Banks, however, it is arguably the simplest. We have examined the behaviour of a number
of alternative voting schemes though the differences in outcome schemes are negligible.
2
1 This threeway voting system is not, in practice, often implemented by the Central
Banks, however, it is arguably the simplest. We have examined the behaviour of a number
of alternative voting schemes though the differences in outcome schemes are negligible.
3
Student No.: 119047265
TESTING FOR PURCHASING POWER PARITY BETWEEN
DENMARK, SWEDEN AND UK
Course project prepared within the EC3064 module
Applied Econometric Project
University of Leicester, 2014
Date of Submission: 15/05/2014
EXECUTIVE SUMMARY
This project concentrates on explaining the relationship between nominal exchange rates
and price levels with particular focus on the relationships macroeconomic policy
implications. Using annual data from 19752013, longrun models were developed to test
how the price level ratio effects the bilateral nominal exchange rate. All data was taken
from the OECD database and the countries investigated were Denmark, Sweden and UK
with Denmark acting as the domestic country. These models failed to prove that
movements in the nominal exchange rate were solely down to changes in the price ratio.
Therefore it must be concluded that strong purchasing price parity doesnt hold. However,
they did indicate that the bilateral nominal exchange rates between Denmark and Sweden,
and Denmark and UK were, in part, determined by the differencing price levels between
the two countries. The tests indicated that some of the deviations in the nominal exchange
rates were down to fluctuations in the price ratio. However, the low r squared values
indicated that only a small percentage of nominal exchange movements can be explained
by changes in the price level. The results might be weakened as the same time period
might not have been long enough. Some of the tests used have also been criticized for
their lack of strength.
1. INTRODUCTION (OR PROBLEMS ADDRESSED)
Purchasing Power Parity (PPP) is an economic theory which claims that the nominal
exchange rate between two countries should be equal to the ratio of aggregate price levels
between those two countries, so that a unit of currency of one country will have the same
purchasing power in a foreign country. While some economists argue that PPP doesnt
hold in the shortrun most intrinsically believe that real exchange rates converge to the
PPP level. PPP is based on the law of one price (LOP) which is an economic assumption
that assumes that prices should be the same across the world when adjusted for their real
exchanges rates. If this was not the case rational agents would take advantage of the
resulting arbitrage opportunities to make risk free profit until prices equalised. Much of
economic theory is based on the assumption that agents act rationally so to reject the
theory of PPP is to dispute the assumption that underpins much of economic theory.
However, empirical studies have only found limited evidence in support of the theory.
This has led economists such as Taylor and Taylor (2004) to call it the PPP puzzle
1
. The
existence of PPP will not only have a huge impact on macroeconomic policy making but
will also provide evidence in support of wider economic theory as a whole.
In this paper the theory of purchasing price parity will be tested using three European
countries. These countries are Denmark, Sweden and UK with the Denmark Krona acting
as the domestic currency. Strong and weak PPP was tested between Denmark and the
Sweden and between Denmark and the UK.
PPP is often used to develop economic comparisons of market conditions in different
countries. It is an essential tool for converting nominal measures into real measures. For
example, PPP is often used to equalise estimates of gross domestic product. It is also
widely used in the formulation of real exchange rates and comparing the cost of living
between countries.
The understanding of how exchange rates adjust is central to exchange rate policy. PPP is
seen as an anchor
2
for long run exchange rates and hence provides an insight into the
1
Taylor, A.L and Taylor, M.P (2004). The Purchasing Power Parity Debate Journal of Economic Perspectives
18, P135158
2
Rogoff K, The Purchasing Power Parity Puzzle, Journal of Economic Literature, Volume 34, 1996, pages
647
2
potential under and overvaluation of a countrys currency. In theory PPP allows policy
makers to extrapolate the economic impacts caused by longterm fluctuations in exchange
rates. The importance attached to predicting the movements of exchange rates and price
levels has stimulated a considerable amount of literature and the significance of the
concept of PPP to macroeconomic policy making is the inspiration for this paper.
In this paper I will use cointegration analysis and test the stationarity of real exchange
rates in order to test for the long run existence of PPP between Denmark and Sweden and
Denmark and the UK. The hypotheses will be tested using the two step EnglGranger
cointegration test and the stationarity of the real exchange rates will be determined by
augmented DickeyFuller tests.
2. ECONOMIC THEORY AND PPP
Purchasing power parity (PPP) states that a unit of currency will have the same
purchasing power anywhere in the world. Ever since the term was coined in 1918 by
Cassel the theory has been the subject of much debate and empirical study. PPP is the
best known theory of the real exchange rate and understanding how exchange rates move
is an essential prerequisite to understanding exchange rate policy and ultimately
successful macroeconomic policy making. If PPP holds then exchange rates will adjust to
a level predetermined by PPP. This determines the extent to which the macroeconomic
system is selfequilibrating and as a result dictates the optimum level of macroeconomic
intervention in the financial markets
3
.
PPP claims that a unit of currency should be able to buy the same number of goods in one
country as it can in a different country. In other words there is parity in the purchasing
power of currency across economies. The theory of PPP is based on the law of one price.
This law states that a good must sell for the same price in all locations. If this was not the
case there would be arbitrage opportunities. The mechanism behind PPP is the idea that
differences in price of the same good in different countries will open up the prospect of
risk free profits to be made by buying the good in one country and selling it in another.
PPP states that all currencies must have the same real value in every country.
3
Rogoff K, The Purchasing Power Parity Puzzle, Journal of Economic Literature, Volume 34, 1996, pages
647
3
Although in its simplest form PPP is an attractive and intuitive theory there are some
complications when the theory is put into practice. The presence of transportation costs
for example can rule out a simple comparison of relative prices. While profitable
arbitrage situations might still exist, purchasing parity will no longer exist as traders will
have additional costs associated with transporting goods. Many countries also impose
trade barriers and tariffs which violate the law of one price. The border effect, first
observed by Engel and Rogers (2006), found that cities that were further away had
greater differences in prices.
4
This price difference increased dramatically across
international borders. On top of this it is often hard to compare the price level across
economies. Different countries tend to produce differentiated goods which are not
perfectly substitutable. Moreover, goods in aggregate price indices will have different
weights. For example if the USA drinks more coffee than the French, then coffee will
have a heavier weight in the USA price index than in France. This makes it difficult to
compare price indices internationally. On top of this, many goods and services are not
traded between all counties.
Despite these objections it is often assumed that PPP will hold to a degree due to the
possibility of international arbitrage. There are two ways in which the PPP theory might
hold. Absolute PPP holds when the purchasing power of one unit of currency is equal in
the domestic country and in a foreign country. Relative PPP implies that the changes in
the exchange rate offset the changes in the price level of the countries concerned. If
absolute PPP holds then relative PPP should also hold. However, it is possible for relative
PPP to hold while absolute PPP is void, since common changes in nominal exchange
rates are happening at different levels of purchasing power for the two currencies. Due to
goods in different countries often being differentiated it is hard to determine whether an
identical basket of goods is available in each country. Therefore it is more common to
test for relative PPP.
PPP has been the subject of much empirical investigation. According to Sideris (2005)
many empirical studies on PPP have found little or no evidence of the theory. Most of the
early studies on PPP were based on the timeseries analysis of short spans of data of 10
years or less Nusair (2003). Almost all of this analysis found no evidence in favour of
4
Charles Engel & John H. Rogers, 2006"The U.S. Current Account Deficit and the Expected Share of World
Output,"NBER Working Papers 11921, National Bureau of Economic Research, Inc.
4
PPP. One possible reason for this is that most studies are unable to use data that spans a
great enough time period. Before 1971 most currencies were tied to the price of gold
through the BrettonWoods agreement and the major currencies have only been floating
against each other in since 1973. Frankel (1986) claimed that the reason many studies
failed to find evidence in favour of PPP was due to the lack of power. Even if arbitrage
corrects unbalanced exchange rates the adjustment may not be immediate. Information
about arbitrage opportunities might take time to become available. It is possible that PPP
may not hold in the short run but that it may hold in the long run. There can be short run
deviations from PPP even if there is a long run tendency for PPP to hold. Rogoff
summarized this paradox neatly with the quote While few empirically literate
economists take PPP seriously as a shortterm proposition, most instinctively believe in
some variant of PPP as an anchor for real exchange rates.
5
Depending on the sample
size, and type of tests employed, some evidence in the favour of long run PPP has been
found. Yet, the speed of convergence to PPP is extremely slow Nusair (2003)
6
.
The failure of PPP in the short run led to a series of formal tests for evidence of PPP as a
longrun phenomenon. These studies claimed that PPP is retained in the longrun but
allowed for shortrun deviations from the PPP price level. If the real exchange rate
displays reversion to its mean then it will settle at a certain value. The only necessary
condition for relative PPP to hold in the long run is the presence of mean reversion. For
absolute PPP to hold the real exchange rate must revert to the PPP real exchange rate
level. Adler and Lehmann (1983) investigated whether real exchange rates converged on
a particular level or whether they followed a random walk by checking the stationarity of
the real exchange rates.
7
However, this early study suffered from low power and was
unable to reject the null hypothesis of a random walk.
In the late 1980s more sophisticated econometric methods were employed in the
investigation of longrun PPP. These new empirical studies were based on the concept of
unit roots. If a unit root is present in a time series function then shocks persist forever and
the function is described as nonstationary. If there are no unit roots present,the time
5
Rogoff K, The Purchasing Power Parity Puzzle, Journal of Economic Literature, Volume 34, 1996, pages
647
7 Nusair, S.A 2003. "Testing The Validity Of Purchasing Power Parity For Asian Countries During The Current
Float" Journal of Economic Developmentvol. 28(2), pages 129147
7
Deviations from Purchasing Power Parity in the Long Run Adler M and Lehmann Bruce
5
series will eventually converge on a level. In other words relative PPP holds when there
are no unit roots present. However, empirical studies applying these tests on real
exchange rates unanimously failed to reject the null hypothesis of a unit root. This
resulted in a failure to find evidence of longrun PPP.
On the one hand Frankel (1990) pointed out that not being able to find sufficient evidence
required to reject the null hypothesis does not mean that we have to accept the null
hypothesis as true.
8
Frankel explained that the statistical tests used to examine the long
run stationarity of the real exchange rate may not have enough power to reject the null
hypothesis. This would have been especially true at the time when empirical studies were
based on data covering little more than 15 years. The argument was that even if the real
exchange rate did tend to revert to its mean over long periods of time, examination of the
real exchange rate over a relatively short time period may not be enough to detect this
mean reversion (Taylor and Taylor 2004). Due to sticky prices the real exchange rates
could take a long time to revert to its mean and this makes it almost impossible to reject
the unit root hypothesis without utilising a large dataset.
One the other hand Adler and Lehmann (1983) observed fixed and floating exchange
rates and were unable to reject the null hypothesis of a random walk.
9
This led them to
question cointegration and claim that real exchange rates followed a random walk. This
rejection of cointegration led many to question PPP with Dornbusch stating that even if
PPP held in the long run the mean reversion time was so long that the theory was of little
or no use empirically.
One way to glean more information about the longrun behaviour of real exchange rates
is to increase the number of years of data. However, as we have already discussed this
was not a feasible option due to the Bretton Woods agreement. Another approach which
was used was to use more countries. The more information used in the tests should
increase their power.Taylor and Sarno (1998) suggested testing the hypothesis that that at
8
Frankel, Jeffrey A., 1990. "And Now won/Dollar Negotiations? Lessons from the Yen/Dollar Agreement of
1984,"Department of Economics, Working Paper Series qt851615wb
6
least one real exchange rate was not mean reverting.
10
If this hypothesis could be rejected
then it could be claimed that all of the tested real exchange rates were mean reverting.
However, once again this test suffered from a lack of power and no conclusive
conclusions could be formed.
As time passed more advanced econometric techniques were used to combat the problem
of low statistical power. Unit root tests with improved power, and the capability to
account for possible structural breaks were used by Parkes and Savvides (1999).
11
Multivariate cointegration techniques such as the Johanson Technique were also used to
analyse possible cointegrating relationships between multiple variables. These modern
econometric techniques delivered mixed results with some studies finding some evidence
in support of longrun PPP.
Since the 1970s, the theory of PPP has undergone a large amount of empirical study and
theoretical debate. For a long time, theoretical work suggested that exchange rates should
be linked to the price level with shortrun deviations while empirical analysis found little
in the way of evidence to back up the theory. However, the gap between theoretical and
empiric understanding is getting smaller due to the increased power of econometric tests
and the ever increasing time frame with which econometricians can operate.
3. DESCRIPTION OF DATA EMPLOYED
Yearly data was employed from 19752013 for three European Countries with no missing
observations. The observed countries are Denmark, Sweden and the United Kingdom. The
data was obtained from the Organisation for Economic Cooperation and Development
(OECD.) The OECD is internationally known for the accuracy and availability of its data. All
results are measured as natural logarithms. Since the PPP hypothesis concerns a relationship
between the nominal exchange rate and the ratio of domestic to foreign price levels in order to
test the theory we need the nominal exchange rates and the price levels of the observed
countries.
10
Taylor, Mark P. & Sarno, Lucio, 1998. "The behavior of real exchange rates during the postBretton Woods
period,"Journal of International Economics, Elsevier, vol. 46(2), pages 281312
11
Parkes, A & Savvides, A 1999. "Purchasing power parity in the long run and structural breaks: evidence from
real sterling exchange rates,"Applied Financial Economics, Taylor & Francis Journals, vol. 9(2), pages 117127
7
The time period of 19752013 was chosen due to the interesting changes in exchange rate
policy during this period. After World War II the U.S signed the Bretton Woods agreement
which tied the U.S dollar to the price of gold. At this time almost all of the Worlds major
currencies were pegged to the U.S dollar. Currencies were not free to float against each other
until the 1970s after the agreement had been dissolved. The time period is as large as
possible to give our tests as much power as possible without mixing fixed exchange rate
systems with floating exchange rate systems. Although the data set has been made as large as
possible, exchange rates seem to take a long time to converge on the PPP determined level
and our 38 year analysis might not be long enough to find evidence of long run PPP. Yearly
data was chosen for simplicity and because increasing the frequency of the observations wont
increase the power for longrun testing because increasing the amount of detail regarding
shortrun movements will not give any extra information on longrun movements.
A key concern was the choice of the price index with the CPI (Consumer Price Index) and the
PPI (producer price index) being the most popular indexes. PPI places a heavier weight on
tradable goods than the CPI and tend to yield more favourable test results to PPP Cheung and
Lai (1993). However, it could be argued that the relationship between traded goods, prices
and exchange rates comes close to a truism. The CPI was chosen, firstly due to its availability
and consistency and secondly because the CPI was used in the actual formulation of the PPP
hypothesis by Casel. The CPI figures used in this study were obtained from the Consumer
Prices (MEI) dataset made available by the OECD. The dataset is a subset of the Main
Economic Indicators (MEI) database. The database consists mainly of monthly data for the 34
OECD member countries. The MEI database contains a wide variety of Shortterm Economic
Statistics. The CPI data series measures changes in the price level of a market basket of goods
and services purchased by consumers. When comparing the CPI it is very hard to determine
whether the goods and services available in each country is exactly the same. However, using
some sort of price index is essential to testing absolute PPP and according to the OECD a lot
of effort has been made to ensure that the data are internationally comparable across all
countries.
12
It could be argued that the nominal exchange rate is the unadjusted weighted
average value of a country's currency relative to all major currencies being traded within an
index or pool of currencies. The exchange rates used in this analysis are a geometric average
12
http://stats.oecd.org/Index.aspx?DataSetCode=MEI_PRICES
8
of the rates against the US dollar. The nominal exchange rate is defined as the amount of
domestic currency required to purchase one unit of foreign currency.
Not all of the nominal bilateral exchange rates were available and some of the exchange rates
had to be created. The nominal exchange rates between Denmark and Sweden and Sweden
and the UK were calculated using the following equation:
(1)
I am aware that some slight inaccuracies might be present in this methodology but this was
the only available method for the less popular currency pairs.
Additionally, the logarithmic expressions of all series were calculated. Formally, the absolute
version of PPP is defined as:
(2)
Where the summed geometric weights sum to unity. If the price indices are constructed using
a geometric index then we must form the weighted sum after taking logs:
(3)
Again the sum of the geometric weights add up to 1. If the national price levels and the
nominal exchange rates are logarithmically transformed then the popular absolute PPP
condition that is used in the literature takes the form.
13
(4)
Equation 4 is less complicated and more intuitive than equation 2 making it more convenient
to use in terms of estimation. Therefore all variables were estimated in log form.
The exchange rates used in this study are all European exchange rates. In 1979 all of the
countries observed in this study adopted the European Exchange Rate Mechanism. The aim of
this mechanism was to reduce the exchange rate variability and achieve monetary stability in
13
Prof Peter J. Hammond (2008) LongRun PPP: The Case of Greece: Towards the Transition to the Euro: A Maximum
Likelihood Cointegration Approach
9
Europe ahead of the introduction of the Euro. One of the central problems with empirical
studies on PPP is the lack of power arising from the inability to use data from fixed exchange
rate systems. However, under the European exchange rate system currencies were still
allowed to float freely against each other. The European exchange rate mechanism was a
semipegged system based on the concept of fixed exchange rate margins, but with the
exchange rates variable within those margins. So while sudden, substantial fluctuations in the
values of the currencies were restricted, the currencies under investigation were still allowed
to depreciate and appreciate freely. This mechanism was designed to prevent the large
fluctuations in currency values associated with the euro. This should not prevent the exchange
rate from returning to the PPP value. Whilst these currencies might not be perfectly free,
exchange rate policy has been an important part of macroeconomic policy for the last 40 years
and it is almost impossible to find a group of currencies that is completely free of government
intervention.
Static and dynamic properties of the data:
Consumer Price Indexes:
Figure 1 shows the price levels for a basket of
goods in Denmark, Sweden and the UK. From
the graph it is possible to see that the base year
was in 2010. The graph shows that all of the
consumer prices are trending upwards over time.
This could be due to the expansionary monetary
policy that was popular during this time period.
This trend could also have resulted from a rise in
real wages caused by technological advances.
However, the graph shows that the rate at which
the CPI is increasing is falling. This will be due to increased emphasis on economic stability
that central banks have focused on for the last 20 years.
Figure 1 clearly shows that all three countries have similar price level trends. All of our
countries have similar price levels which follow a similar pattern, this could be because all
3
3
.
5
4
4
.
5
5
1970 1980 1990 2000 2010
Time
CPI Denmark CPI Sweden
CPI UK
Price Levels for Denmark, Sweden and the UK
Figure 1
Figure 1
10
three of the sample countries have similar economic structures. They are all developed
countries with similar access to technology.
There appears to be a slight structural break in Swedens CPI between the dates of 20082009.
Swedens CPI doesnt increase by as much as it has done in recent years. This is confirmed by
Figure 2 which shows a sharp reduction in Swedens inflation rate in 2008. Figure 2 shows a
similar but smaller dip in the inflation rates for the UK and Denmark for this time period.
The upward trend present in the time series shows that the CPI rates are not stationary. They
are trending upwards over time and do not appear to revert to a particular mean. It is very
unlikely that they have a constant mean.
To confirm the nonstationarity of the CPIs, autocorrelations graphs of the price indexes were
used. Autocorrelation graphs are a tool commonly used to check the randomness in a data set.
If the time series is stationarity, the lags will oscillate around 0 for all timelag separations. If
the variable is nonstationary there will be one or more autocorrelations which will be
significantly nonzero. The autocorrelation graphs in Appendix Figure 1 and 2 show that there
are lags significantly different from 0 for all three CPI values confirming that the time series
are probably not stationary. The autocorrelation graphs dont appear to be stationary until the
second difference this is shown in Appendix Figure 2. The second differenced time series
appear to be much more stationary. The autocorrelation graphs of these variables quickly fall
to zero implying stationarity. The graphical analysis of the CPI rates implies that they are all
integrated to I(2).
Yearly Inflation Rates:
Figure 2 shows the inflation rates for
Denmark, Sweden and the UK. All three
countries have very similar average
inflation rates as previously suggested by
Figure 1.
The inflation rate is the annual change in
0
.
0
0
0
0
0
.
0
5
0
0
0
.
1
0
0
0
0
.
1
5
0
0
I
n
f
l
a
t
i
o
n
R
a
t
e
(
%
)
1970 1980 1990 2000 2010
Time
Denmark Inflation Rate Sweden Inflation Rate
United Kingdom Inflation rate
Yearly Rates of Inflation for Denmark, Sweden and the UK
Figure 2
11
the CPI. Like Figure 1, Figure 2 shows that the price levels in all three countries are
increasing with time. Sweden is the only country that has experienced deflation in the last 40
years. One of these times was in 2008 during the great recession. All three countries
experienced lower inflation in this year probably due to the reduction in demand caused by
lower demand caused by the depression. However, the inflation rates of all three sample
countries spiked after the initial 2008 dip pushing inflation to the highest level since the mid
1990s. This will be due to the loose monetary policy adopted by central banks in an attempt
to boost demand and offset the negative effects of the recession. An increase in fuel and food
prices also pushed inflation higher.
Exchange Rates:
Figure 3
Figure 3 shows the bilateral nominal exchange rates where foreign currency is expressed in
units of domestic currency. The CPI ratio between the home and domestic countries is also
plotted on the same graphs. The motive behind this is that if relative PPP is to hold between
the two countries then there should be a parallel movement in the nominal exchange rate and
the price ratio between the two countries. As each exchange rate has been defined as the
domestic cost of buying a unit of foreign currency, an upward movement in the graph depicts
a depreciation in the nominal exchange rate i.e. the cost of obtaining a unit of foreign
currency has increased. Similarly a downward movement represents an appreciating in the
nominal exchange rate. Over the sample period the Denmark Kroner has generally
depreciated against the Swedish Kroner and the British pound. However, the price level ratio
has stayed relatively constant over this time period suggesting that strong PPP may not hold.
The price ratio does not seem to follow the nominal exchange rate very well indicating that
.
6
.
8
1
1
.
2
1
.
4
1970 1980 1990 2000 2010
Time
NER DenmarkSweden DenSwe priceratio
Denmark:Sweden
0
.
2
.
4
.
6
.
8
1
1970 1980 1990 2000 2010
Time
NERDenmarkSweden DenUK priceratio
Denmark:UK
12
PPP doesnt hold. In both cases there is no tendency for the series to revert to a constant
mean. The visual impression suggests that the nominal exchange rates are consistent with
random walk and hence an I(1) process.
This impression is backed up by Appendix Figures 3 and 4 which show that the bilateral
exchange rates dont appear to be stationary until they are first differenced.
Nominal Exchange Rates:
Figure 4
Figure 4 plots the nominal exchange rates of the three economies in dollar terms. As shown
the UK nominal exchange rate is very stable and doesnt appreciate or depreciate much at all.
The Swedish Krona and the Danish Krona depreciated against the dollar in the mid to late
1970s as the currencies were allowed to float freely against the dollar. The graph shows that
all three of the currencies experienced fluctuations in their exchange rates during the
European Exchange rate mechanism (EERM). The Swedish Krona suddenly devalued in
1993 when the Swedish Government announced that it would be leaving the European
monetary union. Informal graphical analysis seems to suggest that the Danish Krona and the
Swedish Krona fluctuate together despite these currencies not being pegged to each other. In
all three cases there appears to be no tendency for the series to revert to a constant mean. The
visual impression given off by the graphs is that, in each case, the nominal exchange rate is
consistent with random walk and hence appears to be a nonstationary I(1) process.
Appendix Figure 5 shows the autocorrelation graphs of the nominal exchange rates at levels.
The three nominal exchange rates do show a tendency to die out, but only at very long lags.

1
0
1
2
3
1970 1980 1990 2000 2010
Time
ERDenmark ERSweden
ERUK
Nominal Exchange Rates
13
Appendix Figure 6 shows the autocorrelation graph of the first difference of the bilateral
exchange rate. In general these approach the zero axis quickly. These autocorrelation graphs
indicate that the first difference of the nominal exchange rates do not have any significant lags
and therefore appear to be stationary.
Stationarity of the Data:
A timeseries variable that has a mean and variance that dont change over time are described
as stationary. A weakly stationary timer series satisfies the following criteria:
14
1. (
) (constant mean)
2. (
(constant variance)
3. (
(covariance stationarity)
Augmented DickeyFuller test:
The unit root test is the formal way to check the stationarity of a time series variable. In this
paper the augmented DickeyFuller test was used to check the variables for unit roots. Dickey
and Fuller (1979) developed a procedure to formally test for nonstationarity based on a
stimple AR(1) model. The test used in this paper took the following form:
(5)
The augmented DickeyFuller test allows for the addition of lagged lengths. The number of
lagged difference terms that were included was chosen so that the error term is serially
uncorrelated. The number of lags selected were based on the number of significant lags in the
AC graph.
The null and alternative hypothesise of the augmented DickeyFuller test can be written as:
)(6)
Here again the null hypothesis is the presence of a unit root and the alternative hypothesis is
stationaity.
Unit Root Test:
Augmented DickeyFuller test
results
PhillipsPerron test results
Variable
Name
tvalue 5%
Critical
Level
Hypothesis
result
tvalue 5%
Critical
Level
Hypothesis
result
lCPIDenmark
(3)
2.374 2.972 Fail to
Reject
2.635 12.916 Fail to
Reject
lCPISweden
(3)
3.110 2.969 Reject 2.959 12.916 Fail to
Reject
lCPIUK(3) 2.268 2.972 Fail to
Reject
2.635 12.916 Fail to
Reject
ERDenSwe 1.206 2.972 Fail to 2.417 12.916 Fail to
15
The Purchasing Power Parity: Evidence from the Great Financial Crisis
15
Reject Reject
ERDenUK 1.411 2.969 Fail to
Reject
6.839 12.916 Fail to
Reject
Table 1
Table 1 contains the results of the unit root tests on the CPI and nominal exchange rates at
levels. As expected the null hypothesis of a unit root was not rejected meaning that the
variables are nonstationary. These unit root results were backed up by the autocorrelation
graphs, analysed in the data description section, which show several lags that are significantly
different from zero. However, the augmented DickeyFuller test did manage to reject the null
hypothesis for the consumer price index of Sweden concluding that this series was stationary.
The PhillipsPerron test on the other hand failed to reject the presence of a unit root. Arguably
the PhillipsPerron test has more power than the augmented DickeyFuller test. The non
stationarity result is also backed up by the graphical analysis of the trend and the
autocorrelation graphs. Therefore it is possible to assume that all variables are nonstationary
at levels.
Augmented DickeyFuller test
results
PhillipsPerron test results
Variable Name tvalue 5%
Critical
Level
Hypothesis
result
tvalue 5%
Critical
Level
Hypothesis
result
dlCPIDenmark
(2)
1.760 2.972 Fail to
Reject
2.487 12.884 Fail to
Reject
dlCPISweden
(3)
1.138 2.975 Fail to
Reject
2.639 12.916 Fail to
Reject
dlCPIUK(3) 3.400 2.975 Reject 6.523 12.884 Fail to
Reject
dERDenSwe(0) 5.941 2.966 Reject 32.079 12.884 Reject
dERDenUK(0) 4.900 2.966 Reject 32.079 12.916 Reject
Table 2
Table 2 shows the results of the unit root test of the first differenced values of the CPI and
bilateral nominal exchange rates. The unit root tests performed on the two bilarteral nominal
exchange rates found enough evidence to reject the null hypothesis of a unit root. The first
16
difference of the bilateral nominal exchange rates are stationary and integrated to order I(1).
However, the unit root tests failed to reject the null hypothesis for the first deference of the
CPI series. The only test that managed to reject the presence of a unit root was the ADF test
on the UKs CPI. The PhillipsPerron test and the AC graphical analysis.
4. EMPIRICAL MODEL(S) AND ITS (THEIR) ESTIMATION
Methodology:
This econometrics project aims to test weak and strong relative PPP in three countries.
All results were obtained and graphed using the Stata program. Relative PPP holds when
the ratio between the nominal exchange rate and the domestic price levels (this doesnt
make sense or is incomplete). If
is the
overseas price level then PPP can be expressed by this equation:
(7)
Taking logs of both sides gives us:
(8)
The left hand side of this equation is the logarithm of the real exchange rate. This implies
that if PPP holds then the real exchange rate must be constant. As we have already
discussed it is likely that the real exchange rate does fluctuate over the short term. To
allow for these random deviations a stochastic shock
(9)
Strong PPP:
The empirical analysis starts with the strictest version of PPP which assumes that
and
is stationary.
The real exchange rates are a function of nominal exchange rates and price indexes.
) (10)
Where
(11)
Equation 12 was used to graph the real exchange rate for the three economies. The graph
of a stationary time series is not supposed to show a time trend so informal graphical
analysis was used to observe the real exchange rates stationarity.
To formally test the stationarity of the real exchange rate the augmented DickeyFuller
and the PhillipsPerron tests was used. The tests took the same form as in equations 6 and
7. Where we test the null hypothesis
and
and
. All restraints in
equation (10) are relaxed. In this weak case the nominal exchange rate and domestic and
foreign price levels are related in the longrun but with the assumption that a 1% increase
in domestic and foreign prices does not completely offset the nominal exchange rate. The
requirement for weak form PPP is an existence of any linear relationship among the
variable that has stationary properties.
(12)
18
For the weakest form of PPP we are looking for a linear relationship among the observed
variables which means that cointegration analysis should be used.
To investigate if weak longrun PPP holds, the relationship between the nominal
exchange rate and the price indexes needs to be examined. To examine the link between
the variables, ordinary least squares regression analysis was used. The main motivation
for using cointegration analysis is to identify meaningful longrun relationships among
nonstationary variables. As previously discussed the PPP relationship is likely to be a
long run one, it is also likely that the most of our variables are nonstationary. This is
because although our variables measure different aspects of the economy, they are closely
related.
The cointegration test employed in this paper is the EngleGranger test which follows two
steps. The first test is to conduct a unit root test in order to ensure that all variables are
integrated to the same order. Nonstationary variables that become stationary after being
n times become integrated to the nth order. If all the variables are integrated to the same
order, the next step is to estimate the model and test whether the residuals of the model
are stationary.
PPP holds in the long run if the sum of the nominal exchange rate and the foreign
countrys CPI are cointegrated with the CPI of the domestic country. Thus the PPP
relation for the EngleGranger test is:
(
) () (
) (13)
To allow for short run deviations from the PPP exchange rate value an error term is
added.
(
()
()
The resulting residuals from equation 8 are then tested for a unit root. If these residuals
are found to be stationary it can be concluded that weak PPP holds.
19
However, before we apply this test we must make sure that all of the observed variables
are stationary. If all of the observed variables were not stationary there would be a risk of
spurious regression. We would be unable to tell whether our conclusions were based on
economic relationships or simply the randomness in our data. Thus augmented Dickey
Fuller tests were performed for individual CPIs and nominal exchange rates following the
same method as described in the previous section. If we are unable to reject this null
hypothesis at a 5% confidence interval that variable is nonstationary and could therefore
suffer from spurious regression.
When we have nonstationary variables we can try to difference them until they become
stationary and then using the differenced variables in our analysis. However, differencing
to achieve stationary requires the mean to be stochastic. Differenced variables are also
often not statistically significant and have no economic interpretation.
The error correction model:
As noted earlier, there are nonstationary variables in our regression model which could
result in our estimates being spurious. So if we regress equation 15 the estimated
parameters might not be accurate.
One way of resolving this problem is to difference the variables until they are stationary
and then run the regression. In this case the regression will produce correct estimates of
the parameters and the problem of spurious regression is resolved. However, this
regression only measures the shortrun relationship between the variables. As PPP is
generally seen as a longrun phenomenon this would be an inappropriate regression.
The error correction model is a convenient model that measures the correction from
disequilibrium of the previous period. If
and
(15)
Results:
20
Real exchange rate Stationarity:
As discussed one way to test for strong PPP is to test the stationarity of the real exchange
rates. This paper uses graphical analysis and employs the augmented DickeyFuller and
PhillipsPerron tests for unit roots in order to test the stationarity of the real exchange
rates.
Graphical Analysis:
The first step in assessing the empirical evidence is to graph the data. Figure 5 plots the
real exchange rates and the bilateral nominal exchange rates between the three sample
countries.
Figure 5 and 6
Figures 5 and 6 plot the real and the nominal bilateral exchange rates for Denmark,
Sweden and the UK. The real exchange rate for the Denmark and Sweden does not look
like it is stationary. There does not appear to be a constant mean and the time series
appears to be integrated to order I(1). The real exchange rate for Denmark and Sweden
appears to be more stationary with values appearing to oscillate around a constant mean.
This implies that strong PPP might hold between Denmark and the UK. This hypothesis
is backed up by the AC graphs in Appendix Figure 7 and 8 which shows significant lags.
The first differenced AC graphs, in Appendix Figure 8 and 9, show that the real exchange
rates are stationary when they are first differenced.
Unit root testing:
To formally test for stationarity in the real exchange rates unit root tests were performed
on the two variables.

.
4

.
2
0
.
2
.
4
1970 1980 1990 2000 2010
Time
REDenSw ldenmarksweden
Nominal and Real ER Between Denmark and Sweden

2
.
8

2
.
6

2
.
4

2
.
2

2
1970 1980 1990 2000 2010
Time
REDenuk ldenmarkuk
Nominal and Real ER between Denmark and UK
21
Augmented DickeyFuller test
results
PhillipsPerron test results
Variable Name t
value
5%
Critical
Hypothesis
result
t
value
5%
Critical
Hypothesis
result
REldenswe(3) 1.335 2.972 Fail to
Reject
2.543 12.916 Fail to
Reject
REldenuk(2) 0.907 2.969 Fail to
Reject
5.796 12.916 Fail to
Reject
Table 3
Table 3 contains the results of applying the augmented DickeyFuller and PhillipsPerron
unit root tests to the real exchange rate series at levels. All of the tests have failed to find
evidence of stationarity in the real exchange rates. Therefore it must be assumed that the
variables are nonstationary and that strong PPP doesnt hold between the sample
countries.
Augmented DickeyFuller test
results
PhillipsPerron test results
Variable Name t
value
5%
Critical
Hypothesis
result
t
value
5%
Critical
Hypothesis
result
REldenswe(0) 5.461 3.668 Reject 34.229 12.916 Reject
REldenuk(0) 5.057 3.668 Reject 27.911 12.916 Reject
Table 4
Table 4 confirms the AC graphs in Appendix 7 and 8. The unit root test were able to
reject the null hypothesis concluding that the real exchange rates were stationary when
difference. It is therefore possible to conclude that the real exchange rates are integrated
to I(1). This implies that the real exchange rates do not converge on a mean value and
therefore there is no evidence to suggest that strong PPP exists between any of the
countries.
Cointegration Analysis:
Denmark and Sweden:
22
To assess whether there is a purchasing power parity relationship between Denmark and
Sweden equation 15 is estimated by ordinary least squares.
Figure 7
Using the results expressed in Figure 13 our model becomes:
(16)
The model shown in equation (16) shows that as the price level in Sweden increases the
nominal exchange rate between Denmark and Sweden appreciates. It also shows that if
there is an increase in the price level in Denmark the nominal exchange rate between
Denmark and Sweden will depreciate.
(17)
The negative coefficient on the lagged residuals supports the theory of purchasing price
parity. The coefficient suggests that if the nominal exchange rate differs from the
predetermined PPP level by 1% in one period it will move 0.39% closer to the PPP level
in the next term. This shows that the nominal exchange rate is mean reverting and
converging on a particular level. This supports the theory of weak longrun PPP.
However, the error correction model is not without faults. The Rsquared value only 0.26
which is shows that the model does not explain much of the nominal exchange rate
deviations. The
and
(17)
In this new error correction model the R squared value has not fallen much at all
suggesting that the new model is still relevant. All the variables are significant and the
coefficient of the residuals is still negative implying a weak PPP relationship.
The regression and the error correction model provide weak evidence to suggest that a
weak PPP relationship exists between Denmark and Sweden despite the model not fitting
very well.
25
Denmark and UK:
To assess whether there is a purchasing power parity relationship between Denmark and
UK equation 14 is estimated by ordinary least squares. A problem with running a linear
regression between Denmark and UK is the variables are integrated to a different order.
The CPI variable for Denmark and UK are integrated to I(2) whereas the bilateral
nominal exchange rate between the two countries is integrated to I(1). This could result in
spurious regression.
Figure 12
Figure 13 and 14
(18)
The model shown in equation (18) shows that as the price level in Sweden increases the
nominal exchange rate between Denmark and UK appreciates. It also shows that if there
is an increase in the price level in Denmark the nominal exchange rate between Denmark
and Sweden will depreciate.
(19)
The negative coefficient on the lagged residuals supports the theory of purchasing price
parity. This again suggests that weak longrun PPP applies to Denmark and UK.
However, once again the Rsquared values are very low meaning that our model doesnt
fit very well.
The regression and the error correction model provide weak evidence to suggest that a
weak PPP relationship exists between Denmark and UK.
Policy Conclusion:
The theory of PPP has many uses in macroeconomic theory. Weak PPP can explain why
the exchange rate moves in certain ways. PPP can indicate the desirable exchange rate,
when the actual nominal exchange rate differs from the PPP determined level. The central
bank may try to narrow the gap between the actual nominal exchange rate and the
preferable nominal exchange rate by expanding or contracting monetary policy. Weak
PPP between Denmark and Sweden and Denmark and the UK allows the government of
those countries to set an appropriate target level nominal exchange rate which the
Government can use to stabilise their exchange rates.
Finding weak PPP between the sample countries indicates that those countries have a
similar level of competitiveness. The real exchange rate is a frequently used measure of
international competitiveness. When the real exchange rate rises, the home country gains
competitiveness because domestic prices are lower. The tests employed in this paper
found evidence of weak PPP. This indicates that domestic prices are relatively similar
between the Denmark, Sweden and UK. However, when the strong PPP relationship was
analysed using real exchange rates it was found that Denmark appeared to loose
competitiveness when compared to Sweden and the UK. The lack of strong PPP could
imply that Denmarks Government needs to concentrate on policies designed to boost
international competitiveness. Denmark could for example devalue its nominal exchange
rate or reduce the amount of red tape to boost the competitiveness of its domestic
companies.
28
Whether or not PPP holds can provide an insight into the inflation level in a country.
Imported inflation is inflation caused by an increase in the cost of imported goods.
16
If
the nominal exchange rate is below the PPP optimum level inflation is likely to increase.
If a currency is overvalued when compared to the PPP level imports will be relatively
more expensive. The Government can therefore use expansionary and monetary fiscal
policy to bring the nominal exchange rate back to its PPP optimum level without
increasing inflation.
PPP can also be used to develop comparative statistics. PPP can be used to compare
prices and wages internationally. These comparisons can then be used to analyse the
standard of living in each country.
All three of the sample countries are members of the European Union but are not part of
the euro currency area. Finding a PPP relationship between the three sample countries
insinuates that there are few barriers to trade between Denmark, Sweden and the UK.
Agents are able to correct price differentials through arbitrage. One of the main reasons
for countries giving up their monetary freedom by adopting the Euro was to eradicate
currency uncertainty and by doing so increasing trade within the European Union.
However, the three countries studied have a high degree of free trade without the need to
be members of the monetary union.
5. CONCLUSIONS:
The purpose of this paper is to determine the validity of the PPP hypothesis for three
European countries. The two approaches used for the empirical analysis was the unit root test
on the stationarity of the real exchange rate and the cointegration test between price indexes
and nominal exchange rates. The empirical analysis has found some evidence of weak PPP for
Denmark and Sweden and for Denmark and the UK. However, no strong PPP was found as
the ADF tests on the real exchange rates were unable to reject the null hypothesis of a unit
root. These results are realistic. Weak PPP is likely to exist between the 3 sample countries
due to their highly correlated economies and close proximity. Strong PPP is less likely to exist
due to transportation costs and lack of information. The real exchange rate is also known to be
16
http://www.investorwords.com/15442/imported_inflation.html
29
The EngleGranger two step test was used to assess the cointegration between price indexes
and nominal exchange rates. This method allows us to test for the existence of a more general
cointegrating relationship between the nominal exchange rate and price levels. However, there
might be more than one cointegrating relationship between three variables. The Engle
Granger does not allow for more than one cointegrating relationship and is often critised for
over rejecting PPP. To account for multiple cointegrating vectors the more powerful Johanson
test could have been used.
The test indicates that the nominal exchange rate between Denmark and Sweden reverts to its
mean quicker than the nominal exchange rate between Denmark and UK. This is shown by
the value of the coefficient on the lagged residuals in the error correction model. This is to be
expected due to the closer proximity of Denmark and Sweden. This reduces the transport
costs and increases the speed of information.
The data used in this paper is complete with no missing observations. It comes from the
OECD a reliable source. However, as mentioned in description of data section the bilateral
exchange rate series were manually constructed. This may have created minor errors,
although in theory the way in which exchange rates were calculated should be the same.
Finally, another problem encountered was the different order of integration of the variables.
The bilateral exchange rates were both intergrated to I(2) whereas all three price indexes were
only integrated to I(1). For the EngleGranger test to different orders of integration multi
cointegration had to be used. In this paper the trivarite system was employed were the
cointegrated relationship between the bilateral exchange rate and the price levels of the
domestic and foreign countries. In general it is possible for corresponding combination of two
I(2) variables to be cointegrated with a I(1) variable giving rise to a I(0) linear combination
among the three variables.
REFERENCE
30
1 Adler M and Lehmann B.F, Jeffrey A. Deviations from Purchasing Power Parity in the
Long Run, 1990. "And Now won/Dollar Negotiations? Lessons from the Yen/Dollar
Agreement of 1984,"Department of Economics, Working Paper Series qt851615wb
2.http://stats.oecd.org/Index.aspx?DataSetCode=MEI_PRICES
3.Parkes, A & Savvides, A 1999. "Purchasing power parity in the long run and structural
breaks: evidence from real sterling exchange rates,"Applied Financial Economics, Taylor &
Francis Journals, vol. 9(2), pages 117127
4.Prof Peter J. Hammond (2008) LongRun PPP: The Case of Greece: Towards the Transition
5.Rogoff K, The Purchasing Power Parity Puzzle, Journal of Economic Literature, Volume
34, 1996, pages 647.Charles Engel & John H. Rogers, 2006"The U.S. Current Account
Deficit and the Expected Share of World Output,"NBER Working Papers 11921, National
Bureau of Economic Research, Inc..
6.Nusair, S.A 2003. "Testing The Validity Of Purchasing Power Parity For Asian Countries
During The Current Float" Journal of Economic Development vol. 28(2), pages 129147
7.Taylor, A.L and Taylor, M.P (2004). The Purchasing Power Parity Debate Journal of
Economic Perspectives 18, P135158
8.Taylor, Mark P. & Sarno, Lucio, 1998. "The behavior of real exchange rates during the
postBretton Woods period,"Journal of International Economics, Elsevier, vol. 46(2), pages
281312
9.to the Euro: A Maximum Likelihood Cointegration Approach
10. Imported Inflation http://www.investorwords.com/15442/imported_inflation.html
11 Voinea L.G THE PURCHASING POWER PARITY: EVIDENCE FROM THE GREAT FINANCIAL
CRISIS
APPENDIX: RELEVANT COMPUTER OUTPUT AND FIGURES
Appendix A:
Appendix Figures 1 and 2

1
. 0
0

0
. 5
0
0
.
0
0
0
.
5
0
1
.
0
0
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands
Autocorrelations of CPI Denmark

1
. 0
0

0
. 5
0
0
.
0
0
0
.
5
0
1
.
0
0
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands
Autocorrelations of CPI Sweden

1
. 0
0

0
. 5
0
0
.
0
0
0
.
5
0
1
.
0
0
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands
Autocorrelations of CPI UK
31
Appendix Figure 3
Appendix Figure 4
Appendix Figures 5 and 6

1
.
0
0

0
.
5
0
0
.
0
0
0
.
5
0
1
.
0
0
A
u
t
o
c
o
r
r
e
l
a
t
i
o
n
s
o
f
d
e
n
m
a
r
k
u
k
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands

1
.
0
0

0
.
5
0
0
.
0
0
0
.
5
0
1
.
0
0
A
u
t
o
c
o
r
r
e
l
a
t
i
o
n
s
o
f
d
e
n
m
a
r
k
s
w
e
d
e
n
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands
AC Graph of Bilateral Exchange Rates

0
.
4
0

0
.
2
0
0
.
0
0
0
.
2
0
0
.
4
0
A
u
t
o
c
o
r
r
e
l
a
t
i
o
n
s
o
f
D
.
l
d
e
n
m
a
r
k
U
K
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands

0
.
4
0

0
.
2
0
0
.
0
0
0
.
2
0
0
.
4
0
A
u
t
o
c
o
r
r
e
l
a
t
i
o
n
s
o
f
D
.
l
d
e
n
m
a
r
k
s
w
e
d
e
n
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands
AC Graph of Differenced Bilateral Exchange Rates

1
.
0
0

0
.
5
0
0
.
0
0
0
.
5
0
1
.
0
0
A
u
t
o
c
o
r
r
e
l a
t
i o
n
s
o
f
l n
E
R
D
e
n
m
a
r
k
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands

0
.
5
0
0
.
0
0
0
.
5
0
1
.
0
0
A
u
t
o
c
o
r
r
e
l a
t
i o
n
s
o
f
l n
E
R
S
w
e
d
e
n
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands

0
.
5
0
0
.
0
0
0
.
5
0
1
.
0
0
A
u
t
o
c
o
r
r
e
l a
t
i o
n
s
o
f
l n
E
R
U
K
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands
AC Graph of Nominal Exchange Rates

0
.
5
0
0
.
0
0
0
.
5
0
A
u
t
o
c
o
r
r
e
l a
t
i o
n
s
o
f
D
.
l n
E
R
D
e
n
m
a
r
k
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands

0
.
4
0

0
.
2
0
0
.
0
0
0
.
2
0
0
.
4
0
A
u
t
o
c
o
r
r
e
l a
t
i o
n
s
o
f
D
.
l n
E
R
S
w
e
d
e
n
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands

0
.
5
0
0
.
0
0
0
.
5
0
A
u
t
o
c
o
r
r
e
l a
t
i o
n
s
o
f
D
.
l n
E
R
U
K
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands
AC of the Differeneced Nominal Exchange Rates
32
Appendix Firgures 6 and 7
Appendix Figures 7 and 8
Appendix B:
Stata Commands:
Generating Variables:
gen denmarksweden = ERSweden/ERDenmark
. gen denmarkuk = ERUK/ERDenmark
. gen ldenmarksweden=l.denmarksweden
. gen ldenmarksweden= log(denmarksweden)
. gen ldenmarkuk= log( denmarkuk)
. gen lCPIDenmark= log( CPIDenmark)
. gen lCPISweden= log(CPISweden)
. gen lCPIUK=log(CPIUK)
. gen lCPIUnitedKingdom =log( CPIUnitedKingdom )
. gen REdenswe= ldenmarksweden+ lCPISweden lCPIDenmark
Variable Stationarity Testing:
ac lCPIDenmark

1
.
0
0

0
.
5
0
0
.
0
0
0
.
5
0
1
.
0
0
A
u
t
o
c
o
r
r
e
l
a
t
i
o
n
s
o
f
R
E
D
e
n
S
w
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands

1
.
0
0

0
.
5
0
0
.
0
0
0
.
5
0
1
.
0
0
A
u
t
o
c
o
r
r
e
l
a
t
i
o
n
s
o
f
R
E
D
e
n
u
k
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands

0
.
4
0

0
.
2
0
0
.
0
0
0
.
2
0
0
.
4
0
A
u
t
o
c
o
r
r
e
l
a
t
i
o
n
s
o
f
D
.
R
E
D
e
n
S
w
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands

0
.
4
0

0
.
2
0
0
.
0
0
0
.
2
0
0
.
4
0
A
u
t
o
c
o
r
r
e
l
a
t
i
o
n
s
o
f
D
.
R
E
D
e
n
u
k
0 5 10 15 20
Lag
Bartlett's formula for MA(q) 95% confidence bands
33
. dfuller lCPIDenmark, lag(3)
Pperron lCPIDenmark
Ac lCPISweden
Pperron lCPISweden
dfuller lCPISweden, lag(2)
pperron lCPISweden
ac lCPIUnitedKingdom
dfuller lCPIUnitedKingdom (3)
pperron lCPIUnitedKingdom
. ac ldenmarksweden
. dfuller ldenmarksweden, lag(3)
. pperron ldenmarksweden
. ac ldenmarksweden
. ac ldenmarkuk
Dfuller ldenmarkuk, lag(2)
Pperron ldenmarkuk
Dfuller d.ldenmarkuk
pperron d.ldenmarkuk
. ac d.ldenmarksweden
. ac d.ldenmarkuk
. twoway (line ldenmarksweden Time)
. ac d.ldenmarksweden
. dfuller d.ldenmarksweden
Ac d.ldenmarkuk
Dfuller d.ldenmarksweden
Pperron d.ldenmarksweden
Real Exchange Rate Generation:
gen REDenSw= ldenmarksweden+ lCPISweden lCPIDenmark
. twoway (line REDenSw Time) (line ldenmarksweden Time)
. gen REDenuk= ldenmarkuk +lCPIUnitedKingdom lCPIDenmark
. twoway (line REDenSw Time) (line ldenmarksweden Time)
. twoway (line REDenuk Time) (line ldenmarkuk Time)
Gen REDenUK=ldenmarkuk+lCPIUnitedKingdomlCPIDenmark
Real Exchange Rate Stationarity:
ac REDenSw
. dfuller REDenSw, lag(3)
. ac REDenuk
. dfuller REDenuk, lag(2)
. pperron REDenSw
. pperron REDenuk
. ac REDenSw
. ac REDenuk
.ac d.REDenSw
Ac d.REDenuk
Dfuller d.REDenuk
Pperron d. REDenuk
Dfuller d. REDenSw
34
Pperron d.REDenSw
EngleGranger Cointegration:
. reg ldenmarkuk lCPIUnitedKingdom lCPIDenmark
. predict res2, resid
. ac resid1
Dfuller resid, lag(1)
reg ldenmarksweden lCPISweden lCPIDenmark
predict res1, resid
dfuller res1, lag(1)
Error Correction Model:
. reg d.ldenmarkuk d.lCPIUnitedKingdom d.lCPIDenmark l.res2
. ac res2
. reg d.ldenmarksweden d.lCPISweden d.lCPIDenmark l.res1
. ac res2
. dfuller res2, lag(2)
dfuller res2, lag(2)
. reg d.ldenmarkuk d.lCPIUnitedKingdom d.lCPIDenmark l.res2
. reg ldenmarksweden lCPISweden lCPIDenmark
graph combine "Z:\My Documents\ac ex denuk.gph" "Z:\My Documents\ac nex denswe.gph"
. twoway (line REDenuk Time) (line ldenmarkuk Time)
. ac REDenSw
. dfuller REDenSw, lag(3)
. ac REDenuk
. dfuller REDenuk, lag(2)
. dfuller res1, lag(2)
. reg d.ldenmarksweden l.res1
. twoway (line res2 Time) (line ldenmarkuk Time)
. twoway (line res1 Time) (line ldenmarksweden Time)
reg d.ldenmarksweden d.lCPISweden d.lCPIDenmark l.res1
Appendix C:
Computations:
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
Data:
Time CPIDenmark CPISweden CPIUnitedKingdom ERDenmark ERSweden ERUK
1975 22.7 20.1 17.9 5.741258 4.150592 0.451767
1976 24.7 22.1 20.8 6.044192 4.356625 0.556683
1977 27.5 24.7 24.2 6.001142 4.48065 0.573283
1978 30.3 27.1 26.1 5.511075 4.516817 0.521408
1979 33.2 29.1 29.7 5.260533 4.286725 0.4721
1980 37.2 33.1 35 5.635792 4.2292 0.430233
1981 41.6 37.1 39.1 7.120025 5.059825 0.497733
1982 45.8 40.2 42.5 8.333233 6.28225 0.572633
1983 49 43.8 44.5 9.144792 7.667167 0.659783
1984 52.1 47.3 46.7 10.35522 8.273108 0.751567
1985 54.5 50.8 49.5 10.59452 8.60225 0.779342
1986 56.5 53 51.2 8.089108 7.123592 0.682175
1987 58.8 55.2 53.3 6.83835 6.34025 0.611608
1988 61.5 58.4 55.5 6.7302 6.128758 0.562158
57
1989 64.4 62.2 58.4 7.310083 6.446225 0.611408
1990 66.1 68.6 62.4 6.186367 5.918375 0.563025
1991 67.7 75.1 67.1 6.392792 6.045475 0.566883
1992 69.1 76.9 70 6.038375 5.823033 0.569725
1993 70 80.5 71.7 6.482183 7.785442 0.666042
1994 71.4 82.3 73.2 6.360209 7.7157 0.653258
1995 72.9 84.3 75.1 5.60375 7.133616 0.6337
1996 74.4 84.7 77 5.798242 6.707067 0.640817
1997 76 85.3 78.3 6.604075 7.634608 0.6105
1998 77.4 85.1 79.6 6.699308 7.947075 0.603567
1999 79.4 85.5 80.7 6.9799 8.26235 0.618092
2000 81.7 86.2 81.3 8.088034 9.160583 0.660575
2001 83.6 88.3 82.3 8.320808 10.33838 0.694292
2002 85.6 90.2 83.3 7.884284 9.721042 0.66655
2003 87.4 91.9 84.5 6.577 8.078217 0.612283
2004 88.4 92.3 85.6 5.987558 7.346 0.545758
2005 90 92.7 87.3 5.996092 7.473875 0.550117
2006 91.7 94 89.4 5.943008 7.373333 0.543392
2007 93.3 96 91.5 5.442759 6.757733 0.499742
2008 96.5 99.3 94.8 5.098958 6.597108 0.545717
2009 97.8 98.9 96.8 5.359408 7.652608 0.641333
2010 100 100 100 5.62175 7.202192 0.647467
2011 102.8 103 104.5 5.356925 6.489183 0.623817
2012 105.2 103.9 107.4 5.789933 6.768933 0.631067
2013 106.1 103.8 110.2 5.617767 6.513008 0.639742
Molto più che documenti.
Scopri tutto ciò che Scribd ha da offrire, inclusi libri e audiolibri dei maggiori editori.
Annulla in qualsiasi momento.