on the self-fulfilling prophecy of changes in sovereign ...on the self-fulfilling prophecy of...
TRANSCRIPT
On the Self-Fulfilling Prophecy of Changes in Sovereign
Ratings
Ingmar Schumacher
To cite this version:
Ingmar Schumacher. On the Self-Fulfilling Prophecy of Changes in Sovereign Ratings. cahierde recherche 2012-02. 2012. <hal-00668284>
HAL Id: hal-00668284
https://hal.archives-ouvertes.fr/hal-00668284
Submitted on 9 Feb 2012
HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, estdestinee au depot et a la diffusion de documentsscientifiques de niveau recherche, publies ou non,emanant des etablissements d’enseignement et derecherche francais ou etrangers, des laboratoirespublics ou prives.
ON THE SELF-FULFILLING PROPHECY OF CHANGES
IN SOVEREIGN RATINGS
Ingmar SCHUMACHER
February 2012
Cahier n° 2012-02
ECOLE POLYTECHNIQUE CENTRE NATIONAL DE LA RECHERCHE SCIENTIFIQUE
DEPARTEMENT D'ECONOMIE Route de Saclay
91128 PALAISEAU CEDEX (33) 1 69333033
http://www.enseignement.polytechnique.fr/economie/ mailto:[email protected]
On the Self-Fulfilling Prophecy of Changes
in Sovereign Ratings∗
Ingmar Schumacher
Ecole Polytechnique Paris
February 3, 2012
Abstract
We empirically investigate the dynamic interactions between sovereign
ratings and the macroeconomic environment. We use a Panel VAR on
annual data for European countries from 1986-2010. Our results provide
evidence for a significant two-way interaction between the macroeconomic
environment and changes in sovereigns’ ratings. Thus, rating changes are
able to exacerbate a country’s boom-bust cycle.
Keywords: sovereign ratings; Panel VAR; self-fulfilling prophecy.
JEL classification: C33; H6.
∗Ecole Polytechnique Paris, [email protected]. I am grateful for com-
ments to Gaston Giordana, Thomas Matha and Eric Strobl.
1
1 Introduction
The recent changes in sovereign ratings have received considerable attention
from policy makers and researchers alike. Often, changes in ratings are be-
lieved to induce a self-fulfilling prophecy. For example, fears mounted that
governments which are going through a period of crisis would be addition-
ally adversely affected by rating downgrades. As a consequence, downgrades
would induce a self-fulfilling prophecy of instability. Evidence is mounting that
suggests significant effects from changes in ratings to macroeconomic condi-
tions, especially bonds and stock prices as well as defaults (e.g. Cantor and
Packer 1996, Kaminsky and Schmukler 2002, Reinhart 2002, Brooks, Faff, Hillier
and Hillier 2004, Ferreira and Gama 2007), but also from macroeconomic condi-
tions to changes in ratings (e.g. Afonso, Gomes and Rother 2011, Hilscher and
Nosbusch 2010, Mellios and Paget-Blanc 2006). However, the studies just cited
focus on either the effect of ratings on macroeconomic variables or the other way
around.1 In contrast, a study of the self-fulfilling prophecy would require an in-
tegrated framework, allowing for two-way feedbacks between changes in ratings
and changes in macroeconomic conditions. Our contribution in this article is,
thus, to study these feedbacks within a Panel VAR framework.
There have recently been some contributions that question whether changes
in ratings are able to exacerbate a country’s boom-bust cycle, since they find
that changes to ratings were mainly reactions to news (Mora 2006). Our panel
VAR analysis allows us to investigate this question more fully. In particular, the
main result of our study is that we find a significant two-way interaction between
our macroeconomic variables and changes in sovereigns’ ratings, suggesting that
ratings are, indeed, able to exacerbate a country’s boom-bust cycle.1Cantor and Packer (1996) study both but not in a dynamic, interactive way.
2
2 Data and Methodology
Our data consists of consumer sentiment (CCI), Gross Domestic Product (GDP),
Government Deficit (GD) and Population, all of which come from Eurostat,
while the data on the ratings is taken from Fitch’s Complete Sovereign Rating
History and Bloomberg. We focus on European countries since their data is
fully harmonized and thereby comparison is facilitated and meaningful.
The ratings data comes from Fitch, S&P as well as Moody’s and it is the
sovereign long-term rating. We recode the rating in a numerical form, ranging
from 0 for DD to 22 for AAA for Fitch, from 0 for D to 22 for AAA in the
case of S&P, and from 0 for C to 22 for Aaa in the case of Moody’s. We take
the average of the three ratings for each country at each point in time in order
to obtain a balanced picture. In case there are several changes in a sovereign’s
rating within one year we weigh each rating by the number of days that the
rating was active during that year. We then calculate the growth rate of the
rating (in percentage), denoted by g(R), in order to obtain a variable that
better fits within the econometric approach of the Panel VAR (i.e. is essentially
unbounded). A sovereign’s rating has been related to its probability of default
and economic soundness (Reinhart 2002). Thus, changes in ratings should drive
investors’ expectations on their potential returns and household expectations on
their future income. Additionally, changes in ratings affect a sovereign’s cost of
financing its budget deficit (Brooks et al. 2004). In consequence, we also expect
rating changes to impact a sovereign’s deficit.
The variable d(CCI) gives the change in the harmonized consumer senti-
ment index. As described in the background document of the European Com-
mission, (European Commission 2007), the CCI “is the arithmetic average of the
balances (in percentage points) of the answers to the questions on the financial
situation of households, the general economic situation, unemployment expec-
3
tations (with inverted sign) and savings, all over the next 12 months.” Thus,
it is a forward-looking index of the household’s perception on the developments
of their financial situation. With this variable we capture the expectations of
the households in our sample. We anticipate that changes in their sovereign’s
rating should impact their expectations positively.
As the main indicator for the current economic situation we use the growth
rate of GDP per capita. GDP is measured in market prices in Millions of Euro.
We calculate the growth rate of GDP per capita in percentage terms and denote
it by g(GDPpc). We expect a positive impact from g(GDPpc) on a sovereign’s
rating, but a negative impact from changes in a sovereign’s rating on its GDP
growth.
The government deficit is the deficit of the general government and measured
as total expenditure minus revenue. We calculate it relative to GDP in order to
minimize scale effects2 and denote changes in this variable as d(GD/GDP).
Also, this provides us with information on the size of the budget deficit relative
to that of the national economy. This is the only way in which one can quantify
whether a deficit is actually ‘large’. In line with the recent observations, we
expect rating downgrades to follow shocks to d(GD/GDP). Furthermore, given
previous results in the literature we anticipate that government deficits decrease
following rating upgrades.
Based on this data, our sample consists of 26 European countries, namely
Austria, Belgium, Bulgaria, Cyprus, Czech Republic, Denmark, Estonia, Fin-
land, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Lux-
embourg, Malta, Netherlands, Poland, Romania, Slovakia, Slovenia, Spain, Swe-
den and United Kingdom.3
2The results stay essentially unchanged if we use the per capita deficit.3In the Robustness Appendix we exclude those countries that did not have a rating change
during the period of study. We show that this does not lead to qualitative changes to theresults.
4
This gives us an unbalanced dataset consisting of at maximum 363 country-
year observations ranging from 1986 to 2010. The use of the annual data should
minimize potential anticipation effects of fiscal policy changes (Ramey, 2006)
and help us in avoiding spurious results due to cyclical effects. The summary
statistics are provided in Table 1, and the correlations in Table 2.
Table 1: Summary statistics
Variable Mean Std. Dev. Min. Max.
d(GD/GDP) 0 1.711 -6.738 5.945g(GDPpc) 0 5.13 -18.564 19.262d(CCI) 0 6.249 -23.43 18.836g(R) 0 2.768 -17.983 14.454
Looking at the correlations in Table 2 reveals all are significant at the 5%
significance level, apart from the correlations between d(CCI) and d(GD/GDP)
and g(GDPpc) and d(GD/GDP), which are uncorrelated. Changes in consumer
sentiment are positively correlated with per capita GDP growth (0.21). Changes
in a sovereign’s rating are negatively correlated with changes in its budget deficit
(-0.119) but positively with changes in consumer sentiment (0.247).
Table 2: Cross-correlation table
Variables d(GD/GDP) g(GDPpc) d(CCI)
g(GDPpc) -0.001(0.986)
d(CCI) 0.043 0.210(0.419) (0.000)
g(R) -0.119 0.422 0.247(0.024) (0.000) (0.000)
As our estimation strategy we resort to a Panel Autoregressive Regression
(PVAR) with two lags.4 Since we expect all variables to be at least weakly
endogenous we resort to the reduced-form VAR approach as this avoids impos-4The Robustness Appendix shows that the number of lags does not influence the results.
5
ing a detailed structural model. Furthermore, the VAR approach allows us to
identify the dynamic effects of our variables, which we argued in the previous
section to be important for understanding the full interaction between rating
changes and macroeconomic variables. It furthermore allows us to isolate the
individual effects of each variable via orthogonalized impulse responses, which
we decompose based on the Cholesky decomposition (see e.g. Hamilton 1994).
We estimate the model itself via system GMM based on the STATA routine
provided by Inessa Love (see Love and Zicchino 2006). Firstly, we time de-
mean the series,5, which controls for time-specific effects. Secondly, we helmert
transform the variables, which is a forward mean-differencing of the variables
in order to take away fixed effects without introducing serial correlation.
We choose the ordering {d(GD/GDP), g(GDPpc), d(CCI), g(R)}. Due to
the Cholesky decomposition, a variable is allowed to react in the same period
to all variables ordered before it, but does not contemporaneously react to any
of the variables ordered after it.6 Our ordering is based on the view that the
government deficit impacts GDP directly (e.g. Ramey 2011), and that the rat-
ings are responding to macroeconomic conditions only contemporaneously (e.g.
Mora 2006). Thus, we align ourselves with the results in Mora (2006), namely
that ratings react to news, and thereby set the stage against a contemporane-
ous feedback from ratings to macroeconomic variables. If we, even in this case,
find evidence in favor of a two-way relationship, then this would provide the
strongest support for the self-fulfilling prophecy.5This is done via calculating the average of each variable at each point in time, and then
subtracting these from the actual variables.6In the Robustness Appendix we discuss that the results are not qualitatively affected by
the ordering.
6
3 Results
The results for the variance decomposition are shown in Table 3, while the im-
pulse response results are presented in Figure 1. The impulse responses use
5% confidence bands generated by Monte Carlo simulations with 1000 repli-
cations. Overall, we find significant dynamic interactions between changes in
countries’ ratings and their macroeconomic environment, providing support for
the self-fulfilling prophecy.7
Table 3: Variance decomposition
Equation d(GD/GDP) g(GDPpc) d(CCI) g(R)
d(GD/GDP) 98.92 0.30 0.11 0.66g(GDPpc) 1.50 84.43 2.28 11.79d(CCI) 4.55 7.49 84.21 3.75g(R) 2.40 3.50 10.07 84.03
In particular, our results show that 84% of the variance in sovereign ratings
changes can be attributed to an own shock, while the rest of the variance is
explained by changes in a sovereign’s government deficit (2.4%), by per capita
GDP growth (3.5%) and by changes to consumer sentiment (10.07%). Thus,
though in line with the previous literature on the sustainability of government
finances (e.g. Afonso et al. 2011), we also find a relevant role for per capita GDP
growth and for expectations.
> Figure 1 about here <
The impulse responses in Figure 1 show significant feedbacks from the macroe-7In the Robustness Appendix we present the complete robustness studies. We studied sta-
tionarity, all variables were stationary at any lag length with or without trends. We excludedall those countries that did not have a rating change during our period of study (Austria,France, Germany, Luxembourg, Netherlands, United Kingdom), without a change to the re-sults. We changed the ordering of the variables and varied the lag structure and includedadditional controls. We used the variables without controlling for time-specific effects andwe also did not helmert transform them. All of these did not affect the main results of ouranalysis.
7
conomic variables to changes in the ratings. The effect of changes in a country’s
GDP growth rate is short-run and impacts a country’s rating only contempo-
raneously. A one percentage point increase in a country’s GDP growth rate
increases the growth rate of that country’s rating by 0.4 percentage points.
Changes to consumer sentiment lead to a statistically significant and long-term
increase in a sovereign’s rating. Here we find that a one standard deviation
increase in consumer sentiment raises a country’s growth rate of its rating by
roughly 0.5 percentage points. Though marginally statistically insignificant,
we find a sovereign’s rating growth will be reduced following a one standard
deviation increase in that country’s government deficit-to-GDP ratio.8
The effects from changes in ratings to a country’s macroeconomic environ-
ment are also non-negligible. The results indicate that changes in ratings can
explain 0.66% of the variance in the government deficit, 3.75% of the changes
in consumer sentiment, and up to 11.79% of the variance it the country’s per
capita GDP growth rate.
Our impulse response results show that both per capita GDP growth and
changes in consumer sentiment are significantly positively related to changes in
ratings. We find that the effect of rating changes on per capita GDP growth
works its way through changes in consumer sentiment. A one percentage point
increase in a country’s growth rate of its rating increases per capita GDP growth
by roughly 1.1 percentage points after two years, while its consumer confidence
increases by approximately 1.2 points within one year. A country’s government
deficit marginally increases after an increase in its rating. We find that a one
percentage point increase in a country’s growth rate of its rating increases a
country’s deficit-to-GDP ratio by roughly 0.15 points. This effect arises af-
ter two years but it is short-term. Hence, we conclude that there is evidence8In our Robustness Appendix we show that this effect may become statistically significantly
different from zero depending on the time period, the sample or the transformation we applyto the variables.
8
that changes in ratings may induce countries to take a more sustainable fiscal
position.
4 Conclusion
In this article we find evidence for the self-fulfilling prophecy caused by changes
in sovereigns’ ratings, suggesting that ratings are, indeed, able to exacerbate
a country’s boom-bust cycle. Thus, ratings seem to have a similar impact
as marking-to-market of balance sheets. While marking-to-market of balance
sheets may lead to fire sales and additional rounds of feedbacks between asset
sales and asset prices (Plantin, Sapra and Shin 2008), thereby potentially ren-
dering an otherwise sound institution illiquid, rating changes may exacerbate
a sovereign’s boom-bust cycle by two-way feedbacks between its rating and its
macroeconomic condition.
The obvious advantage of sovereign ratings is that they provide debt holders
and investors with an idea about the probability of a sovereign’s default. An-
other advantage is that rating downgrades will place pressure on governments to
address structural problems that otherwise might get postponed and potentially
result in larger costs than those incurred by immediately tackling the problems.
The disadvantage, as we have shown, arises from the fact that changes in rat-
ings can induce a downward spiral and essentially aggravate existing problems.
To find the welfare trade-offs between the advantages and the disadvantages as
well as potential policy solutions should prove to be a fruitful future research
agenda.
9
References
Afonso, A., P. Gomes, and P. Rother, “Short-and long-run determinants
of sovereign debt credit ratings,” International Journal of Finance & Eco-
nomics, 2011, 16 (1), 1–15.
Brooks, R., R.W. Faff, D. Hillier, and J. Hillier, “The national market
impact of sovereign rating changes,” Journal of Banking & Finance, 2004,
28 (1), 233–250.
Cantor, R. and F. Packer, Determinants and impacts of sovereign credit
ratings, Federal Reserve Bank of New York New York, 1996.
European Commission, The joint harmonised EU programme of business
and consumer surveys: User Guide, European Commission Directorate -
General for Economics and Financial Affairs, 2007.
Ferreira, M.A. and P.M. Gama, “Does sovereign debt ratings news spill over
to international stock markets?,” Journal of Banking & Finance, 2007, 31
(10), 3162–3182.
Hamilton, J.D., Time series analysis, Vol. 2, Cambridge Univ Press, 1994.
Hilscher, J. and Y. Nosbusch, “Determinants of Sovereign Risk: Macroe-
conomic Fundamentals and the Pricing of Sovereign Debt*,” Review of
Finance, 2010, 14 (2), 235–262.
Kaminsky, G. and S.L. Schmukler, “Emerging market instability: do
sovereign ratings affect country risk and stock returns?,” The World Bank
Economic Review, 2002, 16 (2), 171–195.
Love, I. and L. Zicchino, “Financial development and dynamic investment
behavior: Evidence from panel VAR,” The Quarterly Review of Economics
and Finance, 2006, 46 (2), 190–210.
10
Mellios, C. and E. Paget-Blanc, “Which factors determine sovereign credit
ratings?,” The European Journal of Finance, 2006, 12 (4), 361–377.
Mora, N., “Sovereign credit ratings: guilty beyond reasonable doubt?,” Jour-
nal of Banking & Finance, 2006, 30 (7), 2041–2062.
Plantin, G., H. Sapra, and H.S. Shin, “Marking-to-Market: Panacea or
Pandora’s Box?,” Journal of Accounting Research, 2008, 46 (2), 435–460.
Ramey, V.A., “Can Government Purchases Stimulate the Economy?,” Journal
of Economic Literature, 2011, 49 (3), 673–685.
Reinhart, C.M., “Default, Currency Crises, and Sovereign Credit Ratings,”
World Bank Economic Review, 2002, 16 (2), 151–169.
11
Fig
ure
1:Im
puls
ere
spon
ses
resp
. of d
(GD
/GD
P)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.7
994
2.05
89
resp
. of d
(GD
/GD
P)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.1
703
0.27
47
resp
. of d
(GD
/GD
P)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.2
411
0.18
61
resp
. of d
(GD
/GD
P)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
824
0.29
00
resp
. of g
(GD
Ppc
) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.9
718
0.36
11
resp
. of g
(GD
Ppc
) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.3
618
4.14
50
resp
. of g
(GD
Ppc
) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.0
507
0.75
42
resp
. of g
(GD
Ppc
) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
0.00
00
1.47
26
resp
. of d
(CC
I) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.8
360
1.64
06
resp
. of d
(CC
I) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-1.3
664
2.26
91
resp
. of d
(CC
I) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.9
901
6.46
21
resp
. of d
(CC
I) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.2
613
1.51
89
resp
. of g
(R)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.7
400
0.20
85
resp
. of g
(R)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.4
313
0.60
94
resp
. of g
(R)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.0
254
0.74
41
resp
. of g
(R)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.0
842
2.14
36
12
ROBUSTNESS APPENDIX to
On the Self-Fulfilling Prophecy of Changes in Sovereign
Ratings
Ingmar Schumacher
Ecole Polytechnique Paris
February 3, 2012
Abstract
We empirically investigate the dynamic interactions between sovereign ratings and the
macroeconomic environment. We use a Panel VAR on annual data for European countries
from 1997-2010. Our results provide evidence for a significant two-way interaction between
the macroeconomic environment and changes in sovereigns’ ratings. Thus, rating changes are
able to exacerbate a country’s boom-bust cycle.
Keywords: sovereign ratings; Panel VAR; self-fulfilling prophecy.
JEL classification: C33; H6.
1
This is the robustness appendix to the article entitled “On the Self-Fulfilling Prophecy of
Changes in Sovereign Ratings”. In Table 1 we present panel data unit root tests based on the
Fisher test (see Baltagi (2005). Table 1 shows that all variables are stationary.
Table 1: Fisher-type unit root tests
1 lag
d(GD/GDP) g(GDPpc) d(CCI) g(R)Inverse chi-squared 430.14 176.1 285.83 125.71p-value 0.00 0.00 0.00 0.00Mod. inv. chi-squared 36.19 11.75 22.3 6.9p-value 0.00 0.00 0.00 0.00
1 lag and trend
d(GD/GDP) g(GDPpc) d(CCI) g(R)Inverse chi-squared 374.53 137.32 201.36 164.6p-value 0.00 0.00 0.00 0.00Mod. inv. chi-squared 30.8 8.01 14.18 10.65p-value 0.00 0.00 0.00 0.00
H0: All panels contain unit roots. Results hold with any number of lags.
We checked whether our results were not driven by the last crisis. Thus, we dropped the
observations from 2008 onwards and re-ran the analysis based on this sub-sample. The results,
shown in 1, are qualitatively the same. However, the response the growth rate in ratings to
the growth of GDP per capita is now more pronounced and is statistically significantly different
from zero from one year after the shock to GDP per capita until six years after the shock. In
addition, the growth rate of the ratings is now statistically significantly affected by a shock in the
deficit-to-GDP ratio.
As an additional robustness check we varied the lag structure. In Figure 2 we present the
model with one lag, while in Figure 3 we present the results with three lags. Again, we find no
remarkable changes to our results above.
The assumed ordering tends to be important for the impulse responses due to the Cholesky
decomposition. Thus, we inspected the robustness of the results with alternative orderings of the
variables, some of which we present in Figure 4 and 5. The ordering in Figure 4 is {g(R), d(CCI),
d(GD/GDP), g(GDPpc)}, while the one in Figure 5 is {d(CCI), g(R), g(GDPpc), d(GD/GDP)}.
We find that the only relevant difference arises in the response of the growth rate in the rating to
that in GDP per capita. Basically, if the growth in GDP per capita cannot contemporaneously
affect the growth rate in the ratings, then the ratings will be unaffected by a shock to GDP per
capita. Apart from this we find no important qualitative differences in the results, indicating that
the ordering has no significant impact on our analysis.
2
We then constrained the sample to only include those countries that had changes in their
ratings during the period of observation. The impulse responses for that case are shown in Figure
6. Our results are unchanged and the analysis is fully robust to the exclusion of those countries.
As an additional robustness analysis we include further variables. Omitted variable bias is
one way in which results in a VAR could be biased. We, thus, include two additional controls,
the inflation rate and changes in the balance of payments relative to GDP. The impulse response
results are shown in Figure 7. The inflation rate has been found to be one determinant of ratings
(see e.g. Cantor and Packer 1996). We find a negative relationship between HICP and changes
in ratings, and thus confirm the results in Cantor and Packer (1996). The variable balance of
payments accounts for outflows or inflows of goods and capital. While we find that there is a
two-way relationship between balance of payments and ratings, we also observe that our previous
results remain unchanged.
As a further robustness analysis we cut all variables at the 1% and 99% tails. In this way
we analyze whether potential outliers may drive our results. This reduces the sample by 39
observations. The results are presented in Figure 8. We find that now the growth rate of the
ratings responds not only negatively but also statistically significantly different from zero to a
shock in the deficit-to-GDP ratio. In contrast, while the response of the growth rate in ratings to
per capita GDP growth is still positive, it is not statistically different from zero any longer.
As final robustness exercises we do not time de-mean the series in order to see whether our
results may be robust without taking care of time-specific effects. The impulse responses of this are
shown in Figure 9. Our results continue to hold. In addition, we find a much stronger statistically
significant relationship between ratings and the deficit-to-GDP ratio. Specifically, we find that a
shock to ratings growth now decreases the deficit-to-GDP ratio, while an increase in sovereign’s
deficit-to-GDP ratio reduces its ratings. It is possible that this statistically stronger result is
driven by the responses of consumer sentiment and per capita GDP, which now react negatively
and statistically significantly different from zero to a shock in its sovereign’s deficit-to-GDP ratio.1
In Figure 10 we present the impulse responses of our model without time de-meaning and without
using fixed effects. The response of the growth rate in ratings to a shock in the deficit-to-GDP
ratio is now even stronger. It is possible that this is driven by the stronger response of the per
capita GDP growth to a shock in the deficit-to-GDP ratio. Since this effect is absent when one
time de-means the data, then we conclude that this result may be driven by time-specific effects.1The negative impact of the deficit-to-GDP ratio on GDP growth has also been found in Bruckner and Pappa
(2010) and Juessen and Linnemann (2012).
3
References
Baltagi, B.H., Econometric analysis of panel data, Wiley, 2005.
Bruckner, M. and E. Pappa, Fiscal expansions affect unemployment, but they may increase it,
Centre for Economic Policy Research, 2010.
Cantor, R. and F. Packer, Determinants and impacts of sovereign credit ratings, Federal Re-
serve Bank of New York New York, 1996.
Juessen, F. and L. Linnemann, Government spending and unemployment in the OECD: Evi-
dence from an annual panel VAR, TU Dortmund University, 2012.
4
Fig
ure
1:R
obus
tnes
sof
impu
lse
resp
onse
s:E
xclu
ding
rece
ntcr
isis
(i.e
.T
<20
08)
resp
. of d
(GD
/GD
P)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.3
594
1.81
63
resp
. of d
(GD
/GD
P)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.3
190
0.11
44
resp
. of d
(GD
/GD
P)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.3
413
0.22
35
resp
. of d
(GD
/GD
P)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
670
0.30
45
resp
. of g
(GD
Ppc
) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.6
878
0.35
36
resp
. of g
(GD
Ppc
) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
0.00
00
3.60
45
resp
. of g
(GD
Ppc
) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.1
630
0.72
05
resp
. of g
(GD
Ppc
) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
878
1.29
18
resp
. of d
(CC
I) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.7
092
1.49
14
resp
. of d
(CC
I) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.7
720
1.43
75
resp
. of d
(CC
I) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.8
575
5.94
88
resp
. of d
(CC
I) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
378
1.20
15
resp
. of g
(R)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.6
763
0.13
53
resp
. of g
(R)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.0
979
0.58
54
resp
. of g
(R)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
0.00
00
0.68
20
resp
. of g
(R)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
0.00
00
1.90
53
5
Fig
ure
2:R
obus
tnes
sof
impu
lse
resp
onse
s:C
hang
ing
lags
(lag
=1)
resp
. of d
(GD
/GD
P)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.5
968
2.03
50
resp
. of d
(GD
/GD
P)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.1
183
0.32
82
resp
. of d
(GD
/GD
P)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.2
307
0.08
60
resp
. of d
(GD
/GD
P)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
326
0.13
68
resp
. of g
(GD
Ppc
) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.6
621
0.32
53
resp
. of g
(GD
Ppc
) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.0
128
4.32
97
resp
. of g
(GD
Ppc
) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.0
475
0.65
98
resp
. of g
(GD
Ppc
) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
0.00
00
0.83
81
resp
. of d
(CC
I) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.3
719
1.24
98
resp
. of d
(CC
I) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.7
179
2.53
35
resp
. of d
(CC
I) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.0
261
6.66
52
resp
. of d
(CC
I) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.0
257
1.48
15
resp
. of g
(R)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.3
966
0.28
90
resp
. of g
(R)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.0
904
0.81
87
resp
. of g
(R)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.0
004
0.80
70
resp
. of g
(R)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
0.00
00
2.25
30
6
Fig
ure
3:R
obus
tnes
sof
impu
lse
resp
onse
s:C
hang
ing
lags
(lag
=3)
resp
. of d
(GD
/GD
P)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.0
422
2.13
18
resp
. of d
(GD
/GD
P)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.4
330
0.30
32
resp
. of d
(GD
/GD
P)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.3
060
0.21
16
resp
. of d
(GD
/GD
P)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.2
244
0.42
20
resp
. of g
(GD
Ppc
) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.6
291
0.71
21
resp
. of g
(GD
Ppc
) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.4
543
4.20
59
resp
. of g
(GD
Ppc
) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.0
787
1.01
20
resp
. of g
(GD
Ppc
) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
0.00
00
1.58
77
resp
. of d
(CC
I) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-2.2
368
1.84
21
resp
. of d
(CC
I) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-1.8
349
2.41
65
resp
. of d
(CC
I) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-1.8
695
6.50
41
resp
. of d
(CC
I) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.4
124
1.86
34
resp
. of g
(R)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.8
540
0.68
91
resp
. of g
(R)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.6
884
0.62
73
resp
. of g
(R)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.1
139
0.75
69
resp
. of g
(R)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
0.00
00
2.09
37
7
Fig
ure
4:R
obus
tnes
sof
impu
lse
resp
onse
s:O
rder
{g(
R),
d(C
CI)
,d(
GD
/GD
P),
g(G
DP
pc)}
resp
. of g
(R)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
152
2.28
70
resp
. of g
(R)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.2
340
0.35
23
resp
. of g
(R)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.7
323
0.11
41
resp
. of g
(R)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.6
763
0.14
22
resp
. of d
(CC
I) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.2
748
2.47
34
resp
. of d
(CC
I) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-1.4
469
6.46
20
resp
. of d
(CC
I) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.8
757
0.51
59
resp
. of d
(CC
I) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-1.3
524
0.34
56
resp
. of d
(GD
/GD
P)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
808
0.27
39
resp
. of d
(GD
/GD
P)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.2
614
0.52
41
resp
. of d
(GD
/GD
P)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.7
860
2.02
81
resp
. of d
(GD
/GD
P)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.1
591
0.29
13
resp
. of g
(GD
Ppc
) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.0
193
1.66
77
resp
. of g
(GD
Ppc
) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.2
910
1.16
30
resp
. of g
(GD
Ppc
) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.0
640
0.36
60
resp
. of g
(GD
Ppc
) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.5
563
3.94
86
8
Fig
ure
5:R
obus
tnes
sof
impu
lse
resp
onse
s:O
rder
{d(
CC
I),g(
R),
g(G
DP
pc),
d(G
D/G
DP
)}
resp
. of d
(CC
I) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-1.2
053
6.76
70
resp
. of d
(CC
I) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.2
816
1.49
78
resp
. of d
(CC
I) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-1.3
483
0.41
10
resp
. of d
(CC
I) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.8
836
0.48
33
resp
. of g
(R)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.0
563
0.82
51
resp
. of g
(R)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.0
995
2.17
99
resp
. of g
(R)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.6
127
0.17
83
resp
. of g
(R)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.7
490
0.10
42
resp
. of g
(GD
Ppc
) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.0
307
1.40
18
resp
. of g
(GD
Ppc
) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.0
006
1.55
95
resp
. of g
(GD
Ppc
) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.4
991
3.94
23
resp
. of g
(GD
Ppc
) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.0
115
0.32
33
resp
. of d
(GD
/GD
P)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.2
350
0.49
22
resp
. of d
(GD
/GD
P)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.2
888
0.29
11
resp
. of d
(GD
/GD
P)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.2
456
0.31
53
resp
. of d
(GD
/GD
P)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.7
471
2.02
13
9
Fig
ure
6:R
obus
tnes
sof
impu
lse
resp
onse
s:E
xclu
ding
coun
trie
sw
ith
nora
ting
chan
ges
resp
. of d
(GD
/GD
P)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.0
810
2.25
91
resp
. of d
(GD
/GD
P)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.1
327
0.52
09
resp
. of d
(GD
/GD
P)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.2
623
0.36
85
resp
. of d
(GD
/GD
P)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
937
0.33
56
resp
. of g
(GD
Ppc
) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.0
151
0.83
74
resp
. of g
(GD
Ppc
) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.4
536
4.31
95
resp
. of g
(GD
Ppc
) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.1
004
0.97
46
resp
. of g
(GD
Ppc
) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.0
505
1.73
46
resp
. of d
(CC
I) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-2.7
152
2.20
79
resp
. of d
(CC
I) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-1.7
993
2.92
69
resp
. of d
(CC
I) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.9
893
6.71
34
resp
. of d
(CC
I) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.4
522
1.79
35
resp
. of g
(R)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.9
423
0.31
94
resp
. of g
(R)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.5
388
0.77
13
resp
. of g
(R)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.0
837
0.90
36
resp
. of g
(R)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
786
2.48
14
10
Fig
ure
7:R
obus
tnes
sof
impu
lse
resp
onse
s:A
ddit
iona
lco
ntro
ls
resp
. of d
(GD
/GD
P)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.0
659
2.21
01
resp
. of d
(GD
/GD
P)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.1
555
0.47
56
resp
. of d
(GD
/GD
P)
to d
(BO
P/G
DP
)s
(p 5
) d(
BO
P/G
DP
)d(
BO
P/G
DP
)(p
95)
d(B
OP
/GD
P)
06
-0.3
764
0.27
70
resp
. of d
(GD
/GD
P)
to H
ICP
s
(p 5
) H
ICP
HIC
P(p
95)
HIC
P
06
-0.1
350
0.46
03
resp
. of d
(GD
/GD
P)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.2
474
0.20
12
resp
. of d
(GD
/GD
P)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
898
0.24
77
resp
. of g
(GD
Ppc
) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.0
878
0.76
26
resp
. of g
(GD
Ppc
) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.4
090
3.93
85
resp
. of g
(GD
Ppc
) to
d(B
OP
/GD
P)
s
(p 5
) d(
BO
P/G
DP
)d(
BO
P/G
DP
)(p
95)
d(B
OP
/GD
P)
06
-1.0
120
0.13
49
resp
. of g
(GD
Ppc
) to
HIC
Ps
(p 5
) H
ICP
HIC
P(p
95)
HIC
P
06
-0.7
444
0.28
64
resp
. of g
(GD
Ppc
) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.3
367
0.64
10
resp
. of g
(GD
Ppc
) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.0
334
1.42
14
resp
. of d
(BO
P/G
DP
) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.0
034
0.00
78
resp
. of d
(BO
P/G
DP
) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.0
086
0.00
55
resp
. of d
(BO
P/G
DP
) to
d(B
OP
/GD
P)
s
(p 5
) d(
BO
P/G
DP
)d(
BO
P/G
DP
)(p
95)
d(B
OP
/GD
P)
06
-0.0
056
0.02
77
resp
. of d
(BO
P/G
DP
) to
HIC
Ps
(p 5
) H
ICP
HIC
P(p
95)
HIC
P
06
-0.0
020
0.00
92
resp
. of d
(BO
P/G
DP
) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.0
055
0.00
13
resp
. of d
(BO
P/G
DP
) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.0
065
0.00
12
resp
. of H
ICP
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.5
110
0.13
33
resp
. of H
ICP
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.1
166
0.31
41
resp
. of H
ICP
to d
(BO
P/G
DP
)s
(p 5
) d(
BO
P/G
DP
)d(
BO
P/G
DP
)(p
95)
d(B
OP
/GD
P)
06
-0.6
444
0.13
72
resp
. of H
ICP
to H
ICP
s
(p 5
) H
ICP
HIC
P(p
95)
HIC
P
06
0.00
00
1.29
50
resp
. of H
ICP
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.2
846
0.16
49
resp
. of H
ICP
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
864
0.20
06
resp
. of d
(CC
I) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-2.3
068
1.91
62
resp
. of d
(CC
I) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-1.4
517
2.41
81
resp
. of d
(CC
I) to
d(B
OP
/GD
P)
s
(p 5
) d(
BO
P/G
DP
)d(
BO
P/G
DP
)(p
95)
d(B
OP
/GD
P)
06
-1.1
847
0.87
87
resp
. of d
(CC
I) to
HIC
Ps
(p 5
) H
ICP
HIC
P(p
95)
HIC
P
06
-1.7
266
0.41
28
resp
. of d
(CC
I) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-1.1
731
6.37
68
resp
. of d
(CC
I) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.3
063
1.70
55
resp
. of g
(R)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.8
589
0.34
00
resp
. of g
(R)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.4
526
0.66
98
resp
. of g
(R)
to d
(BO
P/G
DP
)s
(p 5
) d(
BO
P/G
DP
)d(
BO
P/G
DP
)(p
95)
d(B
OP
/GD
P)
06
-0.8
889
0.18
24
resp
. of g
(R)
to H
ICP
s
(p 5
) H
ICP
HIC
P(p
95)
HIC
P
06
-0.9
816
0.26
94
resp
. of g
(R)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.1
798
0.65
60
resp
. of g
(R)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
131
2.09
03
11
Fig
ure
8:R
obus
tnes
sof
impu
lse
resp
onse
s:C
utti
ng1%
and
99%
tails
resp
. of d
(GD
/GD
P)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.2
312
1.56
12
resp
. of d
(GD
/GD
P)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.1
820
0.22
22
resp
. of d
(GD
/GD
P)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.2
985
0.09
26
resp
. of d
(GD
/GD
P)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
453
0.09
96
resp
. of g
(GD
Ppc
) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.4
873
0.38
85
resp
. of g
(GD
Ppc
) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.0
762
3.81
23
resp
. of g
(GD
Ppc
) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.3
257
0.56
28
resp
. of g
(GD
Ppc
) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.0
726
0.91
23
resp
. of d
(CC
I) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.1
921
1.10
82
resp
. of d
(CC
I) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.8
612
1.84
97
resp
. of d
(CC
I) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-1.2
428
6.15
87
resp
. of d
(CC
I) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
016
1.03
14
resp
. of g
(R)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.4
044
0.13
80
resp
. of g
(R)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.1
342
0.27
70
resp
. of g
(R)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.0
283
0.39
38
resp
. of g
(R)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
0.00
00
1.41
65
12
Fig
ure
9:R
obus
tnes
sof
impu
lse
resp
onse
s:W
itho
utti
me
de-m
eani
ng
resp
. of d
(GD
/GD
P)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.2
171
2.56
16
resp
. of d
(GD
/GD
P)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.5
057
0.22
91
resp
. of d
(GD
/GD
P)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.7
594
0.12
57
resp
. of d
(GD
/GD
P)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.3
165
0.07
73
resp
. of g
(GD
Ppc
) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-2.5
110
0.29
14
resp
. of g
(GD
Ppc
) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.4
992
5.19
68
resp
. of g
(GD
Ppc
) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.2
677
2.11
06
resp
. of g
(GD
Ppc
) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
637
2.15
00
resp
. of d
(CC
I) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-2.3
166
1.24
93
resp
. of d
(CC
I) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-2.6
767
3.60
95
resp
. of d
(CC
I) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-1.4
428
7.71
00
resp
. of d
(CC
I) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.6
277
2.07
35
resp
. of g
(R)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.7
447
0.09
09
resp
. of g
(R)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.3
855
0.75
77
resp
. of g
(R)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
0.00
00
0.59
24
resp
. of g
(R)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
0.00
00
2.32
34
13
Fig
ure
10:
Rob
ustn
ess
ofim
puls
ere
spon
ses:
Wit
hout
tim
ede
-mea
ning
and
wit
hout
fixed
effec
ts
resp
.of d
(GD
/GD
P)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-0.5
745
2.48
75
resp
.of d
(GD
/GD
P)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.2
221
0.37
54
resp
. of d
(GD
/GD
P)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.7
339
0.34
85
resp
. of d
(GD
/GD
P)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.2
858
0.19
67
resp
. of g
(GD
Ppc
) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-2.6
361
0.20
84
resp
. of g
(GD
Ppc
) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.4
751
5.83
87
resp
. of g
(GD
Ppc
) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.4
695
2.28
11
resp
. of g
(GD
Ppc
) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.2
270
2.85
95
resp
. of d
(CC
I) to
d(G
D/G
DP
)s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-2.3
951
1.14
15
resp
. of d
(CC
I) to
g(G
DP
pc)
s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-2.5
872
4.14
19
resp
. of d
(CC
I) to
d(C
CI)
s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-2.8
860
7.27
83
resp
. of d
(CC
I) to
g(R
)s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.8
200
1.55
03
resp
. of g
(R)
to d
(GD
/GD
P)
s
(p 5
) d(
GD
/GD
P)
d(G
D/G
DP
)(p
95)
d(G
D/G
DP
)
06
-1.0
851
0.07
08
resp
. of g
(R)
to g
(GD
Ppc
)s
(p 5
) g(
GD
Ppc
)g(
GD
Ppc
)(p
95)
g(G
DP
pc)
06
-0.4
420
0.98
76
resp
. of g
(R)
to d
(CC
I)s
(p 5
) d(
CC
I)d(
CC
I)(p
95)
d(C
CI)
06
-0.0
288
0.70
02
resp
. of g
(R)
to g
(R)
s
(p 5
) g(
R)
g(R
)(p
95)
g(R
)
06
-0.1
271
2.47
94
14