Transcript The business cycle in historical perspective 1870

The business cycle in historical perspective
1870-2009 – change and continuity
Stud. polit Jeppe Druedahl
Supervisor: Paul Sharp
Opponent: Ole Jahn
Seminar: Topics in Economic History
Department of Economics, University of Copenhagen
Presented 5th of April 2011
Business cycles on the research agenda once again
•
The Great Recession has put business cycles on the
research agenda once again
•
In
•
•
•
•
•
Properties of the business cycle
• Which have remained constant?
• Which have changed?
• In what way have they they?
•
Can help us with
• Testing the validity of theories in a broader context
• Guiding us in which direction to develop the theories
2009 World GDP fell 1.95 percent
Too large stocks
General over-capacity
Unemployment
– It is hard to imagine this being Pareto optimal
What is the business cycle? (I)
•
Typical definition: “Business cycles are a type of fluctuation
found in the aggregate economic activity of nations that
organize their work mainly in business enterprises: A cycle
consists of expansions occurring at about the same time in
many economic activities, followed by similarly general
recessions, contractions, and revivals which merge into
the expansion phase of the next cycle; this sequence of
changes is recurrent but not periodic; in duration business
cycles vary from more than one year to ten or twelve years;
they are not divisible into shorter cycles of similar
character with amplitudes approximating their own.” (Burns
and Mitchell, 1946, p. 3)
•
It is a quasi-cycle: “the length of the period and also the
amplitude [is] to some extent variable, their variations taking
place, however, within such limits that it is reasonable to
speak of an average period and an average amplitude”
(Frisch, 1933)
What is the business cycle? (II)
•
My definition:
• Range: 2 to 8 years
• Proxy: GDP
•
Two different types:
• Classical business cycles (absolute fluctuations)
• Growth cycles (fluctuations around trend) (modern)
•
My approach: Bird’s-eye view with many countries and long
time periods focusing on overall patterns and tendencies
•
Alternative: Country specific analysis (narrative or
model based)
What is the business cycle? (III)
•
Four different characteristics
1.
2.
3.
4.
The
The
The
The
duration and persistence of the cycle
amplitude and volatility of the cycle
co-movement of variables with GDP
synchronization of the cycle across borders
•
Five papers on something similar: Backus and Kehoe
(1992), Bergman et al. (1999), Basu and Taylor (1999),
Helbling (2010) and Artis et al. (2011).
•
Plan:
I. A little bit of theory
II. My data and detrending method
III. Each of the characteristics
Theory: Impulse and propagation
•
Basic idea following Frisch (1933): A shock affects the
system and is then propagated (transmitted) through the
economy and through time by varies mechanism creating the
observed cyclical pattern.
• RBC: Technology shocks and inter-temporal optimization
• Keynesian: Demand shocks and nominal rigidities
•
Changes in the business cycle can come from
• Different shocks
• Different propagation (policy or structural changes)
•
Why different shocks?
• If they have changed – something must have made
them do so
• Be aware: Shocks are always defined as deviations from
the “normal” relationship determined by the particular
model being used.
Data (I) – Variables
•
Variables
• Annual GDP data for 19 countries 1870-2009
• For 12 countries also where possible:
• C – Private consumption
• G – Government consumption
• I – Gross investment
• If – Fixed investment
• X – Exports of goods and services
• M – Imports of goods and services
• P – GDP deflator
•
Full data: Canada, Denmark, Finland, France, Italy, Japan,
Netherlands, Norway, Spain, Swede, UK and USA
Only GDP: Australia, Austria, Belgium, (West) Germany,
New Zealand, Portugal and Switzerland
•
•
Note: Only Western countries + Japan (selection bias)
Data (II) – Sources
•
•
•
•
Recent years: OECD
1. priority: Official national statistic agencies,
2. priority: Research projects at these agencies, central
banks and similar
3. priority: Remaining research projects
•
Only historical data from
• National sources
• Angus Maddison (acclaimed international scholar)
•
Resembles that of Backhus and Kehoe (1992)
•
Note: As general rule in this paper all country averages are
calculated using the same countries in all periods. If a
country lacks consumption data for the prewar period, it is
not used to compute the average in the postwar period to
insure consistency.
Data (III) – Quality
•
Quality: Bad – especially for the component and price series
and at business cycle frequencies
•
Alternative GDP measure: Based on OECD and Maddison
only (US is an exception)
•
It is possible that almost all the results presented here
are figments of the data.
Data (IV) – Deflation, splicing, interpolation
•
•
Deflation of series was done using their own price indices
whenever possible
Splicing was done using the level of the most recent series
and growth rate of the old
•
•
GDP: No missing data points 1870-2009:
World Wars create for the other series
•
We look at the periods:
•
•
•
1870-1913: Prewar
1920-1939: Interwar
1949-2009: Postwar
•
Interpolation done so break is in the middle of the gap
• Components of GDP: Adjusted GDP growth rates
• Prices: Three year moving average growth rate
•
The conclusions of the paper are robust to changes here
Data (V) – Detrending
•
Baxter-King bandpass filter
• Uses spectral analysis to extract only cycles with
periodicities ranging from 2-8 years
• Below 2 years: Short run noise
• Above 8 years: Trend movements
•
Applying the filter reduces the series with three years in both
ends. To avoid this and following Bergman et. al. (1998) and
Stock and Watson (1999) AR(4) forecasts and backcasts
were used instead.
•
The natural logarithm was taken so the cyclical component
can be interpreted as percentage deviations from trend
BK-filtered GDP for Denmark
Duration (I)
•
Spectral analysis: A stationary process can be represented
as the superimposition of infinitely many orthogonal cosine
waves with different frequencies.
•
The power (“probability”) of each wave can be determined by
the power spectrum (shown in the periodogram):
•
Following Levy and Dezhbakhsh (2003) Bartletts method
was used in the estimation (details given in the paper)
•
Note: Probably very large standard errors
Duration (II) - Periodogram for Denmark
•
•
Length of cycle: 1/w
Almost the same peak frequency in all periods
Duration (III) – Longer cycles
•
Note: Lower for the whole postwar period, the interwar
period is very short + Great Depression.
Autocorrelation (I)
•
•
First order: Generally positive – not prewar
Higher order: Negative and more so through time
•
Possible explanation of GDP first order pattern:
Negative first order autocorrelation is typical of white noise –
measurement error or more important agricultural sector.
Autocorrelation (II) – first order
•
•
•
•
Prewar to BW: C (financial innovations + rising living
standards = more smoothing), G (new fiscal policy and
welfare state) and X rising.
BW to postBW: All rising except X – especially I (more
knowledge and investment based economy).
Prices: Persistence does not equal rigidity, but the pattern
matches that of the changes in the duration of the business
cycle (except for GM period).
Similar results in Basu and Taylor (1999)
Autocorrelation (III) – higher order
•
•
•
I think it is hard to see any interesting patterns
The very large autocorrelation at higher order for investment
is notable (same argument as before).
Short run price persistence can in models with
staggered price setting (Taylor or Calvo) still explain
the longer cycle (the higher GDP autocorrelation at higher
lags) – but why more price rigidity? More information should
do the opposite lowering uncertainty and thus menu costs?
Amplitude and volatility (I) – results
Amplitude and volatility (II) – explanations
•
Conclusion: Higher in interwar period, but otherwise secular
fall (generally accepted in the literature)
•
Typical explanations
a.
b.
c.
d.
More important service sector
Activist fiscal and monetary policy
Improved inventory management
Innovations in financial markets
•
Stock and Watson (2003) for the postwar period:
Rather “good luck” (explained partly by fewer price and
productivity shocks) and better monetary policy.
•
My view: The Great Depression is an exception and the
above explanations are necessary in the long historical
perspective.
Amplitude and volatility (III) – ratios
•
•
•
•
•
•
C: Constant around 1, maybe with a falling trend (more
consumption smoothing)
G: Has dropped in the postBW period (absence of wars and
large automatic stabilizers)
If: Constant around 3
I: Has risen – inventories must have become more volatile
X and M: Higher during interwar and BW characterized by
low level of globalization
Prices: Rising until BW, fallen steeply in postBW – could be a
sign of more rigid prices (stability could also leads to more
consumption smoothing)
Co-movements (I)
•
Note: Non-significant coefficietns at a 5 percent level using
the Newey-West estimator have been set to zero.
Co-movements (II)
•
C: Procyclical – correlated at lead and lag in the recent period
(more smoothing)
•
G: Acyclical – only high in the heyday of Keynesianism in the
BW period (notice: no correlation can imply good stabilization
policies)
•
I and If: Procyclical – now as much as C (If somewhat
lagging)
•
X: Procylical generally and acyclical in prewar. Interwar and
BW can be explained by restrained capital mobility. postBW
has to be explained by very synchronized business cycles
(accepted – see following slides)
•
M: Procylical generally and acyclical in prewar.
•
Granger causality: Nothing at all – the problem is annual
series.
Synchronization (I) – correlation coefficients
Synchronization (II) – test
•
Conclusion: Clearly rising except for a small fall from
interwar to BW.
• Also for cross of Europa and Anglo-Saxon (not always so
in the literature)
• Japan somewhat disconnected in the recent period
•
Significant: Yes. Generally
• Wilcoxon Rank Sum Test: Null hypothesis of two
independent samples having equally large value
• Some of the small samples is a problem
Synchronization (III) – puzzle
•
•
Puzzle of the U-shaped globalization and the rising
synchronization
Stylized view:
More globalization
More synchronization
Level of
synchronization
Level of
globalization
1870
WWI
WWII
1973
2000
Synchronization (III) – explanation
•
1) Interwar period:
a) Unusual large global shock
b) Short period
c) Same beginning with WWI
•
2) BW>prewar:
a) Less agriculture means fewer idiosyncratic shocks
b) Globalization has more width and depth in BW
c) A'Hearn and Woitek (2000): The low prewar correlation
is a figment of the method (Kitchen and Junglar cycles)
•
Bordo and Helbling (2010) uses a FSVAR to find favorable
evidence for these conclusions.
Conclusion
•
Notice: All the results can be figments of our data.
•
General conclusion: The business cycle is a rather constant
phenomenon across countries and periods – with the interwar
period being an outlier
•
•
•
•
1)
2)
3)
4)
Longer cycles
Less saw-toothed
Lower amplitude
More synchronized across borders
Conclusion
•
Changes in volatility ratios, autocorrelation structures
and co-movements between GDP and the different variables
are possible starting points for explaining this
•
•
•
•
Signs of increasing price rigidity and consumption
smoothing have been found
Other explanations have focused on changes in fiscal and
monetary policy
Sectorial changes is also an important factor
Extending the data set with more variable and using panel
data methods could yield important results
•
•
•
•
•
•
Wages
Money stocks
Interest rates
Sectoral divisions
Employment
Consumer prices
THANK YOU