (6) Denm., Finland, 1880-1913 de-trended GDP (table 2) Norway, Swed. Artis et al. 2011 0.15 (1) Austria-H., 1880-1913 de-trended GDP (table 2) Greece Bordo&Helbling 0.14 (6) Denm., Finland, 1880-1913 GDP growth 2011 (table 1) Norway, Swed. rates Core Backus&Kehoe 0.20 (8) E, G vis-à-vis 1861-1913 de-trended GDP vis-à-vis 1992 (table 4) Denm., Italy, 4 periphery Norway, Swed. Backus&Kehoe 0.29 (6) E, G vis-à-vis 1861-1913 de-trended GDP 1992 (table 4) Denmark, 4 Norway, Swed. Artis et al. 2011 0.04 (12) E, F, G vis-à- 1880-1913 de-trended GDP (table 2) Nordic countries Artis et al. 2011 0.16 (6) E, F, G vis-à-vis 1880-1913 de-trended GDP (table 2) A-H., Greece Bordo&Helbling 0.01 (60) 6 core vis-à-vis 1880-1913 GDP growth 2011 (table 1) 10 peripheral5 rates

Notes: 1 E: England; F: France; G: Germany; CH: Switzerland: A-H: Austria-Hungary. 2 As explained in the main text (chapter 4), the concordance index cannot be directly compared to the correlation coefficient. 3 Bilateral correlations vis-à-vis Denmark only starting in 1870. 4 Bilateral correlations vis-à-vis Norway only starting in 1865, vis-à-vis Denmark and England in 1870. 5 Core countries: England, France, Germany, Netherlands, Switzerland, US; peripheral countries: Australia, Canada, Denmark, Finland, Italy, Japan, Norway, Portugal, Spain, Sweden.

Sources: Provided in column 2. Table 2 Annual data series for common dynamic factor analysis

Sectoral output indicators #1 agricultural production #2 communication #3 industrial output #4 mining #5 construction #6 transportation #7 fixed investment

Fiscal indicators #8 government expenditure #9 government revenue

Financial indicators #10 narrow money #11 broad money #12 consumer price index #13 short term interest rate

Trade indicators #14 terms of trade #15 real effective exchange rate #16 exports #17 imports

Other indicators #18 external spread #19 real wage #20 population Table 3 Business cycle correlations 1879 – 1913 by country pairs: Full period1, 1893-1913, 1903-1913

AH Gr Ro Se Bu England France Germany Greece 0.06 0.23 0.60 * Romania 0.30 * 0.26 0.31 0.33 0.51 0.45 Serbia 0.34 * 0.27 0.19 0.55 ** 0.22 0.32 0.87 *** 0.76 ** 0.64 ** Bulgaria 0.34 * -0.07 0.27 0.42 ** 0.28 -0.25 0.37 * 0.25 0.03 -0.45 0.22 0.15 England 0.67 *** 0.25 0.13 0.24 -0.07 0.49 ** 0.37 0.07 0.09 -0.29 0.45 0.49 0.29 0.36 -0.55 * France 0.78 *** 0.23 0.17 0.48 ** 0.25 0.78 *** 0.85 *** 0.12 0.19 0.46 ** 0.08 0.66 *** 0.88 *** 0.59 * 0.56 * 0.70 ** -0.19 0.73 ** Germany 0.84 *** 0.08 0.19 0.34 * 0.18 0.81 *** 0.81 *** 0.83 *** 0.25 0.23 0.42 ** 0.07 0.77 *** 0.86 *** 0.78 *** 0.57 * 0.59 * 0.73 ** -0.18 0.84 *** 0.87 ***

Notes: Entries are bilateral correlations of the cyclical component (as computed according to the description in the main text). The first entry in each cell refers to the full period and the second and the third entries refer to 1893-1913 and 1903-1913, respectively. *, ** and *** denote statistical significance levels of 10%, 5% and 1%, respectively. 1 The full period is given by pair-wise intersection of the estimation period, i.e., it might differ between country pairs (for details see data appendix).

Sources: Cf. main text. Table 4 Business cycle correlations 1879 – 1913 (summary statistics) Full period, 1893-1913, 1903-1913

AH Gr Ro Se Bu England France Germany Summary statistics for SEE-5 Average r 0.26 0.13 0.26 0.30 0.24 0.25 0.38 0.32 vis-à-vis 0.34 0.13 0.34 0.34 0.17 0.15 0.34 0.36 SEE-5 0.50 0.34 0.45 0.61 -0.01 0.21 0.51 0.50

0.24 0.32 0.26 0.28 0.38 0.40 Average r 0.76 0.19 0.17 0.35 0.12 vis-à-vis 0.72 0.25 0.16 0.33 -0.04 E, F, G 0.70 0.55 0.48 0.60 -0.31

0.32 0.28 0.40 Summary statistics for SEE-5 and England, France, Germany (28 bilateral correlations) Average r 0.34 0.33 0.44

Notes: Entries are bilateral correlations of the cyclical component (as computed according to the description in the main text). The first entry in each cell refers to the full period and the second and the third entries refer to 1893-1913 and 1903-1913, respectively.

Sources: Table 3. Table 5 Correlation of underlying time series with the business cycle 1875-1913 Correlation in percentage at t = 0 (1st entry) and maximum (absolute) correlation at t = -2, -1, 0, + 1 or +2 (where + indicates a lead-factor and – a lag-factor)

Austria-H. Bulgaria Greece Romania Serbia Ø Sectoral indicators agriculture -1 27 -1 20 20 0 -35 -35 0 28 41 1 8 34 2 4 communication -24 -29 2 64 64 0 43 43 0 34 34 0 73 73 0 38 industry 63 63 0 44 44 0 23 23 0 44 55-1 56 56 0 46 mining 53 53 0 13 13 0 14 41-1 14 30-1 42 42 0 27 construction 36 36 0 69 69 0 33 33 0 24 42-1 41 transportation 76 76 0 59 59 0 13 42 1 65 65 0 59 59 0 54 fixedinvestment 56 56 0 41 41 0 12 12 0 55 55 0 41 Fiscal indicators gov.expenditure -13 23 -2 -21 -21 0 -9 37 2 -27 -27 0 57 57 0 -3 gov.revenue 4 56 -2 -29 -29 0 14 12 2 34 34 0 56 56 0 16 Financial indicators narrowmoney(M0) 54 54 0 1 16 1 88 88 0 94 94 0 19 44 1 51 broadmoney(M3) 26 42 1 0 15 2 87 87 0 94 94 0 17 44 -1 45 CPI 67670 -940-1 536-1 2340-122 short term interest rate 55 55 0 0 32 2 14 14 0 -56 -56 0 -15 34 -2 0 Trade indicators termsoftrade 38 39 1 45 45 0 39 39 0 27 27 0 -20 35 -2 26 realeffectivexr -38 -38 0 4 -39 -1 23 23 0 19 19 0 12 12 0 4 exports 55 55 0 36 36 0 29 29 0 32 40 1 12 39-1 33 imports 59 59 0 35 35 0 44 44 0 64 64 0 17 33-1 44 Other indicators External spread 11 34 -1 Real wage 0 19 -2 Population 23 23 0 12 12 0 8 14 -1

Sources: Own calculations as described in the main text. Table 6 Exports plus imports as share of GDP (openness ratio) European periphery, 1870 - 1913

South-East 1870 1880 1890 1900 1913 Avg. Avg. Europe (1880- 1913) Austria-Hungary 29.0% 25.5% 25.2% 26.8% 24.1% 26.1% 25.4% Bulgaria n.a. 16.6% 20.0% 11.7% 16.7% 16.3% 16.3% Greece 45.6% 42.3% 39.4% 42.3% 29.4% 39.8% 38.4% Romania 29.5% 37.5% 23.4% 35.9% 31.6% 31.6% Serbia 40.6% 32.2% 35.1% 32.5%1 35.1% 35.1%

Southern Europe Italy 18.3% 18.3% 15.9% 19.0% 23.9% 19.1% 19.3% Portugal 33.7% 43.8% 45.3% 48.9% 57.4% 45.8% 48.9% Spain 12.1% 14.8% 18.8% 22.6% 22.3% 18.1% 19.6%

Eastern Europe Russia 14.4% 15.0% 11.4% 13.8% 13.7% 13.7%

Nordic countries Denmark 35.7% 45.8% 48.0% 52.8% 61.5% 48.8% 52.0% Finland 31.7% 50.8% 39.3% 47.6% 56.2% 45.1% 48.5% Norway 33.9% 36.1% 43.6% 43.4% 50.9% 41.6% 43.5% Sweden 29.4% 37.3% 44.9% 39.4% 34.7% 37.1% 39.1%

Notes: 1 Data refer to 1910.

Sources: Bulgaria, Romania, Serbia: own calculations based on Ivanov (2012) for Bulgaria, Axenciuc (2012) for Romania and data kindly provided by Michael Palairet for Serbia. All other entries: Daudin&Morys&O’Rourke (2010). Table 7 Trade shares of South-East European countries, 1872 - 1913 Percentage of total trade (exports plus imports) Austria-Hungary Bulgaria 18802 1890 1900 1913 avg.3 1882 1889 1901 1910 avg. South-East Europe Austria-Hungary 21.9 17.0 16.0 18.1 18.3 Bulgaria 0.01 0.4 0.8 0.4 Greece 0.01 0.9 0.9 0.6 0.6 0.9 2.3 2.2 1.5 Romania 7.2 2.5 2.1 3.5 2.7 11.3 2.6 2.0 2.4 4.6 Serbia 1.7 3.4 1.9 1.3 2.2 2.3 0.9 0.9 0.9 1.3 Total 5.9 5.1 6.5 5.8 36.3 21.4 21.2 23.6 25.6 Total excl. Austria-H. 5.9 5.1 6.5 5.8 14.4 4.4 5.2 5.5 7.4 Core countries France 5.2 3.4 3.1 3.9 8.8 14.2 5.7 8.0 9.2 Germany 61.0 56.2 45.0 39.1 46.7 0.9 3.4 12.2 15.8 8.1 United Kingdom 11.2 9.6 7.9 9.6 17.2 22.0 19.5 12.4 17.8 Total 72.6 58.0 50.1 60.2 26.9 39.6 37.4 36.1 35.0 Ottoman Empire 0.0 2.9 3.6 2.2 18.3 26.3 22.6 21.3 22.1 Greece Romania 1872 1887 1897 1912 avg. 1880 1890 1900 1911 avg. South-East Europe Austria-Hungary 13.9 10.1 9.7 14.9 12.2 44.1 9.6 22.9 15.8 23.1 Bulgaria n.a. 0.01 0.01 2.8 0.9 3.4 0.6 1.2 0.6 1.5 Greece 1.6 0.2 0.3 0.2 0.6 Romania 4.0 2.8 0.7 1.3 2.2 Serbia 0.01 0.01 0.01 0.01 0.0 0.5 0.2 0.3 0.1 0.3 Total 17.9 12.9 10.4 19.0 15.1 49.6 10.7 24.7 16.7 25.4 Total excl. Austria-H. 4.0 2.8 0.7 4.1 11.6 5.5 1.1 1.8 0.9 2.3 Core countries France 11.7 13.4 11.2 9.3 11.4 9.7 8.9 4.8 6.7 7.5 Germany 0.01 2.9 6.4 9.5 4.7 5.2 19.1 15.1 17.2 14.1 United Kingdom 36.9 31.0 28.0 21.0 29.2 24.0 40.5 9.7 11.2 21.4 Total 48.6 47.4 45.6 39.9 45.3 38.9 68.5 29.5 36.1 43.0 Ottoman Empire 14.1 11.2 7.7 4.7 9.4 6.6 3.0 4.7 2.7 4.3 Serbia 1884 1890 1900 1910 avg. South-East Europe Austria-Hungary 72.9 74.2 68.3 18.7 58.5 Bulgaria 0.3 1.4 2.8 2.5 1.8 Greece 0.3 0.1 0.4 0.2 0.2 Romania 1.4 1.5 3.3 4.2 2.6 Serbia Total 74.9 77.1 74.8 25.7 63.1 Total excl. Austria-H. 2.1 3.0 6.5 7.0 4.6 Core countries France 0.6 0.7 3.3 2.6 1.8 Germany 8.5 4.4 15.7 31.1 14.9 United Kingdom 4.5 6.1 3.1 7.2 5.2 Total 13.6 11.2 22.1 40.8 21.9 Ottoman Empire 3.8 5.2 2.2 16.1 6.8 Notes: 1 Not separately reported in national trade statistics. 2 Trade statistics before 1890 do not provide data on final destinations but on the border to which goods were transported. 3 The average only relates to 1890, 1900 and 1913 for the reasons explained in fn. 2. Sources: Trade tables for Greece and Statistical Yearbooks for all other countries. Table 8 Business cycle correlations 1919 – 1941 by country pairs: 1919 – 1941, 1919 – 1929, 1929 - 1941

Au Bu Ro Yu Gr England France Bulgaria 0.59 *** 0.75 *** 0.59 * Romania 0.68 *** 0.71 *** 0.75 *** 0.80 *** 0.90 *** 0.63 ** Yugoslavia 0.79 *** 0.72 *** 0.82 *** 0.70 ** 0.91 *** 0.81 *** 0.92 *** 0.63 ** 0.97 *** Greece -0.09 0.14 0.19 0.24 0.07 0.31 0.23 0.39 -0.10 0.19 0.32 0.26 England 0.51 ** 0.46 ** 0.41 * 0.56 ** 0.29 0.46 0.58 0.12 0.38 0.03 0.58 0.52 * 0.68 ** 0.71 ** 0.57 * France 0.49 ** 0.22 0.32 0.51 ** 0.08 0.51 ** -0.29 0.06 -0.26 -0.09 -0.13 0.32 0.65 * 0.39 0.66 ** 0.79 *** 0.26 0.58 * Germany 0.54 ** 0.26 0.42 * 0.46 ** -0.05 0.56 ** 0.34 0.49 0.22 0.14 0.20 -0.32 0.53 0.02 0.77 * 0.52 0.93 *** 0.89 *** 0.54 * 0.75 *** 0.66 **

Notes: Entries are bilateral correlations of the cyclical component (as computed according to the description in the main text). The first, second and third entry in each cell refer to 1919-1941, 1919-1929 and 1929-1941, respectively. Periods might be shorter, depending on the estimation period (table 3). *, ** and *** denote statistical significance levels of 10%, 5% and 1%, respectively.

Sources: Cf. main text. Table 9 Business cycle correlations 1919 - 1941 (summary statistics) 1919 – 1941, 1919 – 1929, 1929 - 1941

Au Bu Ro Yu Gr England France Germany Summary statistics for SEE-5 Average r 0.49 0.54 0.60 0.64 0.12 0.44 0.32 0.32 vis-à-vis 0.57 0.69 0.65 0.70 0.25 0.31 -0.14 0.15 SEE-5 0.58 0.51 0.71 0.69 0.17 0.61 0.55 0.73

0.48 0.36 0.57 0.11 0.53 0.63 Average r 0.51 0.31 0.38 0.51 0.11 vis-à-vis 0.22 0.29 0.00 0.16 -0.14 E, F, G 0.66 0.48 0.76 0.80 0.45

0.36 0.11 0.63 Summary statistics for SEE-5 and England, France, Germany (28 bilateral correlations) Average r 0.42 0.29 0.60

Notes: Entries are bilateral correlations of the cyclical component (as computed according to the description in the main text). The first, second and third entry in each cell refer to 1919-1941, 1919-1929 and 1929-1941, respectively.

Sources: Table 8.

Table 10 Wilcoxon rank sum test for equality of correlations across periods

Countries Median correlation coefficient Difference of medians across periods (# of bilateral (average correlation coefficient) [p value based on correlations) Wilcoxon rank sum test] 1919-1929 1929-1941 1919-1941 1919-1929 vs. 1929-1941 SEE-5 & core 0.27 0.63 0.46 0.36 (28) (0.29) (0.60) (0.42) [0.00] SEE-5 0.72 0.61 0.64 -0.11 (10) (0.57) (0.53) (0.48) [0.85] Core 0.32 0.66 0.51 0.34 (3) (0.29) (0.67) (0.47) [0.08] SEE-5 vis-à- 0.12 0.65 0.42 0.53 vis core (15) (0.11) (0.63) (0.36) [0.00]

Sources: Cf. main text. Table 11 Correlation of underlying time series with the business cycle 1919-1941 Correlation in percentage at t = 0 (1st entry) and maximum (absolute) correlation at t = -2, -1, 0, + 1 or +2 (where + indicates a lead-factor and – a lag-factor)

Austria Bulgaria Greece Romania Yugoslavia Ø Sectoral indicators agriculture 9 -48 -1 1 31 -1 31 40 -1 -2 60 1 32 32 0 14 communication 65 65 0 28 28 0 -36 -36 0 71 71 0 27 27 0 31 industry 89 89 0 13 40-1 71 71 0 78 78 0 37 37 0 58 mining 36 36 0 47 47 0 45 52-1 18 32 2 41 41 0 37 construction 89 89 0 28 47-1 69 69 0 58 58 0 61 transportation 88 88 0 60 60 0 59 59 0 47 47 0 70 70 0 65 fixedinvestment 28 31 1 64 64 0 61 61 0 87 87 0 60 Fiscal indicators gov.expenditure 74 74 0 68 68 0 47 47 0 -11 36 -2 86 86 0 53 gov.revenue 69 69 0 86 86 0 45 45 0 69 69 0 74 74 0 69 Financial indicators narrowmoney(M0) 71 71 0 66 66 0 86 86 0 49 49 0 37 37 0 62 broadmoney(M3) 81 81 0 55 55 0 65 65 0 83 83 0 50 50 0 67 CPI -3457-23535071710 24 short term interest rate -27 -41 1 -38 -38 0 29 49 -1 -12 Trade indicators terms of trade -26 59 2 -10 23 1 58 58 0 -2 23 2 18 25 1 8 exports 68 68 0 60 60 0 -51 -51 0 83 83 0 69 69 0 46 imports 61 61 0 79 79 0 -61 -61 0 51 51 0 79 79 0 42 Other indicators External spread 23 51 1 23 Realwage 31310 31 Population 22 22 0 28 28 0 27 30 1 26

Sources: Own calculations as described in the main text.

Table 12 Turning points of South-East European business cycles, 1919 - 1941

Austria Bulgaria Romania Yugoslavia Greece 1st peak 1920 1920 1920 1921 1923 1st trough 1927 1926 1926 1927 1926 2nd peak 1930 1929 1929 1929 1929 2nd trough 1933 1934 1933 1933 1931 3rd peak 19371 1937 1937 1937 1934

Notes: 1 The Austrian data ends in 1937 (cf. data appendix).

Sources: Own calculations as described in the main text. Table 13 Trade shares of South-East European countries in 1924, 1929, 1933 and 1939 Percentage of total trade (exports plus imports) Austria Bulgaria 1924 1929 1933 19361 avg. 1924 19302 1933 1939 avg. South-East Europe Austria 10.7 7.3 8.2 n.a. 8.7 Bulgaria 0.0 0.0 0.0 0.0 0.0 Greece 0.0 0.0 0.0 0.0 0.0 5.5 3.0 0.4 0.8 2.4 Romania 4.0 4.3 4.9 7.5 5.2 4.2 3.6 2.5 1.9 3.1 Yugoslavia 6.5 5.4 8.0 5.6 6.4 1.5 0.8 0.2 0.3 0.7 Total 10.5 9.8 12.9 13.1 11.6 21.9 14.7 11.3 3.0 12.7 Total excl. Austria 10.5 9.8 12.9 13.1 11.6 11.2 7.4 3.1 3.0 6.2 Total incl. Austria, 39.3 34.8 34.6 32.3 35.3 30.8 25.4 16.4 8.2 20.2 Hung. & Czechoslov. Core countries France 0.0 0.0 0.0 0.0 0.0 10.0 6.9 3.7 1.0 5.4 Germany 14.3 18.9 18.0 16.7 17.0 19.1 24.9 37.0 66.7 36.9 United Kingdom 0.0 0.0 0.0 0.0 0.0 7.2 4.7 4.0 2.9 4.7 Total 14.3 18.9 18.0 16.7 17.0 36.3 36.5 44.7 70.7 47.1 Turkey 0.0 0.0 0.0 0.0 0.0 3.8 1.6 0.9 0.7 1.8 Greece Romania 19263 1929 1933 1939 avg. 19263 1929 1933 19384 avg. South-East Europe Austria 0.5 1.6 1.5 n.a. 1.2 13.2 11.0 7.8 n.a. 10.7 Bulgaria 3.5 1.3 0.1 0.4 1.3 0.8 1.3 0.5 0.5 0.8 Greece 2.9 2.1 1.7 4.2 2.7 Romania 5.8 4.8 3.7 6.3 5.2 Yugoslavia 4.0 4.1 2.8 2.8 3.4 1.1 1.3 1.1 1.2 1.2 Total 13.8 11.9 8.2 9.5 10.9 18.0 15.7 11.0 5.9 12.7 Total excl. Austria 13.3 10.2 6.7 9.5 9.9 4.8 4.7 3.2 5.9 4.7 Total incl. Austria, 17.3 16.0 11.5 12.8 14.4 34.7 33.6 21.6 21.6 27.9 Hung. & Czechoslov. Core countries France 7.5 6.5 6.4 2.0 5.6 7.4 5.0 11.5 6.1 7.5 Germany 12.6 14.1 14.0 28.9 17.4 17.6 25.8 5.8 31.3 20.1 United Kingdom 11.6 12.3 16.6 6.7 11.8 7.3 6.9 15.2 9.7 9.8 Total 31.7 33.0 37.0 37.6 34.8 32.3 37.7 32.5 47.1 37.4 Turkey 1.8 1.7 1.9 1.2 1.7 1.2 1.2 0.7 1.1 1.1 Yugoslavia 19263 1929 1933 1939 avg. South-East Europe Austria 20.3 16.5 19.1 n.a. 18.6 Bulgaria 0.1 1.0 0.1 0.2 0.4 Greece 4.7 4.2 2.8 1.9 3.4 Romania 7.2 7.8 1.7 1.9 4.7 Yugoslavia Total 32.4 29.5 23.6 3.9 22.4 Total excl. Austria 12.0 13.0 4.5 3.9 8.4 Total incl. Austria, 52.4 47.5 38.7 19.3 39.5 Hung. & Czechoslov. Core countries France 3.5 4.0 3.1 2.3 3.2 Germany 10.6 12.0 13.6 39.2 18.9 United Kingdom 3.3 3.4 5.9 5.9 4.6 Total 17.4 19.4 22.6 47.4 26.7 Turkey 0.4 0.1 0.0 0.1 0.2 Notes: 1 Last data point available for Austria before the Anschluss. 2 No data available for 1929. 3 No data available for 1924. 4 No data available for 1939.

Sources: Statistical Yearbooks of Austria, Bulgaria, Greece, Romania and Yugoslavia.

Table 14 Exports plus imports as share of GDP (openness ratio) South-East Europe, 1920-1937

1920 1925 1929 1933 1937 Austria 46.8 45.1 21.3 27.2 Bulgaria 20.5 28.4 26.2 14.2 21.5 Greece 39.0 61.5 62.5 35.4 43.8 Romania 28.2 21.8 15.0 22.0 Yugoslavia 27.51 25.9 17.3 25.6

Notes: 1 Data refer to 1926.

Sources: Own calculations based on Butschek (1999) for Austria, Ivanov (2012) for Bulgaria, Kostelenos (2007) for Greece, Axenciuc (2012) for Romania and Maddison (2003) for Yugoslavia.