Journal of Economic Studies
Are shocks t o nat ional income persist ent ? New global evidence
Seema Narayan, Paresh Kumar Narayan,
Article information:
To cite this document:
Seema Narayan, Paresh Kumar Narayan, (2011) "Are shocks to national income persistent?
New global evidence", Journal of Economic Studies, Vol. 38 Issue: 2, pp.218-230, https://
doi.org/10.1108/01443581111128433
Permanent link t o t his document :
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
https://doi.org/10.1108/01443581111128433
Downloaded on: 11 June 2017, At : 18: 57 (PT)
Ref erences: t his document cont ains ref erences t o 9 ot her document s.
To copy t his document : permissions@emeraldinsight . com
The f ullt ext of t his document has been downloaded 491 t imes since 2011*
Users who downloaded this article also downloaded:
(2011),"Public debt and risk premium: An analysis from an emerging economy", Journal of Economic
Studies, Vol. 38 Iss 2 pp. 203-217 <a href="https://doi.org/10.1108/01443581111128424">https://
doi.org/10.1108/01443581111128424</a>
(2011),"Sustainable national income: information for attaining sustainability", Management Research
Review, Vol. 34 Iss 11 pp. 1190-1201 <a href="https://doi.org/10.1108/01409171111178747">https://
doi.org/10.1108/01409171111178747</a>
Access t o t his document was grant ed t hrough an Emerald subscript ion provided by emerald-srm: 393177 [ ]
For Authors
If you would like t o writ e f or t his, or any ot her Emerald publicat ion, t hen please use our Emerald f or
Aut hors service inf ormat ion about how t o choose which publicat ion t o writ e f or and submission guidelines
are available f or all. Please visit www. emeraldinsight . com/ aut hors f or more inf ormat ion.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and pract ice t o t he benef it of societ y. The company
manages a port f olio of more t han 290 j ournals and over 2, 350 books and book series volumes, as well as
providing an ext ensive range of online product s and addit ional cust omer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.
*Relat ed cont ent and download inf ormat ion correct at t ime of download.
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/0144-3585.htm
JES
38,2
Are shocks to national
income persistent?
New global evidence
218
Seema Narayan
RMIT University, Australia, and
Received 19 December 2008
Accepted 25 February 2010
Paresh Kumar Narayan
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
School of Accounting, Economics and Finance, Faculty of Business and Law,
Deakin University, Burwood, Victoria, Australia
Abstract
Purpose – This paper aims to investigate the integrational properties of real GDP for 125 countries.
Design/methodology/approach – The paper applies the Kwiatkowski et al. univariate test and a
KPSS-type univariate test that accounts for multiple structural breaks – a test procedure proposed by
Carrion-i-Silvestre et al. The panel versions of the KPSS-type test, proposed by Carrion-i-Silvestre et al.
with and without structural breaks, are also applied.
Findings – The paper finds that, while univariate tests with and without structural breaks provide
mixed results on persistence, the panel test suggests that shocks to national output are persistent.
Originality/value – This is a multi-country study that focuses on both developed and developing
countries and uses more recent data to provide new and comparable evidence on the persistence of
output.
Keywords Breaks, Gross domestic product, Developing countries
Paper type Research paper
Journal of Economic Studies
Vol. 38 No. 2, 2011
pp. 218-229
q Emerald Group Publishing Limited
0144-3585
DOI 10.1108/01443581111128433
1. Introduction
The literature on unit root hypothesis for macroeconomic variables owes to the
influential work of Nelson and Plosser (1982), who found the US real gross domestic
product (GDP) to be a non-stationary process. While a common representation of real
output was the trend stationary model, where business cycles were modelled as
stationary fluctuations around a linear deterministic trend, the Nelson and Plosser (1982)
findings were inconsistent with the notions of the business cycle. Because their findings
questioned the business cycle behaviour of real output, the issue of whether or not real
output is stationary has become an intensively researched topic in macroeconomics.
While there are several studies on developed countries, evidence on developing
countries is scarce (see a review of studies on developing countries in Narayan (2004,
2008). The aim of this paper is to extend this line of research by testing the persistence of
output for large group of countries using more recent data. In doing so, we analyse a total
of 125 countries. The objective of testing for persistence is met through using the
Kwiatkowski et al. (KPSS, 1992) univariate test, which does not take into account
structural breaks and the KPSS univariate test that accounts for multiple structural
JEL classification – C22, O1, O4
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
breaks – a test procedure proposed by Carrion-i-Silvestre et al. (2005) – in the data series.
Equally importantly, using Monte Carlo simulations, we compute specific sample size
critical values, which is also contingent on the location and number of structural breaks
for each of the countries. We conclude our empirical analysis by using a panel unit root
test developed by Carrion-i-Silvestre et al. (2005) that allows for multiple structural breaks.
This study is novel for two reasons. First, given the importance of knowledge on the
integrational properties of the real output series, while the literature is large there is a
general absence of multi-country studies, particularly on developing countries. To this
end, in this study we examine the largest group of countries in this literature. As a
result, this study is timely for researchers and policy makers in terms of comparing
experiences of different countries. Second, while the bulk of the literature, in testing the
issue of stationarity, treats the null hypothesis as non-stationarity, the KPSS test and
its subsequent extensions to cases of univariate multiple structural break test and the
panel multiple structural break test treat the null hypothesis as stationarity. This is an
important treatment likely to have some influence on the results because the null of
stationarity can be considered to be more natural than the null hypothesis of a unit root
for many economic problems (see Bai and Ng, 2004). Particularly relevant to the panel
data models, this implies that there has to be strong evidence against trend stationarity
to conclude in favour of the non-stationarity of the panel.
We organise the balance of the paper as follows. In section 2, we present the
econometric methodology. In section 3, we discuss the empirical results, while in the
final section we conclude.
2. Multiple structural breaks panel unit root test
In this section, we describe the test for the null hypothesis of stationarity, a test
developed by Carrion-i-Silvestre et al. (2005). This procedure allows for two different
types of multiple structural break effects and is based on the panel data version of the
KPSS univariate test developed by Hadri (2000). This test takes the following form:
yi;t ¼ ai þ
mi
X
k¼1
ui;k DU i;k;t þ bi t þ
mi
X
*
gi;k DT i;k;t þ 1i;t
ð1Þ
k¼1
where subscript i ¼ 1; :::; N individuals and t ¼ 1; :::; Ttime periods; the dummy
*
variable DT i;k;t ¼ t 2 T ib;k for t . T ib;k and 0 elsewhere; and DU i;k;t ¼ 1 for t . T ib;k
and 0 elsewhere, where T ib;k denotes the kth date of the break for the ith individual and
k ¼ 1; :::; mi , mi $ 1. Individual structural break effects (break in the mean) result
when bi – 0. The temporal structural break effects (break in the trend) results when
gi;k – 0[1].
Three features of this test are worth highlighting:
(1) structural breaks have different effects on each individual time series;
(2) structural breaks are not restricted so they may occur at different times; and
(3) series are allowed to have different number of structural breaks.
Carrion-i-Silvestre et al. (2005) then use the Hadri (2000) procedure, which is
constructed using a simple average of the univariate stationarity test in Kwiatkowski
Are shocks to
national income
persistent?
219
JES
38,2
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
220
Table I.
KPSS unit root tests
without a break
Regions/countries
Test
statistic
Regions/
Bandwidth countries
Africa
Algeria
Angola
Burkina Faso
Cameroon
Cote d’Ivoire
DR Congo
Egypt
Ethiopia
Ghana
Kenya
Madagascar
Malawi
Mali
Morocco
Mozambique
Niger
Nigeria
Senegal
South Africa
Sudan
Tanzania
Tunisia
Uganda
Zambia
Zimbabwe
Eastern Europe and Central
Asia
Albania
Armenia
Azerbaijan
Belarus
Bulgaria
Croatia
0.1023
0.1464 * *
0.2285 * * *
0.0701
0.1420 *
0.2070 * *
0.2379 * * *
0.2058 * *
0.2071 * *
0.2481 * *
0.1335 *
0.2454 * * *
0.2431 * * *
0.2572 * * *
0.1883 * *
0.1091
0.1548 * *
0.2432 * * *
0.0959 * *
0.1872 * *
0.1984 * *
0.2375 * * *
0.2247 * * *
0.1183
0.1576 * *
5
5
6
5
5
6
6
5
6
5
5
5
5
5
5
5
5
5
5
5
5
6
5
5
5
0.1755 * *
0.1701 * *
0.1696 * *
0.1676 * *
0.1755 * *
0.1667 * *
2
3
3
3
3
2
Czech Republic
Czechoslovakia
Estonia
Georgia
Hungary
Kazakhstan
Kyrgyz Republic
Latvia
Lithuania
Macedonia
Moldova
Poland
Romania
Russian Federation
0.1863 * *
0.1816 * *
0.1649 * *
0.1731 * *
0.1804 * *
0.1693 * *
0.1540 * *
0.1861 * *
0.1632 * *
0.1862 * *
0.1643 * *
0.1399 *
0.1898 * *
0.1685 * *
2
2
3
2
2
3
3
2
3
2
3
2
2
3
Asia
Bangladesh
Cambodia
China
Hong Kong
India
Indonesia
Japan
Malaysia
Myanmar
Pakistan
The Philippines
Singapore
South Korea
Sri Lanka
Taiwan
Thailand
Vietnam
Latin America
Argentina
Barbados
Bolivia
Brazil
Chile
Colombia
Costa Rica
Dominican
Republic
Ecuador
Guatemala
Jamaica
Mexico
Peru
St. Lucia
Trinidad &
Tobago
Uruguay
Venezuela
Western Europe
Austria
Belgium
Cyprus
Denmark
Finland
France
Germany
Greece
Iceland
Ireland
Test
statistic
Bandwidth
0.2341 * * *
0.2361 * * *
0.2481 * * *
0.2476 * * *
0.2603 * * *
0.2461 * * *
0.1173 * *
0.2438 * * *
0.2057 * *
0.2491 * * *
0.1942 * *
0.2460 * * *
0.2480 * * *
0.2448 * * *
0.2496 * * *
0.2436 * * *
0.2294 * * *
6
5
5
6
5
6
5
6
5
6
5
6
6
6
6
6
6
0.105
0.186 * *
0.151 * *
0.136 * *
0.232 * * *
0.264 * * *
0.249 * * *
4
5
5
5
6
5
5
0.132 *
0.163 * *
0.206 * *
0.146 * *
0.184 * *
0.108
0.232 * * *
5
5
5
6
5
5
6
0.136 * *
0.224 * * *
0.074
5
5
4
0.2100 * *
0.1310 * *
0.2455 * * *
0.1619 * *
0.1672 * *
0.0845
0.1228 *
0.1107
0.1909 * *
0.2201 * * *
5
5
6
5
5
5
5
5
5
6
(continued)
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
et al. (1992), to test the null hypothesis of a stationary panel. To obtain the location and
the number of breaks, Carrion-i-Silvestre et al. (2005) recommend using the Bai and
Perron (1998) procedure, which computes the global minimisation of the sum of
squared residuals.
3. Empirical results
3.1 Data
Annual real GDP data for 125 countries across seven regions, namely, Africa, Asia,
Eastern Europe and Central Asia, Latin America, Middle East, North America, and
Western Europe are used. The countries of these regions are displayed in Table I. The
real GDP data covers the period 1950-2008 for most countries. The only exceptions are
Eastern European and Central Asian countries with a study period 1989-2008. These
country data are extracted from the Groningen Growth and Development Centre and
The Conference Board, Total Economy Database, August 2004, www.ggdc.net. All
data was converted into natural logarithmic form before the empirical analysis.
Are shocks to
national income
persistent?
221
3.2 Results
3.2.1 KPSS univariate tests. We begin the empirical analysis with an examination of
the univariate unit root properties of real GDP for each of the regions using the KPSS
Regions/countries
Serbia and Montenegro
Slovak Republic
Slovenia
Tajikistan
Turkmenistan
USSR
Ukraine
Uzbekistan
Yugoslavia
Middle East
Bahrain
Iran
Iraq
Israel
Jordon
Kuwait
Oman
Qatar
Saudi Arabia
Syria
UAE
Yemen
Test
statistic
Regions/
Bandwidth countries
0.1719 * *
0.1706 * *
0.1721 * *
0.1631 * *
0.1682 * *
0.1687 * *
0.1656 * *
0.1704 * *
0.1643 * *
2
2
2
3
3
3
3
3
2
0.2444 * * *
0.1049 * *
0.1736 * *
0.2377 * * *
0.2369 * * *
0.1255 *
0.2482 * * *
0.1133
0.0903
0.2237 * * *
0.1572 * *
0.2306 * * *
5
5
6
6
5
5
6
5
5
6
5
6
Italy
Luxembourg
Malta
Netherlands
Norway
Portugal
Spain
Sweden
Switzerland
Turkey
United Kingdom
North America
Canada
United States
Oceania
Australia
New Zealand
Test
statistic
Bandwidth
0.1811 * *
0.2356 * * *
0.2177 * * *
0.2043 * *
0.2382 * * *
0.2172 * * *
0.2144 * *
0.1343 * *
0.1496 * *
0.2694 * * *
0.2241 * * *
5
6
6
5
6
5
5
5
5
5
6
0.2267 * * *
0.2404 * * *
5
6
0.2347 * * *
0.1764 * *
6
5
Notes: *Denotes statistical significance at the ????? per cent (query author); * *denotes statistical
significance at the 5 per cent level; * * * denotes statistical significance at the 1 per cent level. The finite
sample critical values for the KPSS test are 0.119, 0.146, and 0.216 at the 10 per cent, 5 per cent, and 1
per cent levels, respectively
Table I.
JES
38,2
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
222
test with a time trend but without a structural break. We report the results in Table I
and organise them as follows. Columns 2 and 4 contain the test statistics for real GDP
and its associated bandwidth is reported in columns 3 and 5.
Our main results for the real GDP series can be summarised as follows. First, the
cases of trend stationary real GDP are limited to a few countries in Africa, Latin America,
Middle East, and Western Europe. In the African region, only four (Algeria, Cameroon,
Niger, and Zambia) out of 21 countries are found with trend stationary real GDP series
over the period 1950 to 2008. In Latin American, only three out of 17 cases of stationarity
are found. These countries with trend stationary real GDP are Argentina, Peru, and
Venezuela. Of the 12 Middle Eastern countries, only Qatar and Saudi Arabia are found to
have a stationary real GDP series. Furthermore, of the 21 Western European countries,
only France and Greece, have trend stationary real GDP series.
Second, for the remaining countries in the rest of the four regions, namely Asia, the
Eastern Europe and Central Asia, North America, and Oceania, we are able to reject the
null hypothesis of stationarity in the real GDP. Taken together, these results imply that
for around 92 per cent of the countries real GDP is non-stationary.
Individual results
Table II.
KPSS test on real GDP for
African countries with
break date(s)
Algeria
Angola
Burkina Faso
Cameroon
Cote d’Ivoire
DR Congo
Egypt
Ethiopia
Ghana
Kenya
Madagascar
Malawi
Mali
Morocco
Mozambique
Niger
Nigeria
Senegal
South Africa
Sudan
Tanzania
Tunisia
Uganda
Zambia
Zimbabwe
Bartlett
0.035
0.023
0.045
0.03
0.028
0.023
0.104 * * *
0.032
0.09
0.054
0.029
0.048
0.034
0.122 * * *
0.018
0.027
0.016
0.027
0.047 * * *
0.023
0.153 * * *
0.145 * * *
0.028
0.029
0.02
Finite sample critical
values
99%
95%
90%
0.054
0.157
0.102
0.188
0.053
0.036
0.033
0.07
0.057
0.081
0.107
0.07
0.132
0.053
0.077
0.052
0.103
0.235
0.054
0.164
0.041
0.089
0.097
0.08
0.111
0.042
0.101
0.07
0.124
0.04
0.03
0.026
0.055
0.045
0.06
0.075
0.055
0.087
0.038
0.054
0.038
0.071
0.155
0.04
0.109
0.031
0.061
0.067
0.06
0.079
0.037
0.08
0.055
0.096
0.034
0.027
0.023
0.048
0.039
0.051
0.062
0.048
0.069
0.033
0.044
0.032
0.057
0.12
0.034
0.085
0.027
0.05
0.053
0.052
0.065
Break dates
1961
1974
1969
1977
1962
1957
1957
1962
1959
1965
1969
1962
1972
1962
1966
1962
1969
1981
1962
1975
1960
1967
1968
1964
1969
1983
1992
1977
1985
1980
1975
1966
1990
1974
1991
1994
1980
1980
1975
1974
1972
1980
1996
1980
1998
1975
1987
1978
1993
1997
1995
2000
1985
1993
1991
1990
1975
1995
1999
2000
1983
1991
1982
1995
1991
1984
1983
1999
1991
1984
2000
1990
2000
1995
1994
2000
.
1992
2000
Notes: *Denotes statistical significance at the 10 per cent level; * *denotes statistical significance at
the 5 per cent level; * * *denotes statistical significance at the 1 per cent level. We computed finite
sample critical values by means of Monte Carlo simulations using 20,000 replications
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
One issue that we have ignored so far is the existence of structural breaks and hence
testing for trend stationarity in their presence. It is now well documented that if
structural breaks are present but are not taken into account when testing for
stationarity then the results are likely to be spurious (see, among others, Vogelsang
and Perron, 1998). Hence, we now consider the KPSS test that allows for multiple
significant structural breaks in the data series. We report the results from this exercise
in Tables II to IX, and organise them as follows. In column 2, we report the KPSS test
statistics, generated through accounting for structural breaks; column 3 contains the
finite sample critical values at the 5 per cent and 1 per cent levels, respectively,
obtained through Monte Carlo simulations; and the final column reports the structural
break date(s). Our main finding here is that when we allow for multiple structural
breaks, we are unable to reject the null hypothesis of stationarity for a large number of
countries at the 5 per cent level of significance or better. All these countries real GDP
are found to be affected by one or more significant structural breaks.
The region-wise results are as follows. In the African region, 80 per cent of the
countries are found to have a stationary real GDP series. Some 64 per cent of these
series are found to be stationary only after accounting for structural breaks. The null
hypothesis of stationarity is rejected for only five countries, namely, Egypt, Morocco,
South Africa, Tanzania, and Tunisia. In terms of the structural breaks, the break dates
for Africa seems to be concentrated around the periods: 1971-1974, 1978-1981, and 1987
to 1996. These break dates are mainly associated with oil price shocks and recessions,
including financial crises.
Bangladesh
Cambodia
China
Hong Kong
India
Indonesia
Japan
Malaysia
Myanmar
Pakistan
The Philippines
Singapore
South Korea
Sri Lanka
Taiwan
Thailand
Vietnam
Bartlett
Finite sample critical
values
99%
95%
90%
0.037
0.019
0.084 * *
0.036
0.028
0.032 * *
0.066 * *
0.032 * *
0.07 *
0.075 * *
0.037
0.015
0.062 * *
0.031
0.067 * *
0.022
0.045
0.048
0.069
0.051
0.313
0.201
0.029
0.042
0.032
0.074
0.032
0.056
0.072
0.049
0.085
0.042
0.037
0.078
0.038
0.05
0.04
0.209
0.131
0.023
0.035
0.027
0.054
0.025
0.043
0.049
0.039
0.058
0.032
0.03
0.053
0.034
0.043
0.035
0.159
0.102
0.021
0.031
0.024
0.045
0.022
0.038
0.04
0.034
0.047
0.028
0.027
0.043
Are shocks to
national income
persistent?
223
Break dates
1959
1964
1960
1987
1978
1957
1957
1957
1965
1958
1957
1965
1961
1967
1960
1957
1966
1971
1974
1975
1993
1986
1997
2000
1966
1971
1972
1987
1971
1983
1973
1979
1976
1973
1969
1979
1974
1989
1985
1997
1979
2000
1984
1993
1986
1987
1988
1991
1984
1997
1997
1989
2000
1997
2000
2000
1997
2000
Notes: *Denotes statistical significance at the 5 per cent levels; * *denotes statistical significance at
the 1 per cent levels. We computed finite sample critical values by means of Monte Carlo simulations
using 20,000 replications
Table III.
KPSS test on real GDP for
Asian countries with
break date(s)
JES
38,2
Bartlett
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
224
Table IV.
KPSS test on real GDP for
Eastern European and
Central Asian countries
with break date(s)
Albania
Armenia
Azerbaijan
Belarus
Bulgaria
Croatia
Czech Republic
Czechoslovakia
Estonia
Georgia
Hungary
Kazakhstan
Kyrgyz Republic
Latvia
Lithuania
Macedonia
Moldova
Poland
Romania
Russian Federation
Serbia and Montenegro
Slovak Republic
Slovenia
Tajikistan
Turkmenistan
USSR
Ukraine
Uzbekistan
Yugoslavia
1.068 * * *
0.965 * * *
1.367 * * *
2.101 * * *
0.789 * * *
0.28 * * *
0.787 * * *
0.521 * * *
0.497 * * *
0.595 * * *
1.072 * * *
0.092
0.34 * * *
0.202 * * *
0.058 * *
1.768 * * *
4.53 * * *
0.376 * * *
0.607 * * *
0.161 * * *
0.902 * * *
0.126 * * *
0.061 *
0.509 * * *
0.4 * * *
0.074 * * *
0.076 * * *
1.761 * * *
2.659 * * *
Finite sample critical
values
99%
95%
90%
0.144
0.052
0.039
0.037
0.039
0.07
0.079
0.09
0.074
0.062
0.062
0.186
0.12
0.106
0.073
0.076
0.074
0.061
0.039
0.046
0.072
0.052
0.078
0.091
0.173
0.073
0.073
0.045
0.071
0.107
0.044
0.034
0.032
0.035
0.058
0.062
0.071
0.057
0.05
0.05
0.134
0.09
0.079
0.058
0.058
0.06
0.049
0.035
0.039
0.058
0.043
0.062
0.07
0.126
0.057
0.057
0.039
0.058
0.09
0.04
0.032
0.03
0.032
0.051
0.054
0.06
0.049
0.044
0.044
0.108
0.077
0.068
0.05
0.048
0.054
0.043
0.032
0.035
0.052
0.039
0.054
0.06
0.104
0.05
0.05
0.035
0.052
Break dates
1992
1991
1991
1991
1991
1992
1991
1991
1993
2001
1991
1997
1991
1991
1991
1993
1993
1991
1991
1991
1992
1992
1991
1991
1997
1991
1991
1991
1992
1994
1994
1995
1994
1998
2002
2003
1998
1994
1996
1999
1994
1995
1994
1996
1999
2000
1996
1994
1998
1998
2002
1996
1994
1995
1995
1998
2000
1998
1998
1997
2004
2003
2002
2005
1999
2005
1998
2000
2003
2004
1999
1997
2005
2004
2004
1998
1998
2002
2005
Notes: *Denotes statistical significance at the 10 per cent levels; * *denotes statistical significance at
the 5 per cent levels; * * *denotes statistical significance at the 1 per cent levels. We computed finite
sample critical values by means of Monte Carlo simulations using 20,000 replications
In Asia, 53 per cent of the countries’ real GDP is found to be stationary after accounting
for structural breaks in the series. The null hypothesis of trend stationarity is rejected
for eight countries, namely China, Indonesia, Japan, Malaysia, Myanmar, Pakistan,
South Korea, and Taiwan. Structural breaks seem to clutter around the periods 1957,
early 1980s, 1997, and 2000. Again, these dates are associated with global economic
shocks, such as financial crises and recessions.
The null hypothesis of stationarity was rejected for an overwhelming 93 per cent of
the Eastern European and Central Asian countries. Only two of these countries
(Croatia’s and Kazakhstan’s) real GDP series are found to be trend stationary. Two
common break dates are found: one in the period 1991-94 and the other in 1998. The
first period was marked by the Gulf war while the second period is one associated with
the Asian financial crisis.
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
Bartlett
Argentina
Barbados
Bolivia
Brazil
Chile
Colombia
Costa Rica
Dominican Republic
Ecuador
Guatemala
Jamaica
Mexico
Peru
St. Lucia
Trinidad and Tobago
Uruguay
Venezuela
0.049
0.018
0.048 * *
0.061 *
0.03
0.061 *
0.072
0.254 * * *
0.034
0.104 * * *
0.041 * * *
0.047
0.05
0.184 * * *
0.091 * * *
0.021
0.05 * * *
Finite sample critical
values
99%
95%
90%
0.143
0.143
0.051
0.085
0.154
0.101
0.136
0.071
0.097
0.058
0.04
0.245
0.161
0.047
0.044
0.039
0.05
0.095
0.094
0.041
0.064
0.1
0.069
0.094
0.053
0.068
0.045
0.032
0.162
0.108
0.035
0.036
0.032
0.04
0.075
0.074
0.036
0.055
0.078
0.056
0.076
0.045
0.056
0.039
0.028
0.127
0.085
0.03
0.032
0.028
0.036
Are shocks to
national income
persistent?
Break dates
1973
1973
1959
1964
1974
1969
1972
1964
1968
1962
1957
1982
1975
1961
1959
1958
1959
1991
1981
1977
1977
1982
1981
1981
1976
1978
1978
1972
1987
1973
1978
1972
1979
2000
1991
1985
2000
225
1998
1998
2000
1998
1986
1980
1981
1993
1981
2000
1988
1994
2000
Notes: *Denotes statistical significance at the 10 per cent levels; * *denotes statistical significance at
the 5 per cent levels; * * *denotes statistical significance at the 1 per cent levels. We computed finite
sample critical values by means of Monte Carlo simulations using 20,000 replications
Bahrain
Iran
Iraq
Israel
Jordan
Kuwait
Oman
Qatar
Saudi Arabia
Syria
United Arab Emirates
Yemen
Bartlett
Finite sample critical
values
99%
95%
90%
0.079
0.034
0.036
0.058
0.029
0.033
0.028
0.018
0.026
0.075
0.035 *
0.13 *
0.247
0.047
0.198
0.156
0.104
0.107
0.093
0.081
0.038
0.158
0.03
0.107
0.155
0.038
0.128
0.111
0.071
0.074
0.067
0.061
0.031
0.112
0.024
0.07
0.119
0.034
0.099
0.092
0.058
0.061
0.056
0.052
0.028
0.092
0.022
0.055
Table V.
KPSS test on real GDP for
Latin American countries
with break date(s)
Break dates
1981
1957
1978
1974
1969
1969
1967
1965
1957
1974
1957
1969
1992
1977
1990
2000
1987
1977
1992
1983
1973
1972
1988
1981
1981
1991
1989
1971
1978
1979
1991
1988
2000
2000
Notes: *Denotes statistical significance at the 1 per cent levels. We computed finite sample critical
values by means of Monte Carlo simulations using 20,000 replications
For the Latin American region, the null hypothesis is rejected for nine countries at the
10 per cent level of significance or better. The real GDP of the remaining eight
countries is found to be stationary. Six of these countries – Barbados Chile, Costa Rica,
Ecuador, Mexico, and Uruguay – are found to be stationary only after allowing for
structural breaks. At least one significant structural break is found in the real GDP
Table VI.
KPSS test on real GDP for
Middle Eastern countries
with break date(s)
JES
38,2
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
226
series of all the countries in this region. There is a tendency for some of these breaks to
commonly appear in the periods: 1972-1973, 1977-1978, 1981-1982, 1998 and 2000.
Again, as with structural breaks of other countries, these break dates tend to coincide
with the oil price shocks of the 1970s, the recessions of the early 1980s, and the
financial crisis of 1997.
For most of the Middle Eastern region, around 67 per cent of countries are found to
have a stationary real GDP series after allowing for structural breaks. The only
exceptions are the real GDP series of the United Arab Emirates and Yemen. The
structural break dates for this region is spread across the review period but there
seems to be some concentration of breaks around the dates 1974-1978 and 1991-1992.
In the case of North American countries, the real GDP series of both Canada and the
USA are found to be stationary after allowing for breaks. The Canadian series
experienced two significant structural breaks. These occurred in 1979 and 1990. The
US series, on the hand, was affected by one break which occurred in 1964.
For the Oceania region, we are unable to reject the null hypothesis for the Australian
real GDP after allowing for four structural break dates in 1952, 1960, 1971 and 1981. In
contrast, the stationarity hypothesis is rejected for the New Zealand series even after
accounting for three breaks.
For the Western European countries, we are able to reject the null hypothesis for 12
out of 21 countries after allowing for structural breaks. These countries are Belgium,
Cyprus, Denmark, Iceland, Ireland, Luxemburg, Malta, Norway, Portugal, Spain,
Turkey, and the UK. Around 43 per cent of the region’s countries are found to have a
non-stationary GDP despite accounting for structural shifts in the series.
Notwithstanding, all countries show two or more significant breaks, several of
which appear around the period 1957-1959, 1971-74, 1978-1981, and 1987-1990.
3.2.2 Panel unit root test. In this section, we report the results from the KPSS panel
stationarity test without breaks and with multiple structural breaks. We present the
Finite sample critical values
Bartlett
Table VII.
KPSS test on real GDP for
North American
countries with break
date(s)
Table VIII.
KPSS test on real GDP for
Oceania countries with
break date(s)
Canada
United States
0.068
0.02
0.213
0.133
0.136
0.097
Break dates
0.106
0.081
1979
1964
1990
Notes: We computed finite sample critical values by means of Monte Carlo simulations using 20,000
replications
Australia
NZ
Bartlett
Finite sample critical values
99%
95%
90%
0.036
0.048 *
0.057
0.054
0.042
0.042
0.035
0.037
Break dates
1952
1951
1960
1976
1971
1989
1981
Notes: *Denotes statistical significance at the 5 per cent levels. We computed finite sample critical
values by means of Monte Carlo simulations using 20,000 replications
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
Austria
Belgium
Cyprus
Denmark
Finland
France
Germany
Greece
Iceland
Ireland
Italy
Luxembourg
Malta
The Netherlands
Norway
Portugal
Spain
Sweden
Switzerland
Turkey
United Kingdom
Bartlett
Finite sample critical
values
99%
95%
90%
0.072 *
0.052
0.027
0.020
0.115 *
0.090 * *
0.036 *
0.134 * *
0.063
0.026
0.031 *
0.053
0.024
0.065 * *
0.020
0.022
0.020
0.080 * *
0.058 * *
0.046
0.031
0.095
0.083
0.060
0.089
0.086
0.090
0.045
0.035
0.194
0.051
0.103
0.123
0.045
0.036
0.047
0.036
0.042
0.043
0.034
0.082
0.049
0.069
0.061
0.046
0.064
0.066
0.065
0.036
0.029
0.129
0.038
0.07
0.084
0.033
0.029
0.036
0.028
0.034
0.035
0.029
0.058
0.039
0.058
0.052
0.04
0.054
0.057
0.055
0.032
0.026
0.099
0.033
0.057
0.068
0.027
0.026
0.031
0.025
0.031
0.032
0.026
0.048
0.034
Are shocks to
national income
persistent?
Break dates
1957
1957
1959
1958
1975
1960
1959
1958
1978
1957
1969
1971
1961
1957
1957
1959
1959
1959
1957
1966
1959
1977
1974
1973
1973
1990
1973
1973
1971
1991
1982
1981
1985
1971
1973
1980
1974
1973
1976
1974
1979
1980
227
1981
1978
1980
1996
1992
1990
2002
1979
1981
1988
1983
1987
1992
1988
1990
1990
1987
1998
1997
1992
1995
1999
2001
1996
Notes: *Denotes statistical significance at the 5 per cent levels; * *denotes statistical significance at
the 1 per cent levels. We computed finite sample critical values by means of Monte Carlo simulations
using 20,000 replications
results in Table X. Our results for the real GDP series are based on eight panels. Our
results are based on the assumption that the long-run variance is homogenous and
heterogenous. Under each of these assumptions, we conduct panel tests by allowing for
structural breaks and by not allowing for structural breaks.
The results can be summarised as follows. First, we find that for all the panels when
we do not allow for structural breaks, we reject the null hypothesis of panel
stationarity. This is indicated by the p-value, which has a value of zero for all the
panels. When we allow for multiple structural breaks, we reject the null hypothesis of
stationarity for all, except, the panel consisting of the North American countries.
4. Conclusion
In this paper, we examined the integrational properties of real GDP of 125 countries
from seven regions of the world, namely, Africa, Asia, Eastern Europe and Central
Asia, Latin America, Middle East, North America, and Western Europe, using annual
data for the period 1950-2004. Knowledge on the integrational properties is important
because if income is found to be non-stationary then it is inconsistent with the notion
that business cycles are stationary fluctuations around a deterministic trend. Our main
innovation, which distinguishes our work from the literature, is that we used a recently
developed technique by Carrion-i-Silvestre et al. (2005) that tests the null of stationarity
by allowing for multiple structural breaks in both univariate and panel data.
Table IX.
KPSS test on real GDP for
Western European
countries with break
date(s)
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
JES
38,2
Africa
Asia
Eastern Europe and Central Asia
Latin America
Middle East
North America
Oceania
Western Europe
228
Table X.
Panel test for stationarity
of real GDP
Regions
No breaks
(homogeneous)
p-value
No breaks
(heterogeneous)
p-value
Breaks
(homogeneous)
p-value
Breaks
(heterogeneous)
p-value
15.844 *
17.894 *
13.136 *
9.632 *
6.002 *
5.878 *
5.667 *
15.215 *
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
13.990 *
16.312 *
13.395 *
10.428 *
9.578 *
5.648 *
4.701 *
13.217 *
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.214
15.142 *
75.192 *
13.741 *
5.572 *
-0.051
3.592 *
5.936 *
0.415
0.000
0.000
0.000
0.000
0.520
0.000
0.000
19.153 *
14.970 *
383.100 *
24.985 *
9.653 *
0.783
2.950 *
16.230 *
0.000
0.000
0.000
0.000
0.000
0.217
0.002
0.000
Notes: *Denotes statistical significance at the 1 per cent level
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
Our main finding is that when we do not account for structural breaks only 11
countries have a stationary GDP series; in other words, around 91 per cent of the
countries have a non-stationary GDP series. However, when we allow for multiple
structural breaks, the number of cases of stationary GDP series increases to 53
countries, implying that for around 42 per cent of the countries GDP is non-stationary.
Finally, when we apply panel data models, both with and without structural breaks, we
unravel overwhelming evidence in favour of panel non-stationary GDP for the eight
panels that we considered. In sum then, our results suggest that while based on
univariate tests, slightly over half of the 125 countries have a GDP series where shocks
have a transitory effect, regional panel results suggest that shocks to GDP are
persistent. This means that country-specific results are somewhat different from
region-based results.
Note
1. It is straightforward to follow multiple structural breaks in a univariate series, starting again
with equation (1) by simply dropping off the subscript i.
References
Bai, J. and Ng, S. (2004), “A new look panel testing of stationarity and the PPP hypothesis”,
in Andrews, D. and Stock, J. (Eds), Identification and Inference in Econometric Modelling:
Essays in Honour of Thomas J. Rothenberg, Cambridge University Press, Cambridge.
Carrion-i-Silvestre, J.L., Barrio-Castro, T.D. and Lopez-Bazo, E. (2005), “Breaking the panels: an
application to the GDP per capit”, Econometrics Journal, Vol. 8, pp. 159-75.
Hadri, K. (2000), “Testing for stationarity in heterogenous panel data”, Econometrics Journal,
Vol. 3, pp. 148-61.
Kwiatkowski, D., Phillips, P.C., Schmidt, P.J. and Shin, Y. (1992), “Testing the null hypothesis of
stationarity against the alternative of a unit root: how sure are we that economic time
series have a unit root”, Journal of Econometrics, Vol. 54, pp. 159-78.
Narayan, P.K. (2004), “Are output fluctuations transitory? New evidence from 24 Chinese
provinces”, Pacific Economic Review, Vol. 9, pp. 327-36.
Narayan, P.K. (2008), “Evidence of panel stationarity from Chinese provincial and regional
income”, China Economic Review, Vol. 19, pp. 274-86.
Nelson, C. and Plosser, C. (1982), “Trends and random walks in macroeconomic time series”,
Journal of Monetary Economics, Vol. 10, pp. 139-62.
Vogelsang, T.J. and Perron, P. (1998), “Additional tests for a unit root allowing for a break in the
trend function at an unknown time”, International Economic Review, Vol. 39, pp. 1073-100.
Further reading
Liu, J., Wu, S. and Zidek, J.V. (1997), “On segmented multivariate regressions”, Statistica Sinica,
Vol. 7, pp. 497-525.
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints
Are shocks to
national income
persistent?
229
This article has been cited by:
Downloaded by RMIT University Library At 18:57 11 June 2017 (PT)
1. Huiming Zhu, Peng Guo. 2016. Are shocks to nuclear energy consumption per capita permanent or
temporary? A global perspective. Progress in Nuclear Energy 88, 156-164. [CrossRef]
2. Yanlin Shi, Lingbing Feng. 2016. A discussion on the innovation distribution of the Markov regimeswitching GARCH model. Economic Modelling 53, 278-288. [CrossRef]
3. Yanlin Shi, Kin-Yip Ho. 2015. Modeling high-frequency volatility with three-state FIGARCH models.
Economic Modelling 51, 473-483. [CrossRef]