Content uploaded by Murat Akkaya
Author content
All content in this area was uploaded by Murat Akkaya on May 02, 2018
Content may be subject to copyright.
Annals of “Dunarea de Jos” University of Galati
Fascicle I. Economics and Applied Informatics
Years XXIII – no1/2017
ISSN-L 1584-0409 ISSN-Online 2344-441X
www.eia.feaa.ugal.ro
The Role of Consumer Confidence as a Leading Indicator
on Stock Returns: A Markov Switching Approach
Koy AYBEN, Akkaya MURAT
A R T I C L E I N F O A B S T R A C T
Article history
:
Accepted March 2017
Available online
April
201
7
Investor’s psychological and emotional factors lead to irrationality in financial decision
making and anomalies in prices. Investor sentiment and psychology help to elucidate
phenomena in financial markets that cannot be explained by traditional theory. The aim of
this study is two-fold: it investigates whether mutual regime switching behavior exists
between the consumer indices and equity index, and examines their dynamics in response
to each other in different regimes. This study applies the Markov Regime Switching model
to monthly data from the BIST100 Return Index, Bloomberg Confidence Index, TUIK
Confidence Index, Real Sector Confidence Index for the period between 2007:01 and
2016:06. The results indicate if consumer indices point out negative signals, capital market
still gains in normal periods of economy. If they only in a recession or an expansion regime
do, each of the indices moves in the same direction.
© 201
7
EAI. All rights reserved.
JEL Classification
G10, G11
Keywords:
Behavioral Finance, Consumer
Confidence, Stock Returns, Markov
Regime Switching, Nonlinearity
1. Introduction
Investor sentiment in behavioral finance was made popular by the Prospect Theory developed by
Kahneman and Tversky (1979). This theory claims that investors may not be rational, instead they should be
considered to be "normal". Investors make decisions not only about risk, return, and utility maximization,
they also decide on satisfaction shaped by cognitive and emotional biases. Investor’s intuition and emotions,
systematic errors in financial decisions, and other psychological factors such as cognitive dissonance, lead to
irrationality and anomalies in prices. Moreover, investors overreact or underreact by noise. These behaviors
cause risk in financial markets. Investor sentiment and psychology leads to understand the phenomenon in
financial markets that cannot be explained by traditional theory.
Consumer expectations and confidence are one of the most important variables in studies of investor
sentiment. Consumer confidence and investor sentiments are important indicators in investor’s expectations
and perceptions. While confidence is a difficult concept to pin down, it has a substantial impact on financial
markets. Both confidence and a lack thereof have been important following the financial crises in last two
decades.
The aim of this study is to analyze the relationship between consumer confidence and stock returns
during different phases of the market cycle with regime switching models. The Markov Regime Switching
models represent the most important example of non-linear time series models.
Among nonlinear models, the Markov switching techniques have been popular for studies on finance.
Studies create many interesting new questions, such as if it is possible to distinguish between distinct
regimes in stock market returns, how the regimes differ, how frequently regime switches, or when these
changes occur. Moreover, the predictability of the returns and regime switches are also important questions.
The answers to these questions give us new information about stock market returns (Schaller and Norden,
1997). Having built important strategic partnerships (NASDAQ, London Stock Exchange Group – LSEG,
London Metal Exchange, European Bank for Reconstruction and Development – EBRD,…etc.) as part of its
vision to make Istanbul the largest financial district in the region, Borsa Istanbul is an attractive capital
market between other emerging markets. The main purpose of the study is to investigate whether mutual
regime switching behavior exists between the consumer indices and Turkish equity index. The secondary
purpose is to analyze the relationship between the confidence indices and equity index in different regimes
with probabilities and durations. The Markov switching vector autoregressive (MS(M)-VAR(p)) models with
an intercept successfully capture the heteroskedasticity, skew and fat tails of the stock return distribution.
Accordingly, MSI(M)-VAR(p) models are used with regime shifts including the intercept in this study.
Istanbul Comm erce University, Istanbul, Turkey, Istanbul Gelisim University, Istanbul, Turkey. E- mail addresses: akoy@ticaret.edu.tr (K. Ayben), makkaya68@gmail.com (A.
Murat)
37
2. Literature review
Factors affecting stock prices and prediction of stock prices have always been a popular subject in
finance research. There have been various studies examining the consumer confidence. These studies have
focused mainly on confidence index and stock returns.
The summary literature of the relationship between consumer confidence and stock returns is given
Table 1.
Table 1: The literature of the relationship between consumer confidence and stock returns
Author Date Article Journal
Lee, Shleifer and Thaler 1991 Investor Sentiment and Closed- End Fund Puzzle The journal Of Finance, Vol. XLVI, No.1, March 1991, ss. 75-109
Neal and Wheatley 1998 Do Measures of Investor Sentiment Predict Returns Journal of Finacial and Quantitative Analysis, 33 (4), 1998, ss. 523-548.
Barberis, Shleifer and Vishny 1998 A Model of Investor Sentiment Journal of Financial Economics, 49, 1998, ss. 307-343.
Elton, Gruber and Busse 1998 Do Investors Care About sentiment? The Journal of Business, Vol. 71, No.4, ss. 477-500.
Fisher and Statman 2000 Investor Sentiment and Stock Returns Financial Analysts Journal, 56, Sayı 2, 2000, March-April, 16-31.
Howrey 2001 The Predictive Power of the Index of Consumer Sentiment Brookings Papers on Economic Activity, I, 2001, ss. 175-216
Baker and Stein 2002 Market Liquidity as a Sentiment Indicator NBER, Working Paper 13189, June 2007, ss. 1-31.
Chen 2002 Domestic Investors Herding Behavior in reaction To Foreign Trading International Conference on Finance, 2002
Brown and Cliff 2004 Investor Sentiment and The Near-term Stock Market Journal of Empirical Finance, 11, (2004), ss. 1-27.
Brown and Cliff 2005 Investor Sentiment and Asset Valuation Journal of Business, 78 (2), 2005, ss. 405-440
Baker and Wurgler 2006 Investor Sentiment and The Cross-Section Of Stock Returns Journal of Finance, Vol. LXI, No. 4 August 2006, ss. 1645-1680
Baker and Wurgler 2007 Investor Sentiment in the Stock Market Journal of Economic Perspectives, Volume 21, Number 2, Spring 2007, ss. 129-151
Kumar and Lee 2006 Retail Investor Sentiment and Return Comevements Journal of Finance, Volume 61, Issue 5, 2006, ss. 2451–2486
Barber, Odean and Zhu 2007 Systematic Noise http://faculty.gsm.ucdavis.edu/~nzhu/papers/ systematicnoise.pdf
Verma, Baklacı and Soydemir 2008
Canbaş and Kandır 2009 “Investor Sentiment and Stock Returns: Evidence from Turkey Emerging Markets Finance and Trade, Vol. 45, No. 4, 2009, ss. 36-52.
Lemmon and Ni 2010
Baker, Wurgler and Yuan 2012 Global, Local and Contagious Investor Sentiment Journal of Financial Economics, 104, (2012), ss. 272-287.
Uygur and Taş 2013
Sohn 2013 What Does Sentiment Reflect: Animal Spirit or Risks http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2177181
The Impact of Rational and Irrational Sentiments of Individual and
Institutional Investors on DJIA and S&P 500 Index Returns
The Effects of Investor Sentiment on Speculative Trading and Prices
of Stock and Index Options
The Impact of Investor Sentiment on Returns and Conditional Volatility
of Internatioanal Stock Markets Quality and Quantity, Volume 48, Issue 3, May 2014, ss. 1165-1179.
http://ssrn.com/abstract=1572427.
Applied Financial Economics, 18, 2008, ss. 1303-1317.
Source: Compiled by the authors
The current Michigan University Confidence Index designed by George Katona measures consumer
spending. Established in the late 1940s, it was first study on consumer expectation and confidence to
measure empirically (Katona, 1968).
Consumer confidence has been popular topic in finance studies by Otoo (1999). Using the Michigan
University Confidence Index, Otoo investigates the relationship between consumer confidence index and the
stock prices, observing a strong positive relationship between the two.
The relationship between the stock market and confidence has been observed using different
variables and indices in various studies including, but not limited to: Fisher and Statman (2003), Lemmon
and Portniaguina (2006), Baker and Wurgler (2007), Bremmer (2008) and Schmeling (2009).
Kandır (2006), Korkmaz & Çevik (2009), Topuz (2011), Kale and Akkaya (2015), and Kilic and
Cankaya (2016) analyze the relationship between consumer confidence and the stock market for Turkey’s
capital market. They observe significant relationship and effects of consumer confidence on stock returns.
Korkmaz & Çevik (2009) analyze the causality between BIST100 index return and real sector confidence
index in two stages. They argue that there is a feed-back effect between equity index and confidence index
and that these two factors impact each other simultaneously. Topuz (2011) and Kale and Akkaya (2015) find
one-direction causality from stock prices towards consumer confidence.
Moreover, there is high correlation between consumer confidence and real economic activity found
in studies by Ludvigson (2004) and Howrey (2001). Investors should characterize the fluctuations in the
stock markets to understand the effects of consumer confidence on stock. Hamilton’s Markov-switching
autoregressive (MS-AR) models are successful in studies for real output fluctuations.
Regime Switching and Markov Regime Switching models represent the most important example of
non-linear time series models of current application. Regime is defined as an unobserved state variable
38
affecting the levels, volatility, or correlations of the distributions of stock returns (Perez-Quirors and
Timmermann (2000), Guidolin and Timmermann (2008), Chung and Yeh (2008)).
The predictive power of consumer confidence indexed by Regime Switching and Markov Regime
Switching models has been observed in various studies such as Batchelor (2001), Chen (2009), Chen (2011),
Chung, Hung, and Yeh (2012), Çevik, Korkmaz, and Atukeren (2012).
Batchelor, (2001) studies business confidence, consumer confidence, and the state of the economy in
the US and the UK using a Markov switching model. He concludes that there is significant correlation between
business confidence and state of economy. As consumer confidence decreases, the probability that growth
will subsequently fall, increases. Moreover, rising consumer confidence leads to an increase in the probability
of recovery.
Chen (2011) studies correlation between the lack of consumer confidence and stock returns during
market fluctuations using Markov-switching model. He analyzes asymmetric effects of consumer confidence
on stock returns by using monthly returns on Standard & Poor's S&P 500 price index. It is observed that
market pessimism has larger impacts on stock returns in bear markets. It also finds a higher probability that
the lack of consumer confidence causes a switch to a bear market regime.
Chung, Hung, and Yeh (2012), analyze the effects of investor sentiment on stock returns across
economic expansion and recession states. In this study, the Markov switching model is used to show the
otherwise unobservable dynamics of the changes in the economic regime by using portfolio formed on size,
book-to market equity ratio, dividend yield, earnings-to-price ratio, age, return volatility, asset tangibility,
growth opportunities, and 11 widely documented anomalies. They observe predictive power for the returns.
Furthermore, sentiment performs both in-sample and out-of-sample only in the expansion state. However,
the predictive power of sentiment is generally insignificant in a recession state.
Çevik, Korkmaz, and Atukeren (2012) show evidence in all shares stock returns in the USA by using
time-varying Markov regime-Switching (MS) model. They use the US Institute for Supply Management’s (ISM)
manufacturing and Nonmanufacturing Business Activity Index (NMBAI) in the transition equations. They
observed effects of the developments in the ISM manufacturing index on the regime-switching probabilities
in both bull and bear stock market periods. Also the information content in the ISM manufacturing confidence
index influences the stock market regimes.
3. Data and Methodology
As a confidence index, the Bloomberg Consumer Confidence Index (BCI) and Turkey Statistical
Institution Consumer Confidence Index (TUIKCI) have been conducted monthly in Turkey.
The Bloomberg HT confidence index has covered CNBC-e CCI in name since November 2015. The BCI
(old CNBC-e) has been announced on a monthly basis since January 2002.
The TUIKCI is comprised of the results of the consumer tendency survey carried out in cooperation
with the Central Bank of the Republic of Turkey (CBRT) and has been announced on a monthly basis since
January 2004.
There is also third confidence index called Real Sector Confidence Index (RSCI) in Turkey. It is
measured and announced by the CBRT. The RSCI tracks the general views of the real sector about general
economic outlook. The RSCI has been announced on a monthly basis since January 2007.
3.1. Data
BIST100 Return Index data, BCI data, TUIKCI data and RSCI data have been gathered from official
web site of these institutions.
We use monthly data from 2007:01 to 2016:06 which includes 114 observations, considering data
limitations. Data are analyzed using Oxmetrics program.
Figure 1. BCI, RSCI and TUIKCI
39
Figure 2. BIST 100 Return Index
3.2. Markov Switching Vector Autoregressive Model
A Markov chain is composed of independent random variables. The basic idea for the Markov Regime
Switching Model is to describe a situation or stochastic process which determines the change from one regime
to another via a Markov chain. Markov chain is used to model the behavior of a state variable or combination
of variables which cannot be directly observed but determines the regime. In a Markov Regime Switching, the
state of the economy (st) cannot be directly observed, although the time series variable (vt) can be observed.
Any period of economy whose properties depend on the observation values, is obtained as probability. At the
same, time those observations are supposed to be dependent on the properties of the regime. When the state
of the economy in the Markov regime is determined, the next regime can be expressed as a
probability.(Bildirici et all).
In the Markov Regime Switching Models, there is a K-dimensional vector time series process (yt)
dependent upon an unobservable regime variable. The probability of being in a particular state of the
economy is represented by st (Krolzig, 2000).
f(yt|Yt−1;Xt; ϴ1) ; st = 1 (1)
p(yt|Yt−1;Xt; st) = {
f(yt|Yt−1; Xt; ϴM) ; st = M
Xt: exogenous variables;
ϴ: parameter vector associated with regime M.
The regime-generating process in the Markov Switching models is an ergodic Markov chain. Finite
number of states is defined by the transition probabilities (Krolzig 2000).
; (2)
st follows an ergodic M-state Markov process with an irreducible transition matrix:
(3)
In a two-stated model, the transition probabilities of moving from one state to the other,
P(st+1=1|st = 1) = p11,
P(st+1=2|st = 1) = p12,
P(st+1=1|st = 2) = p21,
P(st+1=2|st = 2) = p22. (4)
Obviously, for the pijs to define proper probabilities, they should be nonnegative, while it should also
hold that p11 + p12 = 1 and p21 + p22 = 1 (Franses and Dijk 2000).
The probability of the regime which is effective at time t, conditional on the information at time t− 1,
and only depends on the statistical inference on st−1:
Pr(st|Yt−1;Xt; St−1) = Pr(st|st−1)
(5)
40
Markov Switching time series analysis was first implemented by Hamilton (1989) in the business
cycle. Hamilton investigates the possibility that macroeconomic variables progress on a cyclical time scale
between calendar time (month, quarter or some other kind of units) and economic time. The time
transformations between economic and calendar time depends on the economic history of the process such
as whether the economy has been in a cyclical expansion or contraction.
The two main types of the Markov Switching Model, are MSM and MSI models. In MSM Model, the
regime switches according to the conditional mean (µt),while in MSI Model, the regime switches according to
the constant (cst).
MSM Model: yt- µt = ɸ(yt-1 - µt-1) + ut (6)
MSI Model: yt– cst= ɸyt-1 + ut (7)
In financial time series, MSIH models’ results are generally the most significant. MSIH model is
derivative from MSI model. An added factor Ω1/2 to the MSI model represents variance covariance matrix of
variable (y) Ω st:
MSIH Model: yt– cst= ɸyt-1 + ut + Ω1/2 (8)
ɸ is an n x n matrix of regime-dependent autoregressive coefficients
ut is an (n*1) unobservable zero mean white noise vector process
Matrix Ω1/2 represents
Ω st =Var[yt| ϗt-1,st] ; (9)
ϗt-1: denotes time t-1 information of all past observations and states.
A two-state bivariate Matrix Ω1/2 is (Guidolin, 2016):
t (10)
To define the nonlinear relationship between multiple variables, we can use the Markov switching
vector autoregressive model. The basic p lagged VAR(p) model process is:
yt= c + [A1yt-1+…+Apyt-p] + ut ; (11)
An is (n*n) coefficient matrices
The general form of a Markov-switching vector autoregressive (MS-VAR) model is (Krolzig 1998,
2000):
yt= c(st) + [A1(st)yt-1+…+Ap(st)yt-p] + ut (12)
A VAR with regime shifts in the mean is called a MSM(M)-VAR(p):
yt= µ(st) + A1(st)(yt-1 - µ(st-1)) +…+ Ap(st)(yt-p - µ(st-p)) + ut (13)
ut ̴̴̴ NID( 0, Ʃ(st))
If the regime shifts affect the intercept of the VAR, this is called a MSI(M)-VAR(p):
yt= c(st) + A1(st)yt-1 +…+ Ap(st)yt-p + ut (14)
The transition in the MSI-VAR is smooth compared to the MSM-VAR model. These models are the
subclass of MS-VAR models. (Krolzig 1998, 2000).
If the regime shifts affect the intercept of the VAR and the model includes a variance covariance
matrix, this is called a MSIH(M)-VAR(p) process:
yt= c(st) + A1(st)yt-1 +…+ Ap(st)yt-p + ut + Ω1/2 (15)
MSIH(M)-VAR(p) model refers to Markov switching, Intercept regime dependent, Vector autoregressive and
heteroskedastic model (Guidolin 2016).
In MS-VAR, it is possible to analyze models how other variables are affected if shocks are applied to a
variable. Impulse response functions of the models show these relations in different regimes. For example, if
the model includes two variables and has three regimes, there should be six (2*3=6) relations to analyze.
4. Results
We apply models to three different consumer confidence index - equity index duals (BCI-BIST100,
RSCI-BIST100 and TUIK-BIST100). The applied models include either two or three regimes and different lags
(0-4). Models are applied to the logarithmic differences of the variables. We select the model between the
41
models with a Davies’ criteria smaller than 0.01 which rejects the Davies criteria’s null hypothesis of linearity.
LR linearity criterion is an important criterion; it shows how much the non-linear model explains the relation
more than the linear model. Moreover, other criterions such as log-likelihood, Akaike (AIC), Hannan-Quinn
(HQ) and Schwarz (SIC) are examined.
4.1 The Bloomberg Confidence Index and the BIST100
The MSIH(3)-VAR(1) Model with 3 regimes and 1 lag is the best model according to the criteria
established above and the distribution of the observations among the different nonlinear models, if any
additional lag of any variable is added; the Akaike's (AIC), Hannan-Quinn (HQ), Schwarz (SIC) criterions
increase and the LR linearity decreases, thus the power of the model decreases.
MSIH(3)-VAR(1) model have three regimes. The first regime includes a recession with high volatility,
the second regime contains moderate growth with low volatility, and the third regime sees expansion with
high volatility. Volatility differentiates the second and third regime in the model.
Table 2: Information Criterions
Model log-
likelihood
AIC HQ SIC LR linearity
test
DAVIES
MSIH(3)-VAR(1) 42.5653 -4.3562 -4.1100 -3.7494 42.5653 0.0002
Table 3 shows the coefficients of the model. The model has three regimes, the first regime is in a
recession, the second one has moderate growth and the last one is experiencing expansion. The constants in
regime 1 (recession) are negative. If the market is in a recession, the consumer confidence index and the
return on the equity index should decline. In regime 2 (moderate growth), there is a weak decline in the BCI
while the BIST100 returns are positive. The constants in regime 3 (expansion) are positive. If the market is in
expansion, the consumer confidence index should rise and the return on equity index should be positive.
Table 3: Coefficients
BCI BIST100
Constant (Reg.1) -0.1284
-0.0883
Constant (Reg.2) -0.0092
0.0128
Constant (Reg.3) 0.0749
0.0368
BCI_1 -0.1432
-0.1061
BIST100_1 0.3528
-0.0338
SE (Reg. 1) 0.0636
0.1097
SE (Reg. 2) 0.0520
0.0613
SE (Reg. 3) 0.1104
0.0943
Table 4 shows the probabilities of regime transitions. If the market is in regime 1, by the following
month the market should be 52.10 % in regime 1, 00.74 % in regime 2 and 47.16 % in regime 3. If the market
in regime 2; the following month the market should be 94.77 % in regime 2, 05.10 % in regime 1 and 00.13 %
in regime 3. If the market is in regime 3; the following observation should be 74.61% in regime 3, 5.17 % in
regime 1 and 20.22 % in regime 2.
Table 4: Matrix of Transition Probabilities
Regime 1 Regime 2 Regime 3
Regime 1 0.5210 0.0074 0.4716
Regime 2 0.0510 0.9477 0.0013
Regime 3 0.0517 0.2022 0.7461
Table 5 shows the transition possibilities for the whole model.
In the examining period, the highest number of observations (80) and the highest probability (72.07
%) and the longest duration (19) belong to the moderate growth regime. The minimum number of
observations (11) belongs to the recession regime.
Table 5: Regime Probabilities
Number of Obs. Probability Duration
Regime 1 11 0.0965 2
Regime 2 80 0.7207 19
Regime 3 21 0.1828 4
42
2008 2009 2010 2011 2012 2013 2014 2015 2016
-0.25
0.00
0.25
MSIH(3)-VAR(1), 2007 (3) - 2016 (6)
DL BCI DL BIST100RI
2008 2009 2010 2011 2012 2013 2014 2015 2016
0.5
1.0 Probabilities of Regime 1
filtered
predicted
smoothed
2008 2009 2010 2011 2012 2013 2014 2015 2016
0.5
1.0 Probabilities of Regime 2
filtered
predicted
smoothed
2008 2009 2010 2011 2012 2013 2014 2015 2016
0.5
1.0 Probabilities of Regime 3
filtered
predicted
smoothed
Figure 3: Regime Probabilities
Figure 4 shows the impulse response test of the MSIH(3)-VAR(1) model of the BCI and BIST100. The
responses of the BIST100 to the shocks on the BCI in different regimes are in the first column. In the first and
second regimes, the BCI shows weak increases when one standard deviation shock is applied to the BIST100.
In the third regime, BCI’s response is negatively weak.
The responses of the BCI to the shocks on the BIST100 in different regimes can be found in the
second column. In all regimes, the BCI increases in the first month when one standard deviation shock is
applied to the BIST100. Any shock on the BIST100 leaves a permanent effect on the BCI. The responses in the
first and third regimes are stronger (0.04, 0.03). The minimum response is shown in the second regime
(around 0.002).
0 25 50 75 100 125
0.02
0.04
0.06
Regime 1: cum. response orth. shock to L BCI
L BCI L BIST100RI
0 25 50 75 100 125
0.05
0.10
Regime 1: cum. response orth. shock to L BIST100RI
L BCI L BIST100RI
0 25 50 75 100 125
0.02
0.04
Regime 2: cum. response orth. shock to L BCI
L BCI L BIST100RI
0 25 50 75 100 125
0.025
0.050
Regime 2: cum. response orth. shock to L BIST100RI
L BCI L BIST100RI
0 25 50 75 100 125
0.00
0.05
0.10
Regime 3: cum. response orth. shock to L BCI
L BCI L BIST100RI
0 25 50 75 100 125
0.05
0.10 Regime 3: cum. response orth. shock to L BIST100RI
L BCI L BIST100RI
Figure 4: Impulse Response Tests
43
4.2 The Real Sector Confidence Index and The BIST100
The MSIH(3)-VAR(1) Model is the most successful among the nonlinear models in explaining the
relationship between the RSCI and the BIST100. If any additional lag of any variable is added, the Akaike's
(AIC), Hannan-Quinn (HQ), Schwarz (SIC) criterions increase and the LR linearity decreases, thus the power
of the model decreases.
Table 6: Information Criterions
Model log-
likelihood
AIC HQ SIC LR linearity
test
DAVIES
MSIH(3)-VAR(1) 390.8101 -6.5323 -6.2861 -5.9255 92.4167 0.0000
Table 7 shows the coefficients of the model. The constants in regime 1 (recession) are negative. If the
market is in recession, the RSCI and the return on the equity index should decline. The constants in regime 3
(expansion) are positive. If the market is in expansion, the consumer RSCI should rise and the return on
equity index should be positive.
Table 7: Coefficients
RSCI BIST100
Constant (Reg.1) -0.0428
-0.0781
Constant (Reg.2) -0.0012
0.0066
Constant (Reg.3) 0.0577
0.1367
BIST100_1 0.1099
0.1297
RSCI_1 -0.0585
-0.5520
SE (Reg. 1) 0.0883
0.0961
SE (Reg. 2) 0.0217
0.0765
SE (Reg. 3) 0.0934
0.0533
Table 8 shows the probabilities of regime transitions. If the market is in regime 1 in any given month,
the market should be 75.66 % in regime 1, 1.89 % in regime 2, and 22.45 % in regime 3 in the following
month. If the market is in regime 2, it should be 98.58 % in regime 2, 01.42 % in regime 1, and 0.00 % in
regime 3. If the market is in regime 3, it should be 58.95 % in regime 3, 01.47 % in regime 1, and 39.58 % in
regime 2.
Table 8: Matrix of Transition Probabilities
Regime 1 Regime 2 Regime 3
Regime 1 0.7566 0.0189 0.2245
Regime 2 0.0142 0.9858 1.1389e-009
Regime 3 0.0147 0.3958 0.5895
Table 9 shows the transition possibilities of the model. The highest number of observations (99) and
the highest probability (91.49 %) belong to the moderate growth regime in this period. Moreover, at 71
months, it also has the longest duration. If the market is in moderate growth, it is expected to last 71 months.
The minimum number of observations (5) belongs to the high growth regime. The probability and duration of
the high growth regime are also at the minimum.
Table 9: Regime Probabilities
Number of Obs. Probability Duration
Regime 1 8 0.0550 4
Regime 2 99 0.9149 71
Regime 3 5 0.0301 2.5
44
2008 2009 2010 2011 2012 2013 2014 2015 2016
-0.2
0.0
0.2
MSIH(3)-VAR(1), 2007 (3) - 2016 (6)
DL BIST100RI DL RSCI
2008 2009 2010 2011 2012 2013 2014 2015 2016
0.5
1.0 Probabilities of Regime 1
filtered
predicted
smoothed
2008 2009 2010 2011 2012 2013 2014 2015 2016
0.5
1.0 Probabilities of Regime 2
filtered
predicted
smoothed
2008 2009 2010 2011 2012 2013 2014 2015 2016
0.5
1.0 Probabilities of Regime 3
filtered
predicted
smoothed
Figure 5: Regime Probabilities
0 25 50 75 100 125
0.06
0.08
0.10 Regime 1: cum. response orth. shock to L BIST100RI
L BIST100RI L RSCI
0 25 50 75 100 125
0.00
0.05
Regime 1: cum. response orth. shock to L RSCI
L BIST100RI L RSCI
0 25 50 75 100 125
0.025
0.050
0.075
Regime 2: cum. response orth. shock to L BIST100RI
L BIST100RI L RSCI
0 25 50 75 100 125
-0.01
0.00
0.01
0.02
Regime 2: cum. response orth. shock to L RSCI
L BIST100RI L RSCI
0 25 50 75 100 125
0.025
0.050
0.075
0.100 Regime 3: cum. response orth. shock to L BIST100RI
L BIST100RI L RSCI
0 25 50 75 100 125
-0.001
0.000
0.001
0.002 Regime 3: cum. response orth. shock to L RSCI
L BIST100RI L RSCI
Figure 6: Impulse Response Tests
Figure 6 shows the impulse response test of the MSIH(3)-VAR(1) model of the RSCI and the BIST100.
The responses of the RSCI to the shocks on the BIST100 in different regimes are in the first column. RSCI
increases in the first month in all regimes when one standard deviation shock is applied to the BIST100. Any
shock on the BIST100 leaves a permanent effect on the RSCI. The highest response is shown in the third
45
regime (over 0.08) and the minimum response is shown in the second regime (over 0.012). The responses of
BIST100 to the shocks of RSCI in different regimes are in the second column. RSCI decreases in the first
month in all regimes when one standard deviation shock is applied to BIST100. The highest response is
shown in the first regime (around -0.04) and the minimum response is shown in the third regime (around -
0.001).
4.1 The TUIK Confidence Index and the BIST100
We find various significant nonlinear models between the TUIKCI and the BIST100. Although they
are econometrically significant, they are not theoretically, economically or financially significant. The
MSIH(3)-VAR(1) Model with 3 regimes and 1 lag is summarized below as an example.
Table 10: Information Criterions
Model log-
likelihood
AIC HQ SIC LR linearity test DAVIES
MSIH(3)-VAR(1) 67.9339 -
6.1044 -5.8582 -5.4976 67.9336 0.000
Table 11 shows the coefficients of the model. The coefficients are explainable. The constants are
negative in regime 1 and positive in regime 3.
Table 11: Coefficients
TUIK CI BIST100
Constant (Reg.1) -0.0018
-0.0155
Constant (Reg.2) -0.0111
0.0236
Constant (Reg.3) 0.0515
0.0800
TUIK CI_1 0.0447
-0.1713
BIST100_1 0.1520
0.0184
SE (Reg. 1) 0.0376
0.0935
SE (Reg. 2) 0.0252
0.0505
SE (Reg. 3) 0.0626
0.0553
Table 12 shows the probabilities of regime transitions. If the market is in regime 1 in any given
month, the market is expected to be 88.78 % in regime 1, 00.27 % in regime 2, and 10.95 % in regime 3 in the
following month. If the market is in regime 2, the market is expected to be 85.78 % in regime 2, 14.22 % in
regime 1, and 0.0000 % in regime 3 in the following month. If the market is in regime 3, it is expected to be
12.04 % in regime 3, 01.01 % in regime 1, and 86.95 % in regime 2 in the following month.
Table 12: Matrix of Transition Probabilities
Regime 1 Regime 2 Regime 3
Regime 1 0.8878 0.0027 0.1095
Regime 2 0.1422 0.8578 3.27e-006
Regime 3 0.0101 0.8695 0.1204
Table 13 shows the transition possibilities for the all model. The highest number of observations (58) and the
highest probability (52.50 %) belong to the moderate growth regime in this period. Moreover, it has a long
duration of 9 months. The minimum number of observations (7), probability (06.54 %) and duration (1)
belong to the high growth regime.
Table 13: Regime Probabilities
Number of Obs. Probability Duration
Regime 1 58 0.5250 9
Regime 2 47 0.4096 7
Regime 3 7 0.0654 1
46
2008 2009 2010 2011 2012 2013 2014 2015 2016
-0.2
0.0
0.2
MSIH(3)-VAR(1), 2007 (4) - 2016 (6)
DLTUIK CI DL BIST100RI
2008 2009 2010 2011 2012 2013 2014 2015 2016
0.5
1.0 Probabilities of Regime 1
filtered
predicted
smoothed
2008 2009 2010 2011 2012 2013 2014 2015 2016
0.5
1.0 Probabilities of Regime 2
filtered
predicted
smoothed
2008 2009 2010 2011 2012 2013 2014 2015 2016
0.5
1.0 Probabilities of Regime 3
filtered
predicted
smoothed
Figure 7: Regime Probabilities
The transition possibility from regime 2 to regime 3 (Table 12) and the dates of the regime
possibilities (Figure 7) are not theoretically, economically, or financially significant.
5. Conclusions
Confidence is crucial topic in financial discussions. Similarly, consumer confidence has been popular
topic in finance since the 2008 global crisis. This paper adds to the literature of studies that focus primarily
on confidence index and stock returns.
The financial markets experienced with fast and frequent fluctuations after the year 2000. These
fluctuations seen in the recession and expansion regimes are captured successfully in non-linear models. In
particular, the Markov Regime Switching Models was found successfully capture these regimes.
We observe that there are various models that explain the mutual nonlinear switching mechanism
between the consumer confidence indices and equity index. Of these models, the Markov Regime Switching
Model is the most appropriate to explain the relationship between these variables in different regimes.
Specifically, the MSIH(3)-VAR(1) Model with 3 regimes and 1 lag is the best model when taking into account
all variable duals among the different nonlinear models. This model contains three regimes; recession,
moderate growth, and expansion. Moreover, it successfully explains the relationships econometrically.
Additionally, the model clarifies the relationships for the BCI – the BIST100 and the RSCI – the BIST100.
However, the model cannot define the relationship between the TUIKCI and the BIST100 return index
economically or financially.
Even if consumer indices point out negative signalrecessions, capital market still gains in normal
periods of economy. If they only in a recession or an expansion regime do, each of the indices moves in the
same direction. If the market is in recession (regime 1), they decline. But if the market is in expanding, they
rise. Moreover, the impulse response functions of the MSIH(3)-VAR(1) models identify the relationships
between the different regimes. The estimated regime-dependent impulse response functions support
theoretical predictions. Both the BCI and RSCI are affected significantly by the shocks on the BIST100 in three
of the regimes. The effects of the shocks of consumer indices on the BIST100 are not strong as the effects of
BIST100’ shocks on consumer indices. These findings support previous studies by Topuz (2011) and Kale and
Akkaya (2015).
References
1. Baker M, Wurgler J (2007), Investor Sentiment in the Stock Market, Journal of Economic Perspectives, 21(2), 129-152. doi:
10.1257/jep.21.2.129.
2. Batchelor R (2001), Confidence indexes and the probability of recession: a Markov switching model. Indian Economic Review, 36(1),
107-124.
3. Bildirici M E, Aykaç Alp E, Ersin Ö Ö , Bozoklu Ü (2010), İktisatta Kullanılan Doğrusal Olmayan Zaman Serisi Yöntemleri, İstanbul:
Türkmen Kitabevi.
47
4. Bremmer D (2008), Consumer confidence and stock prices, Paper presented at the 72nd Annual Meeting of the Midwest Economics
Association Hyatt Regency, Chicago, Illinois, www. rosehulman.edu/~ bremmer/professional/bremmer_midwest2008.pdf/ Accessed
13.02.09.
5. Chen S S (2011), Lack of consumer confidence and stock returns, Journal of Empirical Finance, 18(2), 225–236.
6. Chung, S L, Hung C H, Yeh C Y (2012), When does investor sentiment predict stock returns? Journal of Empirical Finance, 19 (2), 217–
240.
7. Çevik E İ, Korkmaz T, Atukeren, E (2012), Business confidence and stock returns in the USA: a time-varying Markov regime-switching
model, Applied Financial Economics, 22(4), 299–312.
8. Guidolin, M. and Timmermann A (2006), An econometric model of nonlinear dynamics in the joint distribution of stock and bond
returns, Journal of Applied Econometrics, 21(1) 1-22.
9. Guidolin M (2016), M. Modelling, Estimating and Forecasting Financial Data under Regime (Markov) Switching,
http://didattica.unibocconi.it/mypage/dwload.php?nomefile=Lecture_7_-_Markov_Switching_Models20130520235704.pdf, accessed
30.03.16.
10. Fisher K L, Statman M (2003), Consumer confidence and stock returns, The Journal of Portfolio Management, 30(1), 115-127. doi:
10.3905/jpm.2003.319925.
11. Hamilton J D (1989), A new approach to the economic analysis of nonstationary time series and the business cycle, Econometrica:
Journal of the Econometric Society, 57(2), 357-384.
12. Howrey E P (2001), The predictive power of the index of consumer sentiment, Brookings Papers on Economic Activity, 32 (2001–1),
175–216.
13. Kahneman D, Tversky A (1979), Prospect Theory: An Analysis of Decision under Risk, Econometrica, 47(2), 263-291. doi:
10.2307/1914185.
14. Kale S, Akkaya M (2015), The Relation between Confidence Climate and Stock Returns: The Case of Turkey, Istanbul Conference of
Economics and Finance, ICEF 2015, 22-23 October 2015, Istanbul, Turkey.
15. Kandır S Y (2006), Tüketici Güveni ve Hisse Senedi Getirileri İlişkisi: İMKB Mali Sektör Şirketleri Üzerinde Bir Uygulama, Ç.Ü. Sosyal
Bilimler Enstitüsü Dergisi, 15(2), 217-230.
16. Katona G (1968), Consumer Behavior: Theory and Findings On Expectations and Aspirations, The American Economic Review, 58(2),
19-30.
17. Kilic E, Cankaya S (2016), Consumer confidence and economic activity: a factor augmented VAR approach, Applied Economics, 48(32),
3062–3080.
18. Korkmaz T, Çevik E İ (2009), Reel Kesim Güven Endeksi ile İMKB 100 Endeksi Arasındaki Dinamik Nedensellik İlişkisi, İstanbul
Üniversitesi İşletme Fakültesi Dergisi, 38(1), 24-37.
19. Krolzig H M (1998), Econometric modelling of Markov-switching vector autoregressions using MSVAR for Ox.
20. Krolzig H M (2000), Predicting Markov-Switching Vector Autoregressive Processes, Working Paper 2000W31Oxford University.
21. Lemmon M, Portniaguina E (2006), Consumer confidence and asset prices: some empirical evidence, Review of Financial Studies, 19 (4),
1499–1529.
22. Ludvigson S C (2004), Consumer confidence and consumer spending, Journal of Economic Perspective, 18 (2), 29–50.
23. Otoo M W (1999), Consumer sentiment and the stock market: Board of Governors of the Federal Reserve System,
https://www.federalreserve.gov/pubs/feds/1999/199960/199960pap.pdf/ accessed 16.09.16.
24. Schaller H, Norden S V (1997), Regime switching in stock market returns, Applied Financial Economics, 7(2), 177- 191.
25. Topuz Y V (2011), Tüketici Güveni ve Hisse Senedi Fiyatları Arasındaki Nedensellik İlişkisi: Türkiye Örneği, Ekonomik ve Sosyal
Araştırmalar Dergisi, 7(1), 53-65.