\documentclass{crorr}
%%%%% journal  info -- DO NOT CHANGE %%%%%%%%%
\setcounter{page}{25}
\renewcommand\thisnumber{1}
\renewcommand\thisyear {2014}
\renewcommand\thismonth{xxx}
\renewcommand\thisvolume{5}


\usepackage[T1]{fontenc}
\usepackage{multirow}

\usepackage[displaymath,mathlines,pagewise]{lineno}
\begin{document}
%\begin{linenumbers}


\markboth{Tihana \v Skrinjari\'c and Bo\v sko \v Sego}{Temporary and permanent stock price innovations}
\title[ESTIMATING TEMPORARY AND PERMANENT STOCK PRICE INNOVATIONS ON CROATIAN CAPITAL MARKET]{ESTIMATING TEMPORARY AND PERMANENT STOCK PRICE
INNOVATIONS ON CROATIAN CAPITAL MARKET}

\author[Tihana \v Skrinjari\'c and Bo\v sko \v Sego]{\bf Tihana \v Skrinjari\'c\affil{1}\comma\corrauth  \,and   Bo\v sko \v Sego\affil{2}}

\address{\affilnum{1}\ Department of Mathematics, Faculty of Economics and Business, University of Zagreb\\
Trg J.\,F.\,Kennedyja 6, HR-10\,000 Zagreb, Croatia\\
E-mail: $\langle$tskrinjaric@efzg.hr$\rangle$
\\
\medskip
\affilnum{2}\ Department of Mathematics, Faculty of Economics and Business, University of Zagreb\\
Trg J.\,F.\,Kennedyja 6, HR-10\,000 Zagreb, Croatia\\
E-mail: $\langle$bsego@efzg.hr$\rangle$
}


\begin{abstract}
This paper evaluates the size and duration of temporary and permanent stock price innovations on
Croatian capital market in the structural VAR (vector autoregression) framework with Blanchard and
Quah (1989) decomposition. The purpose is to identify the effects of temporary price innovations in
order to determine to which extent future stock prices can be predicted. Temporary components
present in stock prices are explained throughout the mean-reversion hypothesis. This means that stock
prices deviate from the fundamental values, but they will revert to their mean. In that way, to some
extent, it is possible to predict future price movements. The results show that for the observed
period from January 2000 to September 2013, temporary innovations account for only 2.62\% of price
variability over a two-year horizon. This means that forecasting the future movements of stock prices
on Zagreb Stock Exchange is a difficult task.
\end{abstract}

\keywords{temporary and permanent stock innovations, stock market, structural VAR, random walk hypothesis}

\maketitle


\section{Introduction}

Standard present value models of stock evaluation assume that investors buy stocks in order
to hold them for a certain amount of time. In the holding period the investor realizes
dividends, thus the stock price itself is a present value of future cash flows. This relationship
describes the fundamental factors behind stock price movements and it is related to the
Efficient Market Hypothesis (EMH). If all of the information about the future is already
incorporated in stock prices, we cannot predict what will happen with the prices in the future.
The random walk theory compliments the EMH and suggests that the random walk of price
movements generates permanent innovations in stock price movements, while temporary
innovations are a result of a stationary component of prices. Nevertheless, there has been a debate on the legitimacy of the EMH
and the random walk theory. Authors have started
to recognize temporary and permanent stock price innovations and question them for the last
thirty years, both in academic and investors' circles. Questions emerged whether only
fundamental factors affect stock price movements.
The first papers to appear were Shiller \cite{Shiller}, Fama \cite{Fama} and Summers \cite{Summers}. Shiller \cite{Shiller}
criticized the EMH, by concluding that movements in stock prices are somewhat caused by changes in expected real interest rates.
Fama \cite{Fama} argued that stock returns should be negatively correlated with expected inflation, which is approximated by a
short-term interest
rate. Summers \cite{Summers} introduced temporary stock price innovations in order to explain
deviations between the variance of stock prices and the variance of dividend streams. He
concluded that stock prices take temporary swings from fundamental values. Temporary
components present in stock prices are explained throughout the mean-reversion hypothesis.
This means that stock prices deviate from the fundamental values, but they will revert to their
mean. In that way, to some extant, it is possible to predict future price movements. Some
explanations about this phenomenon are given in Shiller \cite{Shiller2}, Blanchard and Watson \cite{Blanchard2}, etc.
If the evidence exists that temporary innovations (the mean-reverting component) make a
large part of price movements, stock returns are predictable and thus the Efficient Market
Hypothesis does not hold.
Fama and French \cite{Fama2} explain that temporary shocks uncorrelated with shocks to rational
forecasts of dividends have no effect on expected dividends and no long-term effects on
expected stock prices. They suggested taking caution in treating stock prices or stock returns
as rational reflections of fundamental values, especially when implementing macroeconomic
theories such as the \textit{q} investment theory (which assumes that asset prices reflect the present
value of future asset rents). Those first findings were followed by studies of Lo and
MacKinlay \cite{Lo}, Kim et. al. \cite{Kim}, Gallagher and Taylor \cite{Gallahager}, Hess and Lee \cite{Hess}, etc. The
mentioned authors found that temporary stock price innovations exist and affect stock price
movements. The question remains how big the size and the dynamics of these effects are,
which gives us answers to which extent it is possible to predict future movements of stock
prices or returns.
A lot of studies examining the Croatian stock market have emerged in the last couple of
years (Vizek and Dadi\'c \cite{Vizek}, Barbi\'c and \v Condi\'c-Jurki\'c \cite{Barbic2}, Benazi\'c \cite{Benazic}, Hsing
\cite{Hsing}, Barbi\'c
\cite{Barbic}, Dadi\'c \cite{Dadic}, etc.). None of them have analyzed and separated stock innovations into
temporary and permanent factors. The purpose of this paper is to analyze co-movements
between stock market returns and interest rates on the Croatian stock market in order to
identify whether temporary stock price innovations exist and in which way they affect stock
prices by the following explanations given in Shively [18:499]: ``real interest rates are
intrinsically related to real stock prices through standard present-value models and they
should be highly informative to identifying the existence, size and dynamic effect of
independent temporary and permanent stock price innovations''. In that way we hope to
contribute to the existing literature by examining the size and duration of temporary stock
price innovations. Since the theory suggests that the mentioned innovations are temporary
swings from the fundamental values, which have no effect on stock prices in the long run, a
structural VAR (vector autoregression) model with defined long-run zero restriction in order
to identify two structural shocks (temporary and permanent) is applied. This is called the
Blanchard and Quah \cite{Blanchard2} VAR framework.
Finally, Mart\'{i}nez-Moya et al. \cite{Martinez-Moya} explain why interest rate fluctuations have a significant
influence on stock price fluctuations: an increase in interest rates leads to an increase in
interest expense of firms, which reduces cash flows from future dividends, thus negatively
impacting stock prices. Furthermore, fluctuations in interest rates affect the opportunity cost
of assets, and they are also linked to real activity in the economy, which affects the
expectations of future cash flows. In that way, this paper estimates the size and duration of
two structural shocks (temporary and permanent) in Croatian stock movements, by examining
the relationship between real stock returns and real interest rates. The structure of the paper is
as follows. The second section describes the methodology applied in this study, while the
third section deals with data description and results of the empirical analysis. The final, fourth
section concludes the paper.


\section{Blanchard and Quah (1989) framework}

Blanchard and Quah's \cite{Blanchard} approach to modeling structural VARs is as follows. A structural
moving average model of variables of interest is written as:
\begin{equation}
 \boldsymbol{x}_t=\boldsymbol{A}(\boldsymbol{L})\boldsymbol{\epsilon}_t,
\end{equation}
i.e.
\begin{equation}
 \begin{bmatrix}
 sr_t\\
\Delta ir_t
 \end{bmatrix}=
\begin{bmatrix}
A_{11}(L) & A_{12}(L)\\
A_{21}(L) & A_{22}(L)
\end{bmatrix}
\begin{bmatrix}
\epsilon_t^T\\
\epsilon_t^P
\end{bmatrix},
\end{equation}
where $\boldsymbol{x}_t=\begin{bmatrix}sr_t & \Delta ir_t\end{bmatrix}'$ is the vector of stationary endogenous variables, $sr_t$ is
real stock
return and $\Delta ir_t$ is the differenced real interest rate. $\boldsymbol{\epsilon}_t=\begin{bmatrix}\epsilon_t^T &
\epsilon_t^P\end{bmatrix}'$ is a vector of structural disturbances, $\epsilon_t^T$ and $\epsilon_t^P$ are temporary and
permanent stock price innovation, respectively. $\boldsymbol{A}(\boldsymbol{L})=\boldsymbol{A}_0 + \boldsymbol{A}_1L+\boldsymbol{A}_2L^2+\ldots$ is
a matrix polynomial of the lag operator $L$. $A_{ij}(L), i, j\in\{1, 2\}$ is a polynomial in the lag
operator $L$. Moreover, structural disturbances have the following properties:
\begin{equation}
 E\left(\boldsymbol{\epsilon}_t\right)=\boldsymbol{0},\
E\left(\boldsymbol{\epsilon}_t\boldsymbol{\epsilon}_t'\right)=\boldsymbol{I}\ \text{and}\
E\left(\boldsymbol{\epsilon}_t\boldsymbol{\epsilon}_s'\right)=\boldsymbol{0}\ \forall\ t\neq s,
\end{equation}
i.e. structural disturbances are by definition serially uncorrelated and mutually orthogonal,
with a normalized variance matrix.

The vector moving average representation of the VAR model is defined as:
\begin{equation}
 \boldsymbol{x}_t=\Theta(\boldsymbol{L})\boldsymbol{e}_t,
\end{equation}
where $\Theta(\boldsymbol{L})=e_t+\Theta_1e_{t-1}+\Theta_2e_{t-2}+\ldots$ and
$\boldsymbol{e}_t=\begin{bmatrix}\boldsymbol{e}_t^T & \boldsymbol{e}^P\end{bmatrix}'$ is a
vector of reduced form disturbances. Properties of reduced form disturbances are:
\begin{equation}
 E\left(\boldsymbol{e}_t\right)=\boldsymbol{0},\
E\left(\boldsymbol{e}_t\boldsymbol{e}_t'\right)=\Omega\ \text{and}\
E\left(\boldsymbol{e}_t\boldsymbol{e}_s'\right)=\boldsymbol{0}\ \forall\ t\neq s.
\end{equation}
By comparing equations (1) and (4) and taking into account matrix polynomials $\boldsymbol{A}(\boldsymbol{L})$ and $\Theta
(\boldsymbol{L})$ it follows that
\begin{equation}
 \boldsymbol{e}_t=\boldsymbol{A}_0\boldsymbol{\epsilon}_t.
\end{equation}
In order to calculate $\boldsymbol{A}(\boldsymbol{L})$, we need an estimate of $\boldsymbol{A}_0$, which can be obtained by
taking the variance of each side of equation (6):
\begin{equation}
 \Omega=\boldsymbol{A}_0\boldsymbol{A}_0'.
\end{equation}
Since the model given in (4) is under-identified, additional restrictions need to be made in order
to obtain estimates of $\boldsymbol{A}_0$. $\boldsymbol{A}_0$ is a 2\textperiodcentered2 matrix. Thus we need four parameters to
recover structural disturbances. Three are given by estimating $\hat{\Omega}$. This means one additional
restriction is needed to identify the system exactly. Here, it is assumed that, according to the
mentioned literature, temporary stock price shocks do not influence stock returns in the long
run. Formally, the restriction is given as
\[
A_{11}(L)=\sum\limits_{i=1}^\infty a_{ii}(i)=0 \mbox{ in (2)}.
\]


\section{Data and empirical analysis}

This section deals with estimation of a two-variable VAR model, with a vector of endogenous
variables $\boldsymbol{x}_t=\begin{bmatrix}sr_t & \Delta ir_t\end{bmatrix}'$, where $\Delta$ denotes the difference operator. Data on the
Croatian stock
market index, CROBEX, and 91 day Treasury bill interest rates were gathered from Zagreb
Stock Exchange \cite{Zagreb Stock Exchange} and from the Ministry of Finance \cite{Ministry of Finance}. It refers to monthly data
from
January 2000 to September 2013\footnote{This was the longest time span available.}. In order to apply the Blanchard and Quah
technique, at
least one of the variables must be non-stationary since stationary variables do not have a
permanent component. In order to use the method, all of the variables must be stationary in
the VAR model. The CROBEX variable is in natural logarithm, while the interest rates were
adjusted by a consumer price index in order to reflect real interest rates. Figure 1 represents
movements of the two variables.

\begin{figure}[!h]
\centering
\includegraphics[width=\textwidth]{Fig1.eps}
\caption{\it CROBEX and real interest rates in the period from January 2000 to September 2013}
\begin{flushleft}Source: Authors.\end{flushleft}
\end{figure}

To determine the order of integration, an Augmented Dickey-Fuller (ADF) test was
performed on the variables in levels and differences. The lag length was chosen based on the
Schwartz Information Criterion. Two versions of the ADF test were performed on the variables in
levels: one including both a constant and a trend as deterministic components, and one only
with the constant included. Two versions of the test were performed on the differenced
variables as well: one including the constant in the ADF equation, and the other with none of
the deterministic components. Table 1 shows the result test values on the 5\% level of statistical
significance, values in round brackets give information about the chosen lag length and values
in square brackets give information about the \textit{p}-value of each test value. Critical values are
used as referred to in MacKinnon (1996).

\begin{table}[!ht]
\centering
\begin{tabular}{|c|c|c|}\hline
\multirow{2}{*}{\textbf{Variable}}  & \multicolumn{2}{|c|}{\textbf{Deterministic components}}\\
\cline{2-3}
& \textbf{Constant, trend} & \textbf{Constant} \\
\hline
\textbf{Ln(crobex)} & -3.4373 (1) [0.9116] & -2.8789 (1) [0.3386]\\
\hline
\textbf{Real interest rate} & -3.4376 (1) [0.1401] & -2.8792 (1) [0.0361]\\
\hline
 & \textbf{Constant} & \textbf{None} \\
\hline
\textbf{$\Delta$ Ln(crobex)} & -2.8789 (0) [0.0000] & -1.9428 (0) [0.0000]\\
\hline
\textbf{$\Delta$ Real interest rate} & -2.8792 (0) [0.0000] & -1.9434 (0) [0.0000]\\
\hline
\end{tabular}
\caption{\it ADF Unit root tests results}
\begin{flushleft}Source: Authors' calculation.\end{flushleft}
\end{table}

As can be seen, all of the variables are $I(1)$, i.e. integrated of order one, because they have
unit root in levels and are stationary in the first differences. Since the variables are $I(1)$,
there exists a possibility of cointegration between them. If they are cointegrated, a structural VECM
(vector error correction model) has to be applied instead of a structural VAR. In order to
check for a possible cointegration, a Johansen Cointegration test was applied over variables in
levels. Table 2 reports the results of the analysis. Trace and Maximum Eigenvalue tests
suggest that there is no cointegration among variables. This means that it is appropriate to use
the first differences of the variables in the SVAR model.

\begin{table}[!ht]
\centering
{\small
\begin{tabular}{|c|c|c|c|c|c|}\hline
\multirow{2}{*}{\textbf{Test}} & \multirow{2}{*}{$\boldsymbol{H_0}$} & \multirow{2}{*}{\textbf{Eigenvalue}} &
\multirow{2}{*}{\textbf{Test Statistic}} & \textbf{5\% Critical} & \multirow{2}{*}{\textbf{\textit{p}-value}}\\
 & & & & \textbf{Value} & \\
\hline
\multirow{2}{*}{\textbf{Trace}}  & $r=0$\textasteriskcentered & 0.0616 & 13.3936 & 15.4947 & 0.1011\\
\cline{2-6}
& $r\leq 1$ & 0.0189 & 3.0875 & 3.8415 & 0.0789\\
\hline
\textbf{Maximum}& $r=0$\textasteriskcentered  & 0.0616 & 10.3061 & 14.2645 & 0.1926\\
\cline{2-6}
 \textbf{Eigenvalue}& $r\leq1$ & 0.0189 & 3.0875 & 3.8415 & 0.0789\\
\hline
\end{tabular}
}
{\footnotesize Note: * denotes rejection of the null hypothesis at the 0.05 level.}
\caption{\it Results of Johansen Cointegration test for a number of cointegration vectors}
\begin{flushleft}
Source: Authors' calculation.
\end{flushleft}

\end{table}

In order to estimate a VAR model, a lag length needs to be determined. Several criteria have
been used: Final prediction error (FPE), Hannan-Quinn (HQ) and Akaike (AIC) information
criteria have pointed out that a lag length of $k = 1$ is appropriate. Thus, a VAR(1) model was
estimated with intercept included. Another model was estimated with a dummy
variable included accounting for the crisis period, but the results were almost identical to the results of
the initial model. Henceforth, the given results in the rest of this paper refer to the initial
model. The final formal requirement for the use of the Blanchard and Quah framework is that the estimated VAR model is stable. All of
the roots of the model lie within the unit circle, thus
the stability of the model is fulfilled. Now, a structural VAR model can be estimated by
setting the long term influence of temporary stock price innovations to zero.

Since we are interested in the effects of temporary and permanent innovations on stock
returns, an impulse response function was constructed based on the estimated model. Figure 2
shows an impulse response function for stock returns. As can be seen, temporary stock
innovations have a positive effect in the first month, followed by a smaller negative response
in the second, which almost vanishes in the third month. From the fourth month, stock returns
return to their average value. Permanent innovations, on the other hand, have negative effects
on stock market returns and their effects vanish after the third month. Thus, it can be
concluded that temporary shocks have little and short-life effects on stock return movements.
In that way, it is very difficult to predict future movements of return.

\begin{figure}[!h]
\centering
\includegraphics[scale=0.7]{Fig2.eps}
\caption{\it Impulse response function of stock returns to temporary and permanent shocks}
\begin{flushleft}
Source: Authors.
\end{flushleft}
\end{figure}

Furthermore, a variance decomposition of stock returns has been made in order to measure the
importance of temporary innovations in stock price fluctuations. Table 3 shows the results of
the analysis. Temporary price innovations have a marginal impact on stock price movements.
In the first month they account for 1.89\% of price volatility, in the second 2.49\% and up to
2.62\% in other horizons. Therefore, the size of the mean-reverting component is very small,
which means it is difficult to forecast future price movements on the Croatian capital market. On
the other hand, fundamental innovations account for over 97\% of stock price movements.

\begin{table}[!ht]
\centering
\begin{tabular}{|c|c|c|}\hline
\multirow{2}{*}{\textbf{Month}} & \multicolumn{2}{|c|}{\textbf{Type of innovation}}\\
\cline{2-3}
 & \textbf{Temporary} & \textbf{Permanent}\\
\hline
1 & 1.888664 & 98.11134\\
\hline
2 & 2.487404 & 97.51260\\
\hline
3 & 2.608685 & 97.39131\\
\hline
4 & 2.622117 & 97.37788\\
\hline
5 & 2.623420 & 97.37658\\
\hline
6 & 2.623542 & 97.37646\\
\hline
12 & 2.623555 & 97.37645\\
\hline
18 & 2.623555 & 97.37645\\
\hline
24 & 2.623555 & 97.37645\\
\hline
\end{tabular}
\caption{\it Variance decomposition of DLNCROBEX}
\begin{flushleft}
Source: Authors' calculation.
\end{flushleft}
\end{table}

The results are not very surprising. If we examine the autocorrelation and partial
autocorrelation function of returns of CROBEX given in Figure 3, the correlograms suggest
that the returns are a white noise process. All of the estimated coefficients of autocorrelation
and partial autocorrelation\footnote{Except for the coefficients on the sixth lag.} lie within the interval of $\pm$ 2 S.E., i. e.
they are not statistically significant.

\begin{figure}[!h]
\centering
\includegraphics[width=.47\textwidth]{Fig3.eps}
\hspace*{3mm}
\includegraphics[width=.47\textwidth]{Fig4.eps}
\caption{Autocorrelation and partial autocorrelation function of stock returns}
\begin{flushleft}
Note: The gray area on the correlogram refers to the 95\% confidence bands of \linebreak $\pm 2$ S.E.\\
Source: Authors.
\end{flushleft}
\end{figure}

\section{Discussion and conclusion}

Random walks, Efficient Market Hypothesis and stock price predictability have been ongoing
issues in academic and investors' circles for decades. Studies which distinguish
temporary and permanent innovations in stock price index movements in the 1980s started to appear.
Authors started to question the randomness of price movements. Some evidence started to
appear that stock prices or returns can be predicted, at least to some extent.

Temporary components present in stock prices are explained throughout the mean-reversion
hypothesis. This means that stock prices deviate from the fundamental values, but they will
revert to their mean. The bigger the impact of temporary components on stock price index movements, the bigger the
probability to accurately predict future price/return movements.
The theory suggests that temporary innovations have effects on stock prices in the short run,
but in the long run these effects vanish, and permanent innovations have effects
which are present both in the short and long run. Therefore, this paper employs a structural VAR model with Blanchard and Quah [4]
decomposition. This framework estimates
structural innovations by forcing a long run restriction on effects of temporary innovations.

The results of the analysis suggest that for the examined period from January 2000 to
September 2013 temporary innovations have positive but small effects on stock price
index movements. They account for only 2.62\% of price variability over a two year horizon. In
that way, it is not an easy task to predict future price movements on the Croatian stock
market. This paper is a first attempt in Croatia to decompose stock price index movements to
temporary and permanent innovations. Thus, more work needs to be done. Other theories on
stock price movements and their components exist, which might describe the movements on the
Croatian capital market more accurately. Future work includes decomposition stock price
movements by using other relevant theories of stock evaluation and comparison of the results
with previous findings in the literature.

\begin{thebibliography}{99}

\bibitem{Barbic}
Barbi\'c, T. (2010). Testing week form of efficient market hypothesis in Croatia's equity
market. The Proceedings of Zagreb Faculty of Economics and Business, Year 8, No. 1, 2010, 155--172.

\bibitem{Barbic2}
Barbi\'c, T., \v Condi\'c-Jurki\'c, I. (2011). Relationship between macroeconomic fundamentals
and stock market indices in selected CEE countries. Economic Review, 62, 113--133.

\bibitem{Benazic}
Benazi\'c, M. (2008). The relationship between stock prices and exchange rate in the
Republic of Croatia: VEC model. Economic Review, 59, 669--687.

\bibitem{Blanchard}
Blanchard. O. J., Quah, D. (1989). The dynamic effects of aggregate demand and supply
disturbances. American Economic Review, 79, 665--673.

\bibitem{Blanchard2}
Blanchard. O. J., Watson. M. W. (1982). Bubbles, rational expectations and financial
markets, in Crises in the Economic and Financial Structure: Bubbles, Bursts and Shocks
(ed. P. Wachtel). Lexington Books. Lexington. MA: 295--315

\bibitem{Dadic}
Dadi\'c, T. (2009). Testing equity market efficiency in selected CEE countries. Master
thesis. Zagreb: Faculty of Economics and Business, University of Zagreb.

\bibitem{Fama}
Fama, E. F. (1981). Stock returns, real activity, inflation and money. American Economic
Review, 71, 545--564.

\bibitem{Fama2}
Fama, E. F., French, K. R. (1988). Permanent and temporary components of stock prices.
Journal of Political Economy, 96, 241--265.

\bibitem{Gallahager}
Gallahager, L. A., Taylor, M. P. (2000). Measuring the temporary component of stock
prices: robust multivariate analysis. Economics Letters, 67, 193--200.

\bibitem{Hess}
Hess, P. J., Lee. B. (1999). Stock returns and inflation with supply and demand
disturbances. The Review of Financial Studies, 12, 1203--1218.

\bibitem{Hsing}
Hsing, Y. (2011). Macroeconomic variables and the stock market: the case of Croatia.
Ekonomska istra\v zivanja, 24, 41--50.

\bibitem{Kim}
Kim, M. J., Nelson, C. R., Startz, R. (1991). Mean reversion in stock prices? A
reappraisal of the empirical evidence. Review of Economic Studies, 58, 515--28.

\bibitem{Lo}
Lo, A. Q., MacKinlay, A. C. (1988). Stock market prices do not follow random walks:
evidence from a simple specification test. The Review of Financial Studies, 1, 41--66.

\bibitem{Martinez-Moya}
Mart\'inez-Moya, P., Ferrer-Lape\~na, R., Escribano-Sotos, F. (2013). Relationship
between interest rate changes and stock returns in Spain: A wavelet-based approach.
UCLM Working papers, Deparment of Economics and Finance, ULCM, Spain.

\bibitem{Ministry of Finance}
Ministry of Finance, \verb"http://www.mfin.hr/hr/mjesecni-statisticki-prikazi" [Accessed
on 1 November 2013].

\bibitem{Shiller}
Shiller, R. J. (1981). Do stock prices move too much to be justified by subsequent
dividends? American Economic Review, 71, 421--436.

\bibitem{Shiller2}
Shiller, R. J. (1984). Stock prices and social dynamics. Brookings Papers on Economic
Activity, 2, 457--510.

\bibitem{Shively}
Shively, P. A. (2003). International evidence of temporary and permanent stock-price
innovations: a multivariate approach. Applied Economics Letters, 10, 499--503.

\bibitem{Summers}
Summers, L. H. (1986). Does the stock market rationally reflect fundamental values?
Journal of Finance, 43, 591--601.

\bibitem{Vizek}
Vizek, M., Dadi\'c, T. (2006). Integration of Croatian, CEE and EU equity markets:
cointegration approach. Economic Review, 57, 631--646.

\bibitem{Zagreb Stock Exchange}
Zagreb Stock Exchange, \verb"http://zse.hr/default.aspx?id=44101\&index=CROBEX" [Accessed on 1 November 2013].

\end{thebibliography}
%\end{linenumbers}
\end{document}