Understanding low interest rates

ECONOMIC POLICY NOTE 23/10/2015
Understanding low interest rates
AGNIESZKA GEHRINGER and THOMAS MAYER1

The customary neoclassical model of interest rate determination is neither rooted in the institutional set-up of the credit markets nor supported by the data.

The Wicksell-Mises-Hayek model of the credit and business cycle offers a much better description of reality. In accordance with this model, we find short-term interest rates to have a
strong influence on long-term interest rates and not vice versa, as suggested by the neoclassical model. We also find population ageing not to exert downward pressure on long-term interest rates (and find the opposite effect in half of our sample countries).

As central bank policy makers are more likely than market participants to lack the knowledge
to push market rates to levels consistent with economic fundamentals, there is a high chance
of misalignments of market rates.
In this paper we argue that the customary neoclassical model of interest rate determination,
in which long-term market interest rates are
determined by the supply of and demand for
investable funds, is neither rooted in the institutional set-up of the credit markets nor supported by the data. Instead, we find the WicksellMises-Hayek model of the credit and business
cycle to offer a much better description of reality. From this we conclude that central bank
policy has guided long-term interest rates to
their low level and not vice versa. Other varia-
1
We would like to thank Michael Biggs, Daniel Gros, Joscha
Beckmann, and other colleagues for helpful comments on
earlier versions of this paper.
bles, such as the ageing of the population, have
not added to the downward pressure on interest rates on their own.
Our findings have three important implications:
First, long-term market interest rates are
strongly influenced by central banks’ perceptions of reality rather than by the perceptions of
market participants. Second, as central bank
policy makers are more likely than market participants to lack the knowledge to push market
rates to levels consistent with economic fundamentals, there is a high chance of misalignments of market rates. Third, misalignments of
market rates can cause severe economic distortions, and their correction severe economic
disruptions.
glut” as the main reason for the low level of US
interest rates. In his view, large capital flows to
the US from countries with current account
surpluses, particularly from China, depressed US
long-term interest rates. In 2013, former US
Finance Minister Larry Summers suggested that
the economy had entered a period of very low
growth in which “…the short-term real interest
rate that was consistent with full employment
had fallen to negative two or negative three
percent sometime in the middle of the last decade.”3 His view echoed earlier diagnoses of
“secular stagnation” from the first half of the
20th century, even though these were refuted by
subsequent developments.4 In Germany, Carl
Christian von Weizsäcker has argued that in the
aging economies of the OECD and China, people’s desired savings for retirement exceed
companies’ desired investments with the result
that interest rates may turn negative.5
Why interest rates are low: the policy
makers’ view
In a recent speech, Peter Praet, the chief economist of the European Central Bank, offered an
explanation for the presently low level of interest rates:2 “It is important to understand why
interest rates are so low. And a closer examination reveals that the underlying drivers are not
so much central bank policies as global and euro
area-specific economic factors, some of which
are more secular in nature and others which are
more associated with the legacy of the postLehman financial crisis.”
With regard to the factors that “are more secular in nature”, he said: “If one takes the textbook Solow growth model as an organising device for the different forces driving real interest
rates in the long run, they ultimately pertain to
productivity and population growth, and savings
behaviour. The intuition is that these forces
determine investment and therefore the demand for loanable funds, which have to be
matched by savings.” He then went on to explain that the growth rate of total factor
productivity had been slowing in the euro area
for decades, and population growth had declined from about 0.7% in the early 1970s to
around 0.3% in recent years. Looking into the
very distant future, he expected that the
downward pull from adverse demographics –
notably a rising supply of funds provided by an
ageing population – could result in a significant
reduction in the real rate in the long-run.
Why interest rates are low: The WicksellMises-Hayek view
The views of policy makers (and economists)
referred to above are based on a neoclassical
understanding of the credit market. There,
banks act as intermediaries between savers and
investors. Interest rates (i) adjust to equilibrate
the supply of savings (S) to the demand for
funds to finance investment (I), as show in
equation (1):
(1)
() = ()
If economic fundamentals exert upward pressure on savings and/or downward pressure on
investment, market rates fall to ensure equilibrium in the investable funds market. Central
bank rates move along with market rates as
Praet’s comments echo views on the causes of
low interest rates exposed by other influential
economists in recent years. Already in 2005,
Ben Bernanke, member of the Federal Reserve
Board at this time, pointed to a “global savings
3
Transcript of Larry Summers speech at the IMF Economic
Forum, November 8, 2013.
4
Alvin Hansen, “Economic Progress and Declining Population Growth”, Presidential address to the American Economic Association delivered in Detroit, 28 December 1938.
5
Carl Christian von Weizsäcker, Der Vorsorge Albtraum.
Wirtschaftsdienst Sonderheft 2013, pp.7-15.
2
See Peter Praet, “The low interest rate environment in
the euro area”, Keynote speech at a Pension Funds Conference organised by De Nederlandsche Bank in Bussum,
The Netherlands, 10 September 2015
2
monetary policy reacts to the same pressures
from economic fundamentals as market rates.
pants' perceptions of equilibrium or natural
rates.”6
Even though this model has entered many economic textbooks describing banks as intermediators between the demand for and supply of
investable funds, it is a false description of reality. Banks do not fund lending to investors from
existing deposits obtained from savers. They
create deposits for investors by extending credit
to them out of nothing. The interest rate they
charge borrowers are determined by existing
and expected inter-bank lending rates and
premia to compensate for liquidity and credit
risk associated with lending. Inter-bank lending
rates are important because banks have to borrow from other banks when borrowers move
the deposits created by the extension of credit
to them from one bank to another. These rates
are determined by central bank policy either via
open market operations (as in the US) or lending rates for central bank money to meet reserve requirements (as in the euro area).
We need to add to Borio’s description that it is
unclear at the time of credit extension whether
the real demand for capital goods triggered by it
will eventually be matched by real savings. This
would be the case, if all participants had perfect
foresight. Banks would then set credit rates and
savings rates at levels that would induce a shift
from sight to savings deposits equal to the new
money created for investment through credit
extension. The central bank would have to set
present and expected interbank lending rates
such that the sum out of these rates and the
premia for liquidity and credit risk would match
the equilibrium credit rates.
Obviously, none of the actors involved in this
process has necessary information and foresight
to satisfy the conditions for an eventual equilibrium between savings and investment. It is
much more likely that the extension of bank
credit out of nothing leads to dynamic disequilibria between investment spending and saving.
In our credit money system, such disequilibria
manifest themselves in credit and investment
cycles. The emergence of credit and investment
cycles in a credit money system has been described by Knut Wicksell, Ludwig von Mises and
Friedrich von Hayek.7 Figure 1 gives a stylized
summary of this theory (“WMH” in the following).
Reflecting the real life process of credit extension in our monetary system, Claudio Borio of
the Bank for International Settlement describes
the influence of central banks on credit rates in
the following way: “Central banks set the nominal short-term rate and influence the nominal
long-term rate, through signals of future policy
rates and purchases of assets. Market participants adjust their portfolios based on their expectations of central bank policy, their views
about the other factors driving long-term rates,
their attitude towards risk and various balance
sheet constraints. Given these nominal interest
rates, actual inflation determines ex post real
rates and expected inflation determines ex ante
real rates. Thus, the influence of saving and
investment is only indirect, through these proximate factors and, in particular, through their
influence on central banks' and market partici-
6
Claudio Borio, “On the centrality of the current account
in international economics”, Keynote speech at the ECBCentral Bank of Turkey conference "Balanced and sustainable growth - operationalising the G20 framework", Frankfurt, 28 August 2015.
7
Knut Wicksell, Geldzins und Güterpreise, Jena 1898,
Ludwig von Mises, Geldwertstabilität und Konjunkturpolitik. Jena 1928, Friedrich A. von Hayek, Geldtheorie und
Konjunkturtheorie. Wien/Leipzig 1929.
3
Figure 1. Credit and business cycles according to Wicksell, von Mises, and von Hayek.
GDP
market rate <
natural rate
Investment bust
Deleveraging
market rate = natural rate
t
Source: Own elaborations.
In Figure 1 the “natural rate” is the rate ensuring that all real investment is funded by real
savings (as described above). If the market rate
drops below the natural rate, more credit is
extended to fund additional investment while
the move of newly created money from transaction to savings accounts is discouraged. Growth
of economic activity accelerates, but a part of
investment is directed to marginal projects that
are not viable at the natural interest rate and
hence represent “malinvestment”. When saving
funds become scarce relative to the demand for
investment funds as projects move towards
completion the market rate increases above the
natural rate. Credit collapses and many unfinished investment projects cannot be completed.
A part of the capital stock becomes obsolete
and economic activity plunges. Repeated policy
interventions to soften the recessionary effect
of deleveraging on economic activity may lead
to further swings in the business cycle, albeit
probably with higher frequency and smaller
amplitudes, until another large downturn eliminates remaining misallocated investment. In the
WMH model, the central bank drives the credit
cycle by steering inter-bank lending rates in an
error-correction-process around the natural
rate.
How important is the short-term rate for
the long-term rate?
The views reviewed in the previous two sections
on the main drivers of long-term interest rates
differ fundamentally. The policy makers and
mainstream economists reviewed in the first
section argue that long-term interest rates are
determined by economic fundamentals in the
same way as short-term policy rates. The central bank has no better choice but to adjust its
policy rate to fundamentals. The alternative
view, based on the WMH model presented in
the second section, stipulates that short-term
policy rates exert a key influence on long-term
rates. In this section, we aim to empirically test
the two hypotheses to see which one gives a
better description of reality.
For the test of the first hypothesis (policy rates
and market rates move in tandem), we regress
3-month money market rates on 10-year government bond yields (gby10).8 We use Dynamic
Ordinary Least Squares (DOLS) for the estimation, allowing us to test for unequivocal causality running from the independent to the de8
We use government bond yields to exclude that the
relationship between short- and long-term interest rates is
influenced by the variation of credit spreads.
4
pendent variables (see Appendix). The hypothesis (“Policy Rates Model”) cannot be rejected, if
both rate series are cointegrated (i.e., if residuals of the regression follow a stationary stochastic process of order I(0)). We use the DickeyFuller (DF) test to check for I(0) of the residuals.
We estimate equations for US, Germany, UK,
and Japanese interest rates for the period from
the first quarter of 1991 to the first quarter of
2015 (US and Germany) and the fourth quarter
of 2014 (UK and Japan), respectively.9 Results of
the regressions are given in Table 1. The residuals of the regressions are plotted in Figure 2.
as of mid-1990. A possible explanation of the
difference in results between Japan and the
other countries is that Japan was not directly
affected by the credit boom-bust cycle of the
1990s and 2000s. Japan experienced its own
credit boom-bust cycle during the 1980s and
has been in a state of post-crisis paralysis since
the early 1990s.
Next, based on the same sample, we test the
hypothesis that government bond yields can be
explained by short-term (policy rates) and other
variables (as described by Borio). Our “Bond
Yield Model” is of the following form:
With the exception of Japan, the DF-test does
not reject the null hypothesis of a unit root of
regression residuals, and hence rejects cointegration between short-term and long-term interest rates. Also, the plots of the residuals of
the regressions for US, German and UK shortterm rates point to significant serial correlation
and non-stationarity. Consequently, we must
reject the Policy Rates Model for these countries.
(2)
10 =  + 1 3 + β v + 
where gby10t denotes government bond yields
with 10-year maturity, i3mt 3-month money
market rates, and vt a vector of other variables.
Apart from a time trend variable (included in
the equation to abstract from the declining
trend of long-term rates during the observation
period), vt consists of government debt (or
budget deficits) and (old age) dependency ratios. We use nominal rather than proxis for real
rates as subtracting inflation from both lefthand and right hand side nominal interest rates
to proxy expected real rates would not materially change the relationship10.
For Japan, we cannot reject the Policy Rates
Model. The DF-test does not reject cointegration between the two variables and the plot
points to stationary regression residuals at least
Table 1. DOLS estimates of the Policy Rates Model.
gby10
US
1.146***
(0.064)
Germany
1.156***
(0.067)
UK
1.139***
(0.068)
Japan
0.988***
(0.093)
N. obs.
96
96
96
96
R-squared adj.
0.672
0.787
0.760
0.824
DF test
-2.025
[0.276]
-2.153
[0.224]
-2.591
[0.095]
-5.070***
[0.000]
Notes: Dependent variable is the 3-month money market rate. *, **, and *** denote statistical significance at the 10%, 5%,
and 1% level. Robust standard errors are in parenthesis. The last row reports the test statistic and, in squared parenthesis,
its p-values of the Dickey-Fuller (DF) test. The null hypothesis of the test assumes the presence of a unit root.
10
The correlation between nominal and real rates over our
sample period is 0.92 for short rates and 0.89 for long
rates.
9
Descriptive statistics of the variables are summarized in
Table A2 in the Appendix.
5
3
2
1
1
0
0
-1
-2
-2
-3
-3
3
UK
3
2
2
1
0
0
Q1-91
Q3-92
Q1-94
Q3-95
Q1-97
Q3-98
Q1-00
Q3-01
Q1-03
Q3-04
Q1-06
Q3-07
Q1-09
Q3-10
Q1-12
Q3-13
1
-1
Japan
-1
Q1-91
Q3-92
Q1-94
Q3-95
Q1-97
Q3-98
Q1-00
Q3-01
Q1-03
Q3-04
Q1-06
Q3-07
Q1-09
Q3-10
Q1-12
Q3-13
-1
Q1-91
Q3-92
Q1-94
Q3-95
Q1-97
Q3-98
Q1-00
Q3-01
Q1-03
Q3-04
Q1-06
Q3-07
Q1-09
Q3-10
Q1-12
Q3-13
2
Germany
Q1-91
Q3-92
Q1-94
Q3-95
Q1-97
Q3-98
Q1-00
Q3-01
Q1-03
Q3-04
Q1-06
Q3-07
Q1-09
Q3-10
Q1-12
Q3-13
Figure 2. Residuals of regression of the Policy Rates Model.
USA
3
-2
-2
-3
Source: Own estimations.
The results of our estimations over the same
period as before are given in Table 2.
the US, a one percentage point change in shortterm interest rates leads to a 0.56 percentage
point change in long-term rates. Thus, the decline of US short-term rates of 32 basis points
per year during our estimation period contributed to the reduction of US long-term interest
rates by 18 basis points per year. Effects are
smaller in the UK, Japan and Germany, reflecting the existence of spill-over effects from the
US bond market (captured by US bond yields as
explanatory variables in the estimations for
Germany, UK and Japan).
Based on the results of the Dickey-Fuller unit
root test on the residuals from the DOLS estimations, we cannot reject the hypothesis of cointegration for all estimation equations, suggesting that the Bond Yield Model is not misspecified (e.g. due to omitted variables). Since the
DOLS methodology controls for endogeneity by
accounting for possible influences of the past
and future observations of the explanatory variables, the statistically significant variables in the
equation are “super-exogenous”, meaning that
they reliably determine the dependent variables.
Another important question we pursue in this
paper is the influence of demographic variables
on interest rates. For the US and Germany, we
use the total dependency ratio (defined as the
percentage ratio of people aged 0-14 and over
64 to the working age population (people aged
15-64)). For the UK and Japan we use the old
age dependency ratio (defined as the percent-
Our results show a strong influence running
from short-term to long-term interest rates as
suggested by the WMH model. Across all sample countries coefficients of short-term rates
are statistically and economically significant. In
6
Table 2. DOLS estimates of Bond Yield Model.
US
0.562***
(0.088)
Germany
0.193***
(0.050)
UK
0.243***
(0.038)
Japan
0.223**
(0.101)
0.572***
(0.168)
0.077
(0.074)
0.049
(0.067)
0.573**
(0.194)
-0.032
(0.027)
0.068
(0.060)
0.291**
(0.018)
0.023**
(0.010)
44.75
(134.74)
---
---
---
-1.084
(2.178)
---
---
---
---
0.525***
(0.119)
0.765***
(0.124)
0.310***
(0.087)
time trend
-0.036***
(0.006)
-0.025**
(0.013)
-0.013
(0.005)
-0.204**
(0.066)
N. obs.
90
93
93
89
R-squared adj.
0.931
0.975
0.980
0.959
DF test
-5.487***
[0.000]
-4.287***
[0.000]
-5.500***
[0.000]
-5.111***
[0.000]
i3m
a
Dependency ratio
Government debt (USA and Japan) or
b
deficit (Germany and UK)
USA net foreign liability
b
b
US debt x USA net foreign liability
US long rate
a
Total dependency ratio for the US and Germany, old age dependency ratio for the UK and Japan, as percentage.
In percent of GDP.
Dependent variable is the interest rate on 10-year government bonds. *, **, and *** denote statistical significance at the
10%, 5%, and 1% level. Robust standard errors are in parenthesis. The last row reports the test statistic and, in squared
parenthesis, its p-values of the Dickey-Fuller (DF) test. The null hypothesis of the test assumes the presence of a unit root.
b
age ratio of people aged over 64 to the working
age population (people aged 15-64)). The choice
of the different variables was determined by the
availability of data. For practical purposes, however, the two versions of the dependency ratio
make no difference as both moved together
during the observation period due to the aging
of the population.
population aging much faster in Japan, the effect is much more noticeable there: the doubling of the old age dependency ratio between
1991 and today, corresponding to an average
yearly increase by one percentage point over
the sample period, added 0.6 percentage points
per year to long-term interest rates. Thus, our
findings reject the hypothesis that interest rates
decline in aging societies as people save more
for retirement. To the contrary, results for the
US and Japan lend some support to the alternative view that older populations save less and
demand more capital (and other) resources to
finance retirement.11
We found coefficients for the dependency variables to be positive and statistically significant
at least on the 5% level of error probability for
the US and Japan, and not significantly different
from zero for Germany and the UK. In the US,
an average 0.27 percentage point yearly increase in the dependency ratio (from 98% at the
beginning of the observation period to 103% at
the end of 2014) added 0.16 percentage points
each year to long-term interest rates. With the
11
This is consistent with the findings of Mikael Juselius and
Előd Takáts (“Can demography affect inflation and monetary policy?” BIS Working Papers No 485, February 2015.
7
Finally, government budget deficits and debt
exert a statistically significant upward pressure
on government bond yields in the UK and Japan.
In both countries, the government’s fiscal accounts deteriorated considerably over the estimation period. In the case of Japan, the public
debt ratio increased at an average yearly rate of
6.7 percentage points from 48% of GDP in 1991
to 211% in 2014. This added 0.15 percentage
points to long-term interest rates. Similarly, the
yearly rise in the UK government budget deficit
by 0.1 percentage points of GDP from 0.2% in
1991 to 2.8% in 2014 added 0.03 percentage
points per year to long-term interest rates.
the two variables in the Policy Rates Model and
no significant influence of short-term rates on
long-term rates in the Bond Yield Model. Unfortunately, we cannot unambiguously say what
the trend variable may reflect. It may stand for
other variables that we did not include in our
set of explanatory variables. But for the reasons
given above we can be reasonably sure that the
relationship we found between short-term and
long-term rates is economically and statistically
sound, and not due to spurious correlation.
Evidence for the Wicksell-Mises-Hayek cycle
Having found evidence that short-term interest
rates influence long-term interest rates in line
with the proposition of the WMH model, we can
now explore (in a descriptive way) the interaction between credit and the business cycle also
stipulated in this model.
However, government budget deficit or debt
variables are insignificant in the equations for
Germany and the US. This may well reflect the
high demand for US and German government
bonds as safe assets in global capital markets.
Indeed, the government debt ratio alone enters
the US equation with a negative sign (in a regression equation not shown here), suggesting
that higher debt would lead to lower rates. The
counter-intuitive (“wrong”) sign is more likely to
be due to omitted variables, in this case foreign
demand for US debt as safe assets, than reflecting a true influence. If we control for this by
including US foreign liabilities (in percent of
GDP) and the interaction between supply and
demand for US debt (captured by the product of
the debt and foreign liability ratios), the debt
variable turns insignificant. Thanks to strong
foreign demand for safe assets, the rise in US
and German government debt has been neutral
for the government bond yields of these countries.
The relationship between credit cycles and economic cycles is shown for the euro area and the
US in Figures 3-5. Figure 3 shows the change in
credit flows relative to GDP (which we call
“credit impulse”) and real domestic demand
growth in the euro area.12 As proposed by
WMH, cyclical movements of credit flows drive
real demand flows. The lead of credit over demand is clear in both the downturn of the cycle
in 2007 and the upturn in 2009. Figure 4 shows
the same variables for the US. In this case, the
lead of credit flows is not so clear in the downturn in 2008, but clearly visible in the upturn in
2009.
12
It is important to compare credit flows with demand
flows to identify a relationship between the two variables.
Comparisons of credit stocks and demand flows, as has
been customary in the economic literature, fail to capture
the relationship. See Michael Biggs, Thomas Mayer, and
Andreas Pick, “Credit and Economic Recovery: Demystifying
Phoenix
Miracles.”
March
15,
2010
(http://papers.ssrn.com/sol3/papers.cfm?abstract_id=159
5980), and Michael Biggs and Thomas Mayer, “Bring credit
back into the monetary policy framework, PEFM Policy
Brief, Oxford University, August 2013.
Both short- and long-term interest rates have
been subject to a strong downward trend since
the early 1980s, and this is captured in the trend
variable, which is statistically significant in all
equations. However, the trend has not affected
short-term and long-term rates equally. If it had,
we would have found a cointegration between
8
Figure 3. Credit impulse and demand in the euro area, 1996Q1-2015Q2.
6
%
% of GDP
4
8
4
2
0
14 Q4
14 Q1
13 Q2
12 Q3
11 Q4
11 Q1
10 Q2
09 Q3
08 Q4
08 Q1
07 Q2
06 Q3
05 Q4
05 Q1
04 Q2
03 Q3
02 Q4
02 Q1
01 Q2
00 Q3
99 Q4
-2
99 Q1
0
-4
-8
-4
real domestic demand
-6
credit impulse (r.h.s.)
-12
Source: Authors’ calculations, Haver Analytics.
Figure 4. Credit impulse and demand in the US, 1996Q1-2015Q2.
6
% change
% of GDP
2
-4
96 Q1
96 Q4
97 Q3
98 Q2
99 Q1
99 Q4
00 Q3
01 Q2
02 Q1
02 Q4
03 Q3
04 Q2
05 Q1
05 Q4
06 Q3
07 Q2
08 Q1
08 Q4
09 Q3
10 Q2
11 Q1
11 Q4
12 Q3
13 Q2
14 Q1
14 Q4
1
-2
-6
-9
-10
real private domestic demand
-14
credit impulse (r.h.s.)
-14
Source: Authors’ calculations, Haver Analytics.
Figure 5. Credit impulse and demand in the US, 1928-2014.
15
% of GDP
% change
10
8
5
3
-5
1928
1931
1934
1937
1940
1943
1946
1949
1952
1955
1958
1961
1964
1967
1970
1973
1976
1979
1982
1985
1988
1991
1994
1997
2000
2003
2006
2009
2012
0
-7
-10
-15
-2
real private domestic demand
Source: Authors’ calculations, Haver Analytics.
credit impulse (r.h.s.)
9
-12
Figure 5 shows these variables on an annual
basis for the US for the period of 1928 to 2014.
There, we can see a clear lead of the credit variable during the Great Depression of 1929-1934
and the similarity between the Great Depression and the more recent Great Recession.
ing of the population, have not added to the
downward pressure on interest rates on their
own.
We have tried our best to let the data speak.
Naturally, our results stand to be refuted by
studies using more comprehensive data samples or more efficient techniques for analysis.
But until refuted by a more powerful analysis,
our findings have three important consequences: First, long-term market interest rates are
strongly influenced by central banks’ perception
of reality rather than by the perceptions of market participants. Second, as central bank policy
makers are more likely than market participants
to lack the knowledge to push market rates to
levels consistent with economic fundamentals,
there is a high chance of misalignments of market rates.13 Third, misalignments of market rates
can cause severe economic distortions, and
their correction severe economic disruptions.
Conclusions
In this paper, we argued that the customary
neoclassical model of interest rate determination, in which long-term interest rates are determined by the supply of and demand for investable funds, is neither rooted in the institutional set-up of the credit markets nor supported by the data. Instead, we found the WicksellMises-Hayek model of the credit and business
cycle to be a much better description of reality.
From this, we conclude that central bank policy
has been largely responsible for the low level of
interest rates. Other variables, such as the age-
13
We assume here that central planning committees are
inferior decision makers than markets.
10
Technical Appendix: Data and estimation technique
The Dynamic Ordinary Least Squares (DOLS) estimation technique used in this paper allows us to
control for endogeneity of explanatory variables (Stock and Watson 1993; Wooldridge 2009).14 Endogeneity in the form of feedback effects or reverse causality between the dependent and independent variables would lead to a misspecification of our estimation model, in which we want to
identify the effects of short-term rates on long-term market rates. The DOLS procedure controls for
endogeneity of all explanatory variables by inserting leads and lags of the changes of all exogenous
variables. The model to estimate assumes the following form:
 =  + 1 1 + 2 2 + 3 3 + 4 4 + 
where
+
+
+
+
−
−
−
−
 = ∑ 1 ∆1− + ∑ 2 ∆2− + ∑ 3 ∆3− + ∑ 4 ∆4− + 
DOLS is a powerful estimation technique according to Saikkonen (1991) and Stock and Watson
(1993).15 Within this estimation framework, standard errors are corrected for heteroscedasticity and
cross-section correlation. It can be shown that by inserting the leads and lags of the exogenous variables in first differences, these variables become (super-) exogenous and the regression results unbiased (Wooldridge, 2009). The leads and lags enter the error term which can be decomposed into the
endogenous and exogenous changes of the right-hand side variables as shown above.
Application of the DOLS procedure requires the series to be non-stationary and in a long-run relationship, i.e. to be cointegrated over time. Only when cointegration is established can we be sure
that we do not estimate spurious relationships and that omitted variables (which are lumped together in the error term) do not systematically influence the long-run relationship between the endogenous and exogenous variables.
14
James H. Stock and Mark W. Watson. A simple estimator of cointegrating vectors in higher order of integrated systems.
Econometrica, 61(4), 783-820, July 2013, Jeffrey Wooldridge. Introductory econometrics: A modern approach. SouthWestern, Ohio, 2009.
15
Pentti Saikkonen. Asymptotically efficient estimation of cointegration regression. Economic Theory, 7(1), 1-21, March
1991.
11
Table A1 gives a description of the data used in our estimation.16
Table A1. Description of data.
USA
Germany
UK
Japan
gby10
10-year Treasury
bond yield at constant maturity, %
Govt securities with
residual maturities of
9-10 years, %
Govt Bonds, 10-year
nominal par yield, %
10-Year benchmark
govt bond yield, %
i3m
3-month London
Interbank Offered
Rate for US$ fund, %
3-month FIBOR:
Frankfurt Interbank
Offer Rate, %
3-Month London
Interbank Offered
Rate for British Pound
funds, %
Call rate for uncollateralized 3-month
money, %
deficit / debt
Gross Federal Debt as
a percent of GDP
Federal Govt budget
balance as a percent
of GDP
Central Govt budget
balance as percent of
GDP
Gross Federal Debt as
a percent of GDP
US foreign liabilities
US govt and other
long-term liabilities,
% of GDP*
debt x foreign liabilities
product of debt and
foreign liabilities (to
capture interaction)
(old age) dependency
ratio
dependency ratio in
% (pop 0-14 &
65+)/pop 15-64)
dependency ratio in
% (pop 0-14 &
65+)/(pop 15-64)
old age dependency
ratio in % (pop
65+)/(pop 15-64)
old age dependency
ratio in % (pop
65+)/(pop 15-64)
Note: * Own calculation based on Haver Analytics.
Source: Haver Analytics and Eurostat (for the German dependency rate).
Table A2 shows descriptive statistics of the variables included in the analysis.
Table A2. Descriptive statistics.
Mean
St. Dev.
Min.
Max.
USA
gby10
4.777
1.681
1.640
8.130
i3m
3.315
2.261
0.228
6.873
69.7
15.8
53.8
103.6
USA net foreign liability
0.014
0.015
0.001
0.140
US debt x USA net foreign liability
0.934
0.930
0.062
8.246
99.0
1.775
96.5
103.6
government debt
dependency ratio
16
Data used for the estimation are available from the authors on request.
12
Mean
St. Dev.
Min.
Max.
Germany
gby10
4.481
1.972
0.310
8.570
i3m
3.469
2.495
0.050
9.760
1.134
1.217
17
3.390
48.5
2.152
45.0
52.0
government deficit
dependency ratio
-8.000
UK
gby10
5.331
2.185
1.680
10.380
i3m
4.710
2.868
0.507
12.500
government deficit
2.023
2.351
17
6.020
old age dependency
24.6
0.726
23.9
27.0
-2.990
Japan
gby10
2.108
1.502
0.325
6.774
i3m
1.000
1.667
0.000
8.000
government debt
126.4
57.2
46.7
214.0
28.0
6.706
18.0
40.4
old age dependency
17
Budget surpluses in Germany and the UK reflect the receipts from the sale of G3 telephone licenses in 2000.
13
LEGAL NOTICE
The information contained and opinions expressed in this document reflect the views of the author at the time of publication and are subject to change without prior notice. Forward-looking statements reflect the judgement and future expectations of the author. The opinions and expectations found in this document may differ from estimations found in other
documents of Flossbach von Storch AG. The above information is provided for informational purposes only and without any
obligation, whether contractual or otherwise. This document does not constitute an offer to sell, purchase or subscribe to
securities or other assets. The information and estimates contained herein do not constitute investment advice or any
other form of recommendation. All information has been compiled with care. However, no guarantee is given as to the
accuracy and completeness of information and no liability is accepted. Past performance is not a reliable indicator of future performance. All authorial rights and other rights, titles and claims (including copyrights, brands, patents, intellectual
property rights and other rights) to, for and from all the information in this publication are subject, without restriction, to
the applicable provisions and property rights of the registered owners. You do not acquire any rights to the contents. Copyright for contents created and published by Flossbach von Storch AG remains solely with Flossbach von Storch AG. Such
content may not be reproduced or used in full or in part without the written approval of Flossbach von Storch AG.
Reprinting or making the content publicly available – in particular by including it in third-party websites – together with
reproduction on data storage devices of any kind requires the prior written consent of Flossbach von Storch AG.
© 2015 Flossbach von Storch. All rights reserved.
SITE INFORMATION
Publisher: Flossbach von Storch AG, Research Institute, Ottoplatz 1, 50679 Cologne, Germany; Phone +49 221 33 88-291,
[email protected], Directors: Dr. Bert Flossbach, Kurt von Storch, Dirk von Velsen; Registration: No. 30 768 in the Commercial and Companies Register held at Cologne District Court; VAT-No. DE200075205; Supervisory authority: German
Federal Financial Services Supervisory Authority, Marie-Curie-Straße 24 – 28, 60439 Frankfurt / Graurheindorfer Straße 108,
53117 Bonn, www.bafin.de; Authors: Agnieszka Gehringer, PhD and Dr. Thomas Mayer; Editorial deadline: 23. October 2015
14