Transcript Morys_presentx - Viessmann European Research Centre

Financial Markets and Financial Regulation:
Sources of Instability or Growth?
International Historical Perspectives
Muenster, Germany, April 12, 2012
Business Cycles in South-East Europe 1870-2000:
A Bayesian Dynamic Factor Model
Dr Matthias Morys (University of York)
Prof. Martin Ivanov (Bulgarian Academy of Sciences)
Business Cycles in Historical Perspective
• “Business cycle paradox”: Research on market integration
suggests high degree of business cycle synchronisation but it
cannot be detected in the data
• Example: Classical Gold Standard period: product and factor
markets were highly integrated & quasi-universal fixed XRs
(Daudin&Morys&O’Rourke 2010)
• BUT Research on business cycles finds low correlation of GDP
growth / de-trended GDP
Example: Research on core countries 1870s - 1913
Source
Average
bilateral
correlation
Countries
Time frame
Statistical
method:
correlation of
Backus&Kehoe
(1992)
0.03
England,
Germany
1870-1913
de-trended GDP
Artis et al. (2011)
0.09
England, France,
Germany
1880-1913
de-trended GDP
Bordo&Helbling
(2011 )
0.04
US, E, F, G, CH,
Netherlands
1880-1913
GDP growth
rates
Bordo&Helbling
(2011)
0.09
F, G, CH, Neth.
(European core)
1880-1913
GDP growth
rates
• equally low correlation for peripheral countries and for core
countries vis-à-vis peripheral countries
• literature as above & Bergman, Jonung for the Nordic
countries
Business Cycles in Historical Perspective
• “Business cycle paradox”: Research on market integration
suggests high degree of business cycle synchronisation but it
cannot be detected in the data
• Example: Classical Gold Standard period: product and factor
markets were highly integrated & quasi-universal fixed XRs
(Daudin&Morys&O’Rourke 2010)
• BUT Research on business cycles finds low correlation of GDP
growth / de-trended GDP
• Two ways to solve the business cycle paradox:
(a) correlation between business cycle synchronization and
market integration not necessarily positive (Krugman 1993):
trade integration  regional specialization  reduced output
synchronization due to industry-specific shocks
(b) business cycle paradox as a figment of poor GDP data
Business cycle paradox a result of poor GDP data?
• 3 concepts to measure GDP (production, expenditure, income)
often give different results (2001 and 2007-9 US recession;
Übele (2011) for German GDP data 1851-1913)
• historical national accounts: interpolation reduces volatility
• fine for GDP levels, but highly problematic for cycles
• for a detailed critique see Ritschl et al. (2008) & Aiolfi et al. (2011)
South-East Europe: we have to make a virtue out of necessity!
• Pre-WW I: GDP estimates on annual basis only for AustriaHungary (Schulze 2000), Bulgaria (Ivanov 2009) and Greece
(Kostelenos et al. 2007)
• only Schulze (2000) in Maddison (2003)
• Post-WW II: conceptual differences between System of National
Accounts and Material Product Accounting (East bloc counterpart)
Dynamic Factor Methodology as Solution to
‘Business Cycle Paradox’?
• Many time series discarded for GDP purposes but useful for business
cycle studies
• In line with Burns&Mitchell 1946...
(“... a cycle consists of expansions occurring at about the same time in many economic
activities, followed by similarly general ... contractions...”)
• ... and the NBER business cycle dating procedure (with its emphasis on
movements in employment over and above GDP data)
• CDFA has strong similarities to Principal Component Analysis
• A cross-section of economic variables share a common factor; extracting
the common factor for the entire period delivers a business cycle index
that is far more accurate than GDP
• Stock&Watson (2010): 250 series to analyse current US business cycle
• Federal Reserve Bank of Chicago’s National Activity Index (CFNAI)
• Economic history: Aiolfi et al. (2011), Ritschl et al. (2008),
Sarferaz&Übele (2009), Übele (2010)
Dynamic factor model
Static factor model
(1) yt
=
(n x 1)
(2) yit
=
(1 x 1)
λ0
+ λ
(n x 1)
λ0i
(n x o)
+ λi
(1 x 1)
(1 x o)
ft
εt
+
(o x 1)
ft
(n x 1)
εit
+
(o x 1)
(1 x 1)
common component
idiosyncratic component
Dynamic factor model: allowance for dynamic properties
(3) ft
=
(o x 1)
(4) εit
=
Φ1
ft-1
(o x o)
(o x 1)
φi1
εi,t-1
+ Φ2
(o x o)
+ φi2
ft-2
(o x 1)
εi,t-2
(all variables are scalars)
ft follows VAR(p) process, εit AR(q) process
+ … + Φp
(o x o)
+ ... + φiq
ft-p
+ ζt
(o x 1)
(o x 1)
εi,t-q + ηit
What time series are we drawing on?
All sorts of time series contain information on business cycles
Burns&Mitchell (1946): locomotive orders, pig iron production, bank
clearings
In recessions women buy more lipsticks but fewer clothes and men
replace their underwear less frequently (Greenspan apparently
collected data of this sort, The Economist, 29.7.2011)
Our approach: 25 time series for each country
1.
Sectoral output indicators: agriculture; communication; industrial
output; mining; construction; transportation; fixed investment
2.
Fiscal indicators: government expenditure & revenues
3.
Financial indicators: M0, M3, domestic interest rate, mortgage credit
4.
Trade indicators: REER, TOT; export & import
5.
Other indicators: real wage; population; CPI
Data mostly taken from official Statistical Yearbooks
For Soviet bloc countries (Bulgaria, Romania, Yugoslavia) doublechecked with Western sources, mainly Vienna Institute for
International Economic Studies
CDFA versus GDP
England, France, Germany, Austria-H. 1879 - 1913
France
(CDFA)
Germany
(CDFA)
England France
(CDFA) (CDFA)
0.78 ***
Germany
(CDFA)
0.81 ***
0.81 ***
Austria-H. 0.67 ***
(CDFA)
0.78 ***
0.84 ***
AH
England
(CDFA) (GDP)
France
(GDP)
England
(GDP)
0.84 ***
0.66 ***
0.72 ***
0.69 ***
France
(GDP)
0.28
0.55 ***
0.35 **
0.49 ***
0.28 *
Germany
(GDP)
0.22
0.15
0.36 **
0.26
0.02
0.48 ***
0.30 *
0.22
0.39 **
0.34 **
0.35 **
Austria-H. 0.23
(GDP)
Extreme cases: Germany – England:
Germany – Austria-Hungary:
Germany
(GDP)
0.02
0.81 versus 0.02
0.84 versus 0.02
Business cycles England, France, Germany
1879-1913
5
Baring crisis
1890
4
American banking
crisis 1907
3
2
1
0
-1
-2
-3
-4
1880
1885
1890
ENGLAND
1895
1900
FRANCE
1905
GERMANY
1910
Questions we want to ask with respect to South-East Europe
1. Has there been an identifiable regional business cycle ?
2. If so, to what extent it was synchronized with W. Europe ?
3. How volatile and how persistent have SEE business cycles been
compared to the ‘core’?
4. Are global / regional or country-specific shocks dominant?
Countries included in our study
• Austria(-Hungary), Bulgaria, Greece, Romania, Serbia/Yugoslavia
• These 5 countries taken together have consistently accounted for
more than 85% of regional GDP from 1870 to the present day
Questions we want to ask with respect to South-East Europe
1. Has there been an identifiable regional business cycle ?
2. If so, to what extent it was synchronized with W. Europe ?
3. How volatile and how persistent have SEE business cycles been
compared to the ‘core’?
4. Are global / regional or country-specific shocks dominant?
Countries included in our study
• Austria(-Hungary), Bulgaria, Greece, Romania, Serbia/Yugoslavia
• These 5 countries taken together have consistently accounted for
more than 85% of regional GDP from 1870 to the present day
Implications beyond SEE...
• European region best suited for analysis of “regional business
cycles”: since when do they exist?
(cf. Bordo&Helbling 2011, Artis et al. 2011, Aiolfi et al. 2011)
• SEE has more countries than Scandinavia (4), Iberia (2) or
Central&Eastern Europe (1) which go back to the late 19th century
Business cycle correlations 1875 – 1913 by country pairs:
Full period, 1893-1913, 1903-1913
AH
Gr
Greece
0.06
0.23
0.60 *
Romania
0.30 *
0.31
0.51
0.26
0.33
0.45
Serbia
0.34 *
0.55 **
0.87 ***
0.27
0.22
0.76 **
Bulgaria
0.34 *
0.28
0.03
England
0.67 ***
0.49 **
0.45
France
Germany
-0.07
-0.25
-0.45
Ro
Se
Bu
E
F
0.19
0.32
0.64 **
0.27
0.37 *
0.22
0.42 **
0.25
0.15
0.25
0.37
0.49
0.13
0.07
0.29
0.24
0.09
0.36
-0.07
-0.29
-0.55 *
0.78 ***
0.85 ***
0.88 ***
0.23
0.12
0.59 *
0.17
0.19
0.56 *
0.48 **
0.46 **
0.70 **
0.25
0.08
-0.19
0.78 ***
0.66 ***
0.73 **
0.84 ***
0.83 ***
0.78 ***
0.08
0.25
0.57 *
0.19
0.23
0.59 *
0.34 *
0.42 **
0.73 **
0.18
0.07
-0.18
0.81 ***
0.77 ***
0.84 ***
0.81 ***
0.86 ***
0.87 ***
G
Business cycles of Austria-H., Romania and Greece
1881 - 1913
1903 - 1913
6
4
3
4
2
2
1
0
0
-2
-1
-4
-2
-6
-3
82
84
86
88
90
92
AUSTRIA
94
96
98
00
ROMANIA
02
04
06
08
GREECE
10
12
03
04
05
06
AUSTRIA
07
08
09
ROMANIA
10
11
12
GREECE
• For the later period we recognise upswings and downswings which are welldocumented for other countries
Long upswing in the early 1900s
Interrupted by the American Banking Crisis of 1907
Another upswing in the 1910s
• Confirms qualitative evidence on SEE countries (Lampe&Jackson 1982)
13
Business cycle correlations 1875 – 1913 (summary stat.):
Full period, 1893-1913, 1903-1913
AH
Ro
Se
Bu
Summary statistics for SEE-5
Average r
0.26
0.13
vis-à-vis
0.34
0.13
SEE-5
0.50
0.34
0.26
0.34
0.45
0.30
0.34
0.61
0.24
0.17
-0.01
Average r
vis-à-vis
E, F, G
0.24 (0.24 without Bulgaria)
0.26 (0.33
“
“ )
0.38 (0.64
“
“ )
0.17
0.35
0.12
0.16
0.33
-0.04
0.48
0.60
-0.31
0.76
0.72
0.70
Gr
0.19
0.25
0.55
0.32 (0.37 without Bulgaria)
0.28 (0.37
“
“ )
0.40 (0.58
“
“ )
• Business cycle integration starts at high levels (compared to previous studies)
and increases the further we come to WW I: within SEE and vis-à-vis core
• Effect more pronounced when Bulgaria omitted (insufficient time span?)
• But: intra-core synchronization levels higher (0.80)
• no regional cycle; SEE increasingly participating in a pan-European cycle
Business cycles 1875 – 1913: concluding remarks
• Increases in business cycle synchronization are
statistically significant (Wilcoxon rank sum tests)
• Our main finding – increasing synchronization of SEE
countries that, however, fails to reach core country levels –
is consistent with research on market integration:
emphasis on increasingly integrated good and factor
markets before WW I but considerable cross-country
differences remain (O’Rourke&Williamson 1999)
• “geography of synchronization”: Austria-Hungary and
Serbia (on the Western edge of the Balkans) more strongly
synchronized, followed by Romania, Greece, Bulgaria (in
that order)
Interwar period: What are our expectations?
Factors suggesting a decrease of synchronization
• de-globalization (in particular reduction in trade and capital flows)
• Currency instability
Factors suggesting an increase of synchronization
• Great Depression as “global shock” (of extraordinary size)
(Backus&Kehoe 1992, Basu&Taylor 1999)
• Formation of trade areas and currency blocs might lead to increased regional
synchronization
Specific to SEE: border changes after WW I might
increase synchronization
• Yugoslavia, Romania gain at the expense of Austria, Hungary and (to a smaller
extent) Bulgaria
• established trade relationships between cities/regions continue even if in
different countries after WW I
• similar rational holds for bank lending
• some evidence for this in the Yugoslav case (Aleksic 2009)
Business cycles II: Interwar Period
Bilateral correlations of cyclical component (1919-1941)
Bu
Bulgaria
Germany
France
England
0.14
0.26
0.22
0.46
0.68
0.19
0.42
0.32
0.41
0.79
0.24
0.46
0.51
0.56
-0.09
0.54
0.49
0.51
-0.05
0.08
0.29
0.32
0.32
0.44
0.42
0.38
0.48
Ro
Yu
Au
Gr
0.71
0.72
0.59
0.82
Romania
0.71
Yugoslavia
0.72
0.82
Austria
0.59
0.68
0.79
Greece
0.14
0.19
0.24
-0.09
Average
0.54
0.60
0.64
0.49
Average excl. Greece
0.67
0.74
0.78
0.69
0.12
• Very high levels of business cycle integration between Bulgaria, Romania,
Yugoslavia and Austria, but Greek case different
• Average correlation between these 4 countries at 0.72, with little change before
and after 1929 (0.79 versus 0.77)
• Correlation intra-core and core-periphery increase after 1929 (to 0.66 & 0.63) but
do not reach intra-SEE-4 synchronization level  SEE regional business cycle
• Only exception is Germany which in the 1930s is as synchronized with SEE-4 as
SEE-4 among each other (0.73)
How to explain the regional business cycle?
And how to explain the Greek exception?
No accepted “narrative” of interwar business cycles but three factors
are seen as particularly relevant (Ritschl&Straumann 2010):
(1) WW I: wartime recession (standard case) versus wartime boom
(mainly neutral countries)? In case of wartime recession, peak
normally 2 – 3 years after end of WW I
(2) impact on the national business cycle from tying to (in the 1920s)
and untying from (in the 1930s) the gold standard
(3) development of bilateral trade patterns
(1) The effect of World War I
6
Greece
4
2
0
-2
-4
-6
20
22
24
26
28
BULGARIA
YUGOSLAVIA
30
32
34
36
38
ROMANIA
GREECE
• All SEE countries follow standard pattern of wartime recession...
• ... but Greece is involved for longer in hostilities due to Greek-Turkish
war (1919-1922)
• Peaks follow end of hostilities with 2 – 3 year lag:
1920 for Austria, Bulgaria, Romania
1921 for Yugoslavia
1923 for Greece
(2) Business cycles and exchange-rate regime
6
4
2
0
-2
-4
Greece
-6
20
22
24
26
28
BULGARIA
YUGOSLAVIA
30
32
34
36
38
ROMANIA
GREECE
• monetary stabilization in the 1920s almost simultaneous (Bulgaria, Greece: 1928;
Romania, Yugoslavia: 1929); exception: Austria (1923)
• 1930s: Austria, Bulgaria, Romania, Yugoslavia adopt deflationary policies cum
foreign exchange controls  Greece leaves gold standard in 1932 and defaults
• Trough of Great Depression:
• 1931 for Greece
• 1933 for Austria, Romania, Yugoslavia
• 1934 for Bulgaria
• supportive of Eichengreen&Sachs (1985): early departure means quicker recovery
(3) Trade patterns show increasing focus on Germany
1929
1933
1937
Export to Germany
Bulgaria
42.5
45.1
47.1
Hungary
42.1
38.2
41.0
Romania
37.0
27.2
24.1
Yugoslavia
24.1
35.6
35.2
Bulgaria
29.8
44.4
58.2
Hungary
33.2
39.5
44.2
Romania
36.6
27.8
40.1
Yugoslavia
33.0
29.3
42.7
Import from Germany
Source: Warriner, 1939: 58.
Summary
On Dynamic Factor Methodology
• Can be employed successfully even for relatively short periods of time such as
the interwar period
Empirical findings
• CDFA suggests much higher business cycle synchronisation than historical
national accounts
Pre-WW I
• increasing business cycle synchronization, both within SEE and vis-à-vis the
core economies
• SEE increasingly participates in pan-European business cycle (i.e., no regional
business cycle)
Interwar period
• emergence of a regional business cycle based on similar wartime experience
(i.e. wartime recession), monetary developments and trade patterns
• Greece is not part of this business cycle but Germany is
Thank you very much !
We are looking forward to your
comments
[email protected]
[email protected]
Business cycles III: Postwar Period
Does trying to detect a regional business cycle make sense
despite the Iron Curtain?
• Was there business cycle integration in the Soviet bloc at all?
• Common external shocks: oil price shocks 1973/1979
• SEE contains the two “unruly children” of the Soviet bloc:
(a) Yugoslavia: after 1949 (break with Stalin) opens economically
towards the West by increasing trade and importing capital; even labour
mobility (Gastarbeiter experience with West Germany)
(b) Romania: similar (though less pronounced) developments after the
oil price shock; becomes IMF member in early 1980s
 Do these interactions lead to a business cycle with Western Europe?
General finding for Austria/Greece confirmed: relatively high integration
with E, F, G in the 1950s and 1960s, less so in 1970s but then again in
the 1980s and 1990s.
Soviet bloc business cycle integration versus
synchronisation with West Germany, 1950s – 1970s
4
4
.3
3
3
.2
2
.1
1
.0
0
-.1
-1
-.2
-2
-.3
-3
-.4
2
1
0
-1
-2
-3
-4
56
58
60
62
64
66
BULGARIA_
Germany
W. Germany
Yugoslavia
68
70
72
74
76
ROMANIA_
Ro
Bu
0.48
-0.30
-0.53
-0.31
-0.70
Romania
-0.30
-0.31
Bulgaria
-0.53
-0.70
1956
1958
1960
1962
YUGOSLAVIA_
Yu
0.48
-.5
1954
0.56
0.56
1964
1966
1968
1970
1972
GERMANY
- Bulgaria and Romania well integrated
- Yugoslavia fairly well integrated with
West Germany but not at all with
Bulgaria, Romania
Soviet bloc business cycle integration versus
synchronisation with West Germany
1950s - 1970
Germany
W. Germany
Yugoslavia
1970 - 1988
Germany
Yu
Ro
Bu
0.48
-0.30
-0.53
W. Germany
-0.31
-0.70
Yugoslavia
0.42
0.56
Romania
0.34
0.57
Bulgaria
-0.39
-0.12
0.48
Romania
-0.30
-0.31
Bulgaria
-0.53
-0.70
0.56
Yu
Ro
Bu
0.42 0.34
-0.39
0.57
- 0.12
0.28
0.28
• Not only Yugoslavia is connected to West European business cycle, but
from the 1970s onwards also Romania
• Bulgaria appears increasingly isolated; supports the political history
story that Bulgaria was a more “committed” member of the Soviet bloc
Summary
On Dynamic Factor Methodology
• Can be employed successfully even for relatively short periods of time such as
the interwar period
Empirical findings
• CDFA suggests much higher business cycle synchronisation than historical
national accounts
Pre-WW I
• increasing business cycle synchronization, both within SEE and vis-à-vis the
core economies
• SEE increasingly participates in pan-European business cycle (i.e., no regional
business cycle)
Interwar period
• emergence of a regional business cycle based on similar wartime experience
(i.e. wartime recession), monetary developments and trade patterns
• Greece is not part of this business cycle but Germany is
Post-WW II
• Onset of the Cold War almost completely extinguishes regional business cycle
integration
• increasing links with the West of Yugoslavia, later Romania sees re-emergence
of a common business cycle vis-à-vis Austria (and West Germany)
Thank you very much !
We are looking forward to your
comments
[email protected]
[email protected]
Business cycles in the interwar period: The Greek exception
6
4
2
0
-2
-4
-6
20
22
24
26
BULGARIA
28
30
32
ROMANIA
34
36
38
YUGOSLAVIA
6
4
2
0
-2
-4
-6
20
22
24
26
28
BULGARIA
YUGOSLAVIA
30
32
34
ROMANIA
GREECE
36
38
Raw slide related to table 1
Source
Average
correlation
Countries
Time frame
Statistical
method:
correlation of
Morgenstern
(1959)
Backus&Kehoe
(1992)
0.83
E, F, G
1870-1914
concordance index
(NBER tradition)
0.03
E, G
1870-1913
de-trended GDP
Artis et al. (2011)
0.09
E, F, G
1880-1913
de-trended GDP
Bordo&Helbling
(2011 )
0.04
E, F, G, Ne, CH, US
1880-1913
GDP growth rates
Bordo&Helbling
(2011)
0.09
F, G, Ne, CH
1880-1913
GDP growth rates
Uebele (2011)
0.61
E, F, G
1862-1913
CDFA business cycle
indices
Example: Austria-Hungary 1875-1913
4
5
4
4
3
4
3
3
2
3
2
2
1
2
1
1
0
1
0
0
-1
0
-1
-1
-2
-1
-2
-2
-3
-2
-3
-3
-4
-3
-4
-4
-4
-5
1875
-5
1875
1880
1885
1890
BUSINESS_CYCLE
1895
1900
1905
TRANSPORTATION
r = 0.76
1910
-5
1880
1885
1890
1895
BUSINESS_CYCLE
1900
1905
1910
MANUFACTURING
r = 0.63
• transportation: transported goods in million tons (Statistical Yearbooks)
• manufacturing : component of GDP reconstruction by Schulze 2010
• transportation: quintessentially disaggregate data, raw data not influenced by
prices
• the individual series with the highest correlation are in our case: manufacturing,
construction, transportation (often > 0.70), M0, M3, exports, government revenue
Some comments on data reconstruction / estimation
• SEE is often seen as terra incognita by cliometricians but a “new
economic history” of SEE could easily be written
• Main source: official Statistical Yearbooks (publication of which started
shortly after gaining political independence in 1870s & 1880s)
• Comparison with Mitchell’s Historical Statistics often shows major
discrepancies
• For the post-WW II period, local sources for BG, RO & YU were doublechecked with Western sources on Soviet bloc countries (Vienna Institute
for International Economic Studies, Marrer et al.)
• First attempt ever to construct some of the time series based on a unified
methodology for all countries: CPI, REER & TOT
• Time series are de-trended with HP filter (λ = 6.25)
• Bayesian estimation (more flexible when dealing with short time series)