Stock Market Cycles in Sweden

Antonio de la Torre Gallegos

Departamento de Economía Financiera y Dirección de Operaciones

Facultad de Ciencias Económicas y Empresariales

November 2004 (Preliminary Draft)

Abstract

In this paper we analyze the development of the Stock Market in Sweden since the 1950`s.
We identify the cycles, the upward and downward phases of the market, and analyze a
number of their features: the period or length of the cycle, the duration of the upward and
downward phases, the relative position of the highs, the rate of return during periods of
upward and downward trends and the percentage of retracement of move from low-to-high.
We also compare the degree of similarity with in the cycle periods with those of the US, the
international market of reference. We find that the cycles in Sweden are quite similar to
those of the U.S stock market and stock market phases tend to coincide in time across U.S.

Time series theory is of interest for this analysis primarily because it helps identify market
cycles and turning points in those market cycles. In this work we attempted to extract the
trend and the cycle with different techniques such us Hodrick Prescott (HP) Filter, the
moving averages and the spectral analysis.

The paper is organized as follows. Section 1 is an introduction. Section 2 uses the Hodrick-
Prescott Filter to find the lows or turning points of the cycles. Section 3 uses the moving
average to find the lows or turning points of the cycles. Section 4 uses spectral analysis to
measure the period or length of cycle. Section 5 develops a battery of measures of the
behavior of stock prices during the cycles and shows the results of applying these
techniques to the cycles of the stock market in Sweden and the US. Also, sections 5, offers
some comments on the similarities of these phases with those of the US. Section 6 forecasts
through cyclic analysis. Section 7 concludes.

1
1. Introduction

A viewing of the graphs of many time economic series, such us stock market price data,
often suggests the presence of a cycle in the series. The existence of these cycles in the
stock market contradicts the efficient market hypothesis, based on the rationality of the
investors and the theory of random walk. Two explanations have been put forward for the
existence of cycles in the stock markets. One view (Siegel, 1998) states that cycles reflect
large shifts in consensus perceptions of fundamentals and expectations of the future. An
alternative view (Kaheneman and Tversky, 1979) is that cycles are due to irrational
behaviours of the investors1. In this paper we want to analyze the visual evidence in order
to learn what the characteristics of such cycles are.

In the market, we consider a classic cycle exists when a price starts low (point A), rises
smoothly to a high (point B) over a length of time, and then smoothly falls back to the
another low (point C) over the same length of time. See Figure 1.

Market Cycle

High
B

Wave 2

Wave 1 Amplitude

A Period o lenght C Phase

Low Low
1º Cycle 2º Cycle
Figure 1

The three characteristics of a cycle are period, amplitude and phase. The period of the cycle
or the cycle length is the time required to complete the cycle or the time between lows.
Amplitude is the range of the cycle from low (troughs) to high (crests) measured in
whatever units the cycle is in. The phase is a measure of the time location of a wave trough;
so phasing allows the cyclic analyst to study the relationships between the different cycle
lengths. Once the amplitude, period, and phase of a cycle are known, the cycle can
theoretically be extrapolated into the future. Assuming the cycle remains fairly constant, it
can then be used to estimate future peaks and troughs.

1
This set of alternative explanations to the efficiency of the market is conforming a new paradigm in the
Financial Theory called “Behavioral Finance”.

2
Time series theory is of interest primarily for market analysis because it helps identify
cycles and turning points in those market cycles. Every time series can be separated or
decomposed in four types of components: Seasonality, Trend, Cycling and Irregularity. The
first three components are deterministic which are called "Signals", while the last
component is a random variable, which is called "Noise".

Short definitions of the major components:

Seasonal variation (S): Seasonalities are regular fluctuations which are repeated from year
to year with about the same timing and level of intensity. When a repetitive pattern is
observed over some time horizon, the series is said to have seasonal behavior.

Trend (T): Trend is growth or decay that is the tendencies for data to increase or decrease
fairly steadily over time. A time series may be stationary or exhibit trend over time. Long-
term trend is typically modeled as a linear, quadratic or exponential function.

Cyclical variation (C): Cyclic oscillations are general up-and-down data changes; due to
changes e.g., in the overall economic environment (not caused by seasonal effects) such as
recession-and-expansion.

Irregularities (I): Are any fluctuations not classified as one of the above. This component
of the time series is unexplainable; therefore it is unpredictable. Estimation of I can be
expected only when its variance is not too large. Otherwise, it is not possible to decompose
the series. If the magnitude of variation is large, the projection for the future values will be
inaccurate.

Mathematically and as summation a time series, can be expressed as:

Yt = St + Tt + Ct + I

When we want study a temporal series in the long time, because we are interested in long
term cycles, the first and the last components do not have importance and we can express
the series like the sum of the trend and the cycle. Smoothing data removes random variation
and shows trends and cyclic components. In this work we want extract the trend and the
cycle with different techniques like Hodrick Prescott (HP) Filter, the moving averages and
the spectral analysis.

It is desirable to have certain characteristics in data before it is analyzed for cycles. One of
these is that the fluctuations are of similarly meaningful amplitude of the time period.
When a Stock Market Index with a lot of years is constructed we find that, partly due to
inflation, the index has grown in a compound manner. For data of this type taking
logarithms before looking for cycles is highly desirable. Like the cycle in a series Yt can be
expressed in terms of its turning points, which are local maxima and minima in its sample
path. So we work with yt=ln(Yt) rather than Yt. Turning point in Yt and yt are identical so
that the transformation loses no information.

3
2. Hodrick-Prescott Filter

The Hodrick-Prescott filter or H-P filter is a smoothing mechanism used to obtain a long
term trend component in a time series. It is a way to decompose a given series into
stationary and non-stationary components in such a way that their sum of squares of the
series from the non-stationary component is minimum with a penalty on changes to the
derivatives of the non-stationary component.

Let’s suppose that the original series is composed of a trend component (tt) and a cyclical
component (ct).

That is,
yt = tt + ct , t = 1,…,N

Hodrick and Prescott (1997) suggest a way to isolate ct from yt solving this equation:

⎧N N −1
2⎫
min ⎨∑ ( y t − t t ) + λ
{ t }t = 0, KN +1 ⎩ t =1
2
∑ [(t
t =2
t +1 − t t )− (t t − t t −1 )] ⎬
⎭
t

where λ is the penalty parameter. The first term in the loss function, penalizes the variance
of ct, while the second term puts a prescribed penalty to the lack of smoothness in tt. Put it
differently, the HP filter identifies the cyclical component ct from yt by the trade-off to the
extent to which the trend component keeps track of the original series yt (good fit) against
the prescribed smoothness in tt. Note that as λ approaches to 0, the trend component
becomes equivalent to the original series, while as λ diverge to ∞, ct approaches to the
linear trend. It is customary to set λ to 1,600 for quarterly data. For monthly and annual
data, recommended to use 14,400 and 100, respectively.

By taking derivatives of the loss function with respect to ct, t = 1, · · ·, T and rearranging
them, it can be shown that the solution to the first term can be written as the following
matrix form yT= (λF + IT)tT . We resolve it with an informatics program such us excel and
we obtain the trend component (tt) that we denominate “HP-Trend”.

In figure 2 we present the results of this filter to the natural logarithm of the evolution of
the Sweden Stock Market Index (Affärsvärldens General Index) from 1950 to 2004. This
index has been obtained from Affärsvärlden web page. In the inner window of the same
figure we present the cyclical component (ct) that we denominated “HP-Cycles”. To
insolate this cycles the HP-Trend is plotted as a “zero line” and prices are plotted above and
below that zero centerline.

4
 LOG AGFX Monthly

HP Trend

HP Cycles

Figure 2

In order to identify the stock market cycles, firstly we need to identify the most significant
consecutive lows or turning points which correspond to local minima in the series and
secondly we need to calculate the high between this lows. In order to ensure that we do not
identify shorts or spurious cycles we eliminate these lows if the cycle is less than 2.5 year.

Following these rules, we have detected the lows that fulfill these conditions and we have
marked them by upward arrows. Once we have detected the dates of these minimums in the
HP-Cycles, we looked for the corresponding value in the series of stock-exchange price
data. The date of the lows in HP-Cycles are generally exactly with the one of the stock
series, though in some occasions a small variation exist. When it is not similar we have
taken the nearest price data. Once the lows have been detected, the highs in the series
between these lows can be found easily, and normally agree with a maximum in HP-
Cycles. We have marked them, in figure 2, by downwards arrows.

Table 1 presents the results of the cycles of the Swedish Stock Market from 1950 to 2004.
The averages at the bottom of the table show that the cycle averaged 4.1 years from low-to-
low (the period or length of cycle).

5
 Date of Low HP Date of High Cicle Period
Low Close Close Cycle High Close Close Moves Cycle (years)
Dec-52 1.24 -11.74% Mar-55 1.87 Up 3.9
Nov-56 1.65 -8.17% Aug-61 3.12 Up 5.9
Oct-62 2.62 -14.31% Aug-65 4.26 Up 4.2
Dec-66 3.15 -15.69% May-69 5.14 Up 3.8
Oct-70 3.27 -21.45% Apr-74 5.49 Up 4.2
Dec-74 4.55 -15.82% Apr-76 6.93 Up 2.9
Nov-77 4.63 -19.11% Nov-81 11.29 Up 4.4
Apr-82 10.18 -15.83% Sep-87 54.83 Up 5.6
Nov-87 37.38 -23.92% Aug-89 77.72 Up 4.8
Sep-92 40.14 -49.69% Jan-95 87.20 Up 2.5
Mar-95 84.04 -7.75% Jun-98 215.95 Up 3.5
Sep-98 164.93 -17.72% Feb-00 374.34 Down 4.0
Sep-02 130.30 -42.56%
Averages 4.1
Table 1

The same previous study has been applied on the evolution of the Dow Jones IA from 1950
to 2004, obtaining figure 3 and table 2. The averages show that the longer-term cycle in the
U.S. Stock Market averaged 4,1 years from low-to-low, the same ones that for the Swedish
market. Also it is possible to observe that the dates of low close are closely together. Eight
of the thirteen minimums that have been located in both markets appear in the same year
and the rest in the adjacent years, suggesting that both markets are quite synchronous. At
the present time, as it shows in figure 3, a new cycle of 4 years has formed, that could
extend until the 2005-6.

LOG DOW JONES IA Monthly

HP Trend

HP Cycles

Figure 3

6
 Date of Low Low HP Date of High Cicle Period
Low Close Close Cycle *1 High Close Close Moves Cycle (years)
14-09-53 256 -16.05% 06-04-56 521 Up 4.1
22-10-57 420 -17.20% 13-12-61 735 Up 4.7
26-06-62 536 -23.27% 02-09-66 995 Up 4.3
07-10-66 744 -12.54% 03-12-68 985 Down 3.6
26-05-70 631 -23.82% 11-01-73 1052 Down 4.5
06-12-74 578 -32.57% 12-09-76 1015 Up 3.2
28-02-78 742 -13.43% 27-04-81 1024 Up 4.5
12-08-82 777 -15.29% 25-08-87 2722 Up 5.2
19-10-87 1739 -11.50% 17-07-90 3000 Up 3.0
11-10-90 2365 -9.61% 31-01-94 3978 Up 3.5
04-04-94 3593 -10.22% 17-07-98 9338 Up 4.4
31-08-98 7539 -17.29% 13-01-00 11582 Down 4.1
09-10-02 7286 -21.20%
Averages 4.1

Table 2

3. Moving Averages

Moving Average is a method of smoothing by averaging “n” terms of the time series. It
simply takes a certain number of past periods, adds them together; and then divides by the
number of periods. The following formula is used in finding the moving average of order n,
MA(n) for a period t,

MAt = [Dt + Dt-1 + ... +Dt-n] / n

where n is the number of observations used in the calculation.

A moving average may be centered in which case it is plotted at the middle of the time
interval which it is the average. A centered moving average which has a number of terms
equal to the length of a cycle will remove that cycle from the data and shows trends and
cyclic components in the long time. To eliminate the trend in prices and isolate the cyclic
component Bressert (1991) plotted the centered average as a “zero line” and the prices
above and below that zero centerline. Analyzing these shorter term components of the long-
term cycles can be helpful in identifying major turning points in the longer term cycle.

In figure 4 we present the results of the 48 centered moving averages to the natural
logarithm of the evolution of the Sweden Stock Market Index from 1950 to 2004. At the
bottom of the same figure we represent the cycles.

7
 LOG AGFX Monthly

48 Centred Moving Average

Cycles

Figure 4

As important fact we can observe that these cycles that we detected with both techniques
(HP Filter and moving average) are practically the same, as show in figure 5. Therefore,
we can conclude that HP Filter agrees with the moving average for a number of periods
equivalents to the cycle that is desired to isolate.

Figure 5

The same previous study can be applied to the DJ obtains the same conclusions. In figure 6
we present the 48 centered moving averages to the natural logarithm of the evolution of the
Dow Jones from 1950 to 2004. At the bottom of the same figure we represent the cycles.

8
 LOG DOW JONES IA Monthly

48 Centred Moving Average

Cycles

Figure 6

4. Spectral Analysis

Spectral analysis is the standard method for converting any time series into the sum of a
series of sine waves of different frequencies. Each sine wave has a specific cycle length,
amplitude, and phase relationship to the other sine waves. Fast Fourier Transform ("FFT")
is an abbreviated calculation developed by Hutson and Warren in 1983, which can be
computed in a fraction of the time. FFT sacrifices phase relationships and concentrates only
on cycle length and amplitude. The benefit of FFT is its ability to extract the predominant
cycles from a series of data.

In our study we used the "Interpreted" Fast Fourier Transforms found in the MetaStock
computer program. This indicator shows the three predominant cycle lengths and the
relative strength of each of these cycles.

The Interpreted FFT shows that exist a predominate cycle lengths in Sweden market of 3,7
years, near of the cycle previous calculate. In US market shows that exists the same
predominate cycle lengths, 3.7 years. This discrepancy can be eliminated by dividing the
period between year 1982 and 1987 in two cycles. We have not done it since an inferior
cycle to the 2.5 years would be obtained.

9
5. Describing cycle characteristics and statistics.

Once the cycles have been identified, we calculate a battery of statistics that describe the
behavior of stock prices in each of the cycles. This behavior can then be compared across
countries and across cycles, in search for relevant differences that may shed light on the
determinants of stock market evolution. The eight characteristics of the phases of the
market that we study are:

- The period or length of cycle (years from low-to-low)

- The duration of the upward phases (years from low-to-high)
- The duration of the downward phases (years from high- to-low)
- The relative position of the high
- The percentage of advance or decline during the length of cycle (percentage the
advance/decline from low-to-low).
- The percentage of return during periods of upward phases (percentage the advance
from low-to-high).
- The percentage of return during periods of downward phases (percentage the
decline from high-to-low).
- The percentage of retracement of move from low-to-high (decline/advance).

The averages of all cycles, however, are not very helpful in defining a market, so let’s look
at a market qualifier – those cycles that exceeded the initial low versus those that did not
exceeded the initial low. As you can see in tables 3, 4, 5 and 6, there is a distinct difference
between those cycles.

Table 3 presents the cycles of Swedish Stock Market that exceeded the initial low. The
statistics in this table can be summarized as follows. Firstly, the averages at the bottom of
the table show that the cycle averaged 4.2 years from low-to-low, 3.1 years from low-to-
high, 1.1 years from high-to-low and the high occurs in the 73% of length of the cycle. So,
it is clear that in cycles that exceeded the initial low the upward phases tend to be longer
than the downward phases and the high shifts to the right of the ideal midpoint causing
right translation. Secondly, the average advance during the length of cycle (from low-to-
low) was 68.8 %, the average advance or return during periods of upward phases (from
low-to-high) was 122.9 % and the average decline or return during periods of downward
phases (from high-to-low) was 23.5 %. The percentage tends to be strongly higher during
upward phases because the trend is upward. Thirdly, the retracement of move from low-to-
high (decline/advance) implies an average of 51.3 %.

10
 Date of Period Years Year Position % Move % Advance % Decline Decline/
Low Close Cycle (years) Low to High High to Low High L-Lnext L-H H-Lnext Advance
Dec-52 3.9 2.3 1.7 0.57 33.0% 51.2% 12.0% 35.5%
Nov-56 5.9 4.8 1.2 0.80 58.7% 89.5% 16.2% 34.4%
Oct-62 4.2 2.8 1.3 0.68 20.5% 62.8% 26.0% 67.4%
Dec-66 3.8 2.4 1.4 0.63 3.7% 63.3% 36.5% 94.2%
Oct-70 4.2 3.5 0.7 0.84 39.3% 68.1% 17.1% 42.2%
Dec-74 2.9 1.3 1.6 0.46 1.8% 52.3% 33.2% 96.6%
Nov-77 4.4 4.0 0.4 0.91 119.8% 143.6% 9.8% 16.6%
Apr-82 5.6 5.4 0.2 0.97 267.2% 438.6% 31.8% 39.1%
Nov-87 4.8 1.8 3.1 0.36 7.4% 107.9% 48.4% 93.2%
Sep-92 2.5 2.3 0.2 0.93 109.4% 117.2% 3.6% 6.7%
Mar-95 3.5 3.3 0.3 0.93 96.3% 157.0% 23.6% 38.7%
Averages 4.2 3.1 1.1 0.73 68.8% 122.9% 23.5% 51.3%

Table 3

Table 4 presents the unique cycle of Swedish Stock Market since 1950 that did not
exceeded the initial low. The period of the cycle was 4 years from low-to-low, 1.4 years
from low-to-high, 2.6 years from high-to-low and the high occurs in the 35% of length of
the cycle. In this case, it is clear that in cycles that not exceeded the initial low the
downward phases tend to be longer than the upward phases and the high shifts to the left of
the ideal midpoint causing left translation. On the other hand the decline during the length
of cycle (from low-to-low) was 21 %, the advance or return from low-to-high was 127 %
and the decline from high-to-low was 65.2%. Finally, the retracement of move from low-to-
high (decline/advance) was 116.5 %.
Date of Period Years Year Position % Move % Advance % Decline Decline/
Low Close Cycle (years) Low to High High to Low High L-Lnext L-H H-Lnext Advance
Sep-98 4.0 1.4 2.6 0.35 -21.0% 127.0% 65.2% 116.5%
Averages 4.0 1.4 2.6 0.35 -21.0% 127.0% 65.2% 116.5%

Table 4

Table 5 presents the cycles of U.S. Stock Market that exceeded the initial low from 1950.
The cycle averaged 4.1 years from low-to-low, 3.5 years from low-to-high, 0.6 years from
high-to-low and the high occurs in the 84 % of length of the cycle. The average advance
from low-to-low was 53.9 %, the average advance from low-to-high was 103.2 % and the
average decline from high-to-low was 23.2 %. Finally, the retracement of move from low-
to-high (decline/advance) implies an average of 51.3 %.

Date of Period Years Year Position % Move % Advance % Decline Decline/

Low Close Cycle (years) Low to High High to Low High L-Lnext L-H H-Lnext Advance
14-09-53 4.1 2.6 1.5 0.62 64.1% 103.5% 19.4% 38.1%
22-10-57 4.7 4.1 0.5 0.89 27.6% 75.0% 27.1% 63.2%
26-06-62 4.3 4.2 0.1 0.98 38.8% 85.6% 25.2% 54.7%
06-12-74 3.2 1.8 1.5 0.55 28.4% 75.6% 26.9% 62.5%
28-02-78 4.5 3.2 1.3 0.71 4.7% 38.0% 24.1% 87.6%
12-08-82 5.2 5.0 0.2 0.97 123.8% 250.3% 36.1% 50.5%
19-10-87 3.0 2.7 0.2 0.92 36.0% 72.5% 21.2% 50.4%
11-10-90 3.5 3.3 0.2 0.95 51.9% 68.2% 9.7% 23.9%
04-04-94 4.4 4.3 0.1 0.97 109.8% 159.9% 19.3% 31.3%
Averages 4.1 3.5 0.6 0.84 53.9% 103.2% 23.2% 51.3%

Table 5

11
Table 6 presents the cycles of U.S. Stock Market that did not exceed the initial low. In this
case, the cycle averaged 4.1 years from low-to-low, 2.1 years from low-to-high, 2 years
from high-to-low and the high occurs in the 50 % of length of the cycle. The average
decline from low-to-low was 9 %, the average advance from low-to-high was 50.9 % and
the average decline from high-to-low was 39.4 %. Finally, the retracement of move from
low-to-high (decline/advance) implies an average of 121.9 %.

Date of Period Years Year Position % Move % Advance % Decline Decline/

Low Close Cycle (years) Low to High High to Low High L-Lnext L-H H-Lnext Advance
07-10-66 3.6 2.2 1.5 0.59 -15.2% 32.4% 35.9% 146.9%
26-05-70 4.5 2.6 1.9 0.58 -8.4% 66.7% 45.1% 112.6%
31-08-98 4.1 1.4 2.7 0.33 -3.4% 53.6% 37.1% 106.3%
Averages 4.1 2.1 2.0 0.50 -9.0% 50.9% 39.4% 121.9%

Table 6

Table 7 summarizes the results on the basic measures of both stock markets. If we have a
look across Sweden and U.S. market cycles that exceeded the initial low are quite similar in
terms of the length, position of the high and the percentage of retracement of move from
low-to-high (decline/advance), suggesting that both markets are quite synchronous. In
cycles that did not exceed the initial low the differences are higher. The length of the cycle
are very similar, but the years from low to high tend to be longer in U.S. than Sweden and
therefore the years from high to low tend to be shorter in U.S. than Sweden.

Sweden U.S.
CYCLE MOVE UP
Years from low-to-low 4.2 4.1
Years from low-to-high 3.1 3.5
Years from high- to-low 1.1 0.6
Position of the high 0.73 0.84
Percentage the advance from low-to-low 68.8% 53.9%
Percentage the advance from low-to-high 122.9% 103.2%
Percentage the decline from high-to-low 23.5% 23.2%
Decline/advance 51.3% 51.3%
CYCLE MOVE DOWN
Years from low-to-low 4.0 4.1
Years from low-to-high 1.4 2.1
Years from high- to-low 2.6 2.0
Position of the high 0.35 0.5
Percentage the decline from low-to-low -21.0% -9.0%
Percentage the advance from low-to-high 127.0% 50.9%
Percentage the decline from high-to-low 65.2% 39.4%
Decline/advance 116.5% 121.9%

Table 7

12
If we do a correlation analysis we can measure the relationship between the two indexes. In
the inner window of the figure 7 the resulting value of this analysis, the "correlation
coefficient", shows that if changes in U.S. market will result in changes in the Sweden
Market. The figure shows a strong positive relationship mainly since 1995, except during
the year 2000. This means that an increase in the U.S. market usually predicts an increase
in the Sweden market.

LOG Dow Jones Montly

LOG AFGX Montly

AFGX HP Cycles

DJ HP Cycles

Correlation Coefficient

Figure 7

6. Market forecasting through cyclic analysis.

Once the characteristics of cycles are known, assuming remains fairly constant, they can
then be used to estimate future high and lows in terms of probability .We assumed a risk
because the cycles are not exact; they contract, extend and sometimes skip a beat. Firstly,
we need to 1decided if the next cycle move is up (cycles that exceeded the initial low) or
down (cycles that not exceeded the initial low).

If the cycle move is up and we considered the characteristics of the U.S. cycle, to which
they tend to converge all the markets, to calculate the next high and low of the 4-year cycle
follow the next steps. To determine the most probable time of the cycle high, from the most
recent 4 year cycle low, calculate 3.5 years and to determine the most probable price
calculate a percent rise of 103.2 % from the low close of the cycle. Next, to determine the
most probable time for the next 4-year cycle low, from the most recent cycle low calculate
4.1 years or 0.6 years from the actual cycle high. The most probable price will be a percent

13
rise of 53.9 % from the low close of the cycle or a percent decline of 23.2 % from the actual
cycle high.

If the cycle move is down follow the next steps. To determine the most probable time of the
cycle high, from the most recent 4 year cycle low, calculate 2.1 years and to determine the
most probable price calculate a percent rise of 50.9 % from the low close of the cycle. Next,
to calculate the most probable time for the next 4-year cycle low, from the most recent
cycle low calculate 4.1 years or 2 years from the actual cycle high. The most probable price
will be a percent decline of 9 % from the low close of the cycle or a percent decline of 39.4
% from the actual cycle high.

Still, at the beginning of October of 2004, it is premature to know as it will be the present
evolution of the market, is to say if the present cycle around the 4 years will be bullish or
bearish.

S&P 500 Weekly

Sep 98

Oct 02

Dec 94

Oct 90

Dec 87

7. Conclusion

In the previous sections, we have seen how the financial markets movement follows time
cycles. The existence of these cycles contradicts the efficient market hypothesis, based on
the rationality of the investors and the theory of random walk. The explanations of this
behaviour are that the markets are not only based on fundamentals, also it is necessary to
consider the psychology of the investor, determined by irrational but known behaviours.
This set of alternative explanations to the efficiency of the market is conforming a new
paradigm in the Financial Theory called “Behavioral Finance”.

14
This paper uses monthly stock market data to study the characteristics of cycles and trend
of the Sweden and U.S. stock market during the period 1950-2004. In order to detect
cyclical patterns in stock price movements we use the hp filter, the moving average and
spectral analysis. We identify some characteristics in the stock price cycles, such the period
or length of cycle, the duration of the upward and downward phases, the relative position of
the high, the percentage of advance or decline during the length of cycle, of return during
periods of upward and downward phases and the percentage of retracement of move from
low-to-high. We compare these characteristics with those US. Our results show that both
markets are quite synchronous mainly after June of the year 1995, although some
differences still remain. Assuming the cycle remains fairly constant, it can then be used to
estimate future peaks and troughs.

7. References

- Bloomfield, P (2000), Fourier Analysis of Time Series: An Introduction. John Wiley &
Sons. N.Y.

- Boldin, M.D. (1994) Dating Turning Points in the Business Cycle. Journal of Business 67:
97–131

- Box, G. E. P., Jenkins, G. M., and Reinsel, G. C. (1994), “Time Series Analysis:
Forecasting and Control”, 3rd ed. Prentice Hall, Englewood Clifs, NJ.

- Bressert, W.J: The Power of Oscillator/Cycle Combinations. Bressert&Asociates, 1991.

- Gómez Biscarri, J and Perez de Gracia, F. “Stock Market Cycles and Stock Market
Development in Spain”. Working Paper No.01/02 .March 2002.

- Hodrick, Robert, and Edward C. Prescott (1997), “Postwar U.S. Business Cycles: An
Empirical Investigation”. Journal of Money, Credit, and Banking.

- Kaheneman, D and Tversky, A. (1979), “Prospect Theory: An analysis of decision under

risk” . Econometrica, nº 47.

- Pagan, A.R., Sossounov, K.A. (2002), “A Simple Framework for Analyzing

Bull and Bear Markets”. Journal of Applied Econometrics, (forthcoming).

- Siegel, J.J. (1998), “Stocks for the Long Run”. New York: McGraw-Hill.

- Shiller, R.J. (2000): Irrational Exuberance. NewYork: Broadway Books.

- Warren, A and J. Hutson (1983), “Fast Fourier Transform (Fft) Basic language code”
Technical Analysis of STOCKS & COMMODITIES, Volume 1: January.

15