© 2016 Nicolaus Copernicus University. All rights reserved. http://www.dem.umk.pl/dem D Y N A M I C E C O N O M E T R I C M O D E L S DOI: http://dx.doi.org/10.12775/DEM.2016.006 Vol. 16 (2016) 87−116 Submitted November 30, 2016 ISSN (online) 2450-7067 Accepted December 18, 2016 ISSN (print) 1234-3862 Aneta Włodarczyk, Iwona Otola * Analysis of the Relationship between Market Volatility and Firms Volatility on the Polish Capital Market A b s t r a c t. In this paper we investigate if the strength of firm-market volatility relationship has changed after subprime crisis on the Polish Capital Market. The empirical study concern the selected companies listed on the Warsaw Stock Exchange (WSE) from the construction and IT sectors in the 2004–2011 period. The volatility measures were computed on the basis of daily low and high prices for companies shares and WIG index. For each company ARFIMAX-FIGARCH model with additional exogenous variables, which represented market volatility, was estimated in the stable and the turbulent period. Conducted empirical studies have not shown that the negative shocks flowing from the American stock market through investors' behavior channel contributed to the increase in the fraction of firms of the construc- tion and IT sectors listed on the WSE whose volatility is shaped by market volatility. K e y w o r d s: ARFIMAX-FIGARCH, firm volatility, market volatility, subprime crisis, Warsaw Stock Exchange. J E L Classification: G12; D40; C58. Introduction Volatility of prices of listed companies shares is the subject of interest not only to stock exchange investors but also companies which emit them. High volatility means high probability of change of company market value in the future and this in turn influences its competitiveness on the market. This is * Correspondence to: Aneta Włodarczyk, Czestochowa University of Technology, Faculty of Management, 36b Armii Krajowej Street, 42-200 Częstochowa, Poland, e-mail: aneta.w@interia.pl; Iwona Otola, Czestochowa University of Technology, Faculty of Man- agement, 19b Armii Krajowej Street, 42-200 Częstochowa, Poland, iwotola@gmail.com. Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 88 connected with the fact that high market value of an enterprise in relation to its book value is connected with a simultaneous high market share of a given subject (Grabowska, 2013). Enterprise competitiveness depends, among others, on the value for the shareholders and clients, financial capability, determining its ability to act and react quickly in the competitive environ- ment and human potential and technologies used in implementing strategic changes (Feurer and Chaharbaghi, 1994). Moreover, the knowledge about linkages between market volatility and firm volatility is of great importance in risk management process and determining the investment portfolio struc- ture, because it helps to understand the nature of individual firm volatility and identify this process determinants. It is not surprising that there exist many articles that were devoted to this issue. We may summarize some im- portant facts about firm volatility on the basis of literature studies. As sug- gest Karolyi (2001) the existence of excessive volatility in stock prices un- dermines the usefulness of the information about the true value of the com- pany. This clearly does not mean that high volatility is proof of the irrational behavior and inefficient markets and investors. No signals from the market to volatility of stock prices, affirms the conviction that it is the correct meas- ure of risk. Interest in the subject matter of volatility, in particular, the possi- bility of its forecasting is related to the ability to reduce the risk of invest- ment or achieving higher returns from investments. In the literature, there are already several well-documented facts concerning volatility. Chen and Lai (2013) and other researchers (Whaley, 2009; Simon, 2003; Giot, 2005; Peng and Ng, 2012) showed a significant feature, which is the asymmetry of vola- tility. Shin and Stulz (2000) decomposed shares risk into its market and firm- specific components in order to show that changes in market risk are posi- tively correlated with changes in firm value, but changes in firm- specific risk are negatively correlated with changes in firm value. Campbell et al. (2001) used a disaggregated approach to study the volatility of common stocks at the market, industry and firm level. They proposed three variance components, which were estimated monthly using daily data, so their further analysis about volatility components were conducted on the basis of monthly frequency indicators and for this reason they could not consider the impact of volatility clustering effect on evaluation of risk measurement process. Phylatkis and Xia (2009) investigated the equity market comovement at the sector level and confirmed the sector heterogeneity of the contagion. They showed that investors might find the profitable sectors on the capital markets despite of the prevailing contagion on the market level. Chuliá and Torró (2011) estimated a conditional CAPM with multivariate GARCH-M struc- ture in order to investigate an asymmetric volatility spillover effect between Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 89 large and small firms in the Spanish stock market and they proved the exist- ence of bidirectional relationship between these volatility processes with asymmetric influence of bad and good news for firm volatility. Sharma et al. (2011, 2014) showed that firms belonging to different sectors experienced different degrees of association with market volatility, what caused that in some cases firm volatility might be predicted on the basis of market volatil- ity. They proved that commonality in volatility increases with the firm's size. Sequeira and Lan (2003) found that, the most significant component of ag- gregate volatility in the international market, with firm-level volatility form- ing the largest component of total volatility is unsystematic variance of the stock return. They also claim that volatility of the market-level is more im- portant than the volatility of the sector-level in the explanation of the total volatility of stock return, which is also confirmed in earlier research con- ducted by Cavaglia et al. (2000) or Griffin and Karolyi (1998). That is why both managerial staff in enterprises and investors should be interested in knowing the strength and direction of relationships between shares volatility of particular enterprises and stock market volatility. Previous studies on the volatility in the majority relate to market volatili- ty. A small part of the study is devoted to the volatility of the firm-level. Volatility of asset prices is also often explained by the mechanism of trans- mission of crises and the related phenomenon of contagion (Baur, 2003; Corsetti et al., 2005; Claessens and Forbes, 2004; Le and David, 2014). Re- searchers examining the contagion effect on the capital market level concern only a chosen stock indices, which do not reflect the impact of the volatility of share prices of companies in different sectors. In such an approach the heterogeneity effect of the sectors selected within a given capital market and their various immunity to external shocks and different sensitivity to effects of a financial crisis are omitted (Phylatkis and Xia, 2009). The impact of the subprime crisis on individual segments of the Polish fi- nancial market were presented by Konopczak et. al. (2010). The analysis shows the local and global conditions that made the interaction of disorders associated with the global financial crisis on the Polish financial market was as strong as in developed markets. Empirical studies, including the period of the subprime crisis, conducted by Otola (2013), concerning the relationship between the WIG index and the S&P500 index, had to indicate whether neg- ative shocks arising in the United States have been transferred to Polish Cap- ital Market by stock exchange channel. The high volatility in the capital market, which has been observed in an unstable period is not only the result of the interdependence between the markets but the result of the financial contagion of Polish Capital Market. In this context, we rightly seem to con- Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 90 duct the further research on the volatility of the Polish Capital Market on the firm level. Taking above into consideration we are interested in examining the direc- tion and strength of relationship between firm-level volatility and market- level volatility on the Polish Capital Market in the period of pre- and post- subprime crisis. The aim of this paper is to investigating if the market vola- tility and firm volatility are related and if the strength of this relationship increase significantly in the turbulent period compared to the stable period. The empirical study concerns the selected companies listed on the Warsaw Stock Exchange from the construction and IT sectors in the period from 2 January 2004 to 31 December 2011. In this respect, our contribution to the literature is connected with the calibration of Sharma et al. model (2011, 2014) through introducing long memory dynamics and skewed fat-tailed distributions of innovations into the basic model describing the relationship between firm volatility and market volatility. Following Sadigue and Silvapulle (2001) and Kang and Yoon (2012) we expect that on the Polish capital market some investors reacted to new information immediately, whereas others postpone making the investment decision until they confirm the information. It causes that their actions form a non-linear pattern, which characterizes statistical persistence in a time series. We also taking into con- sideration the possible structural changes that may have affected the long memory property, so we divide the analyzed period into two sub-periods, taking the subprime financial crisis of July 2007 as the structural change point. The above considerations allow us to formulate the following hypoth- esis: There is no significant increase in the fraction of firms from a given sector of Polish Capital Market whose volatility is strongly and positively related with market volatility in pre-crisis period compared with post-crisis period. The remainder of the paper is organized as follows. Section 2 describes the methodology, data properties are presented in section 3. In Section 4 we verify the hypothesis about the commonality in volatility in stable and turbu- lent period. And finally, Section 5 concludes the paper. 2. Methodology We have modified the model of Sharma et al. (2011, 2014), which de- scribes the relationship between firm volatility and stock market volatility by capturing the possible long memory effect in both firm and market volatility process. Theoretical frames of the volatility model determines Capital Asset Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 91 Pricing Model (CAPM) and Fractionally Integrated Generalized Autoregres- sive Conditional Heteroscedasticity (FIGARCH). In empirical research presented in the paper the authors use the following regression equation, in which firm volatility is conditioned by previous and present market volatility and also by present return on investment into the market portfolio and squared return of market index (Sharma et al., 2011): , 2 ,4,31,2,10, ittmitmitmitmiitf RRVVV i    (1) where: Vm – stock market volatility, if V – share volatility of i-nth company (i=1, 2, ..,N), Rm,t – logarithmic return of stock market index. Sharma et al. (2011) incorporated market return and squared market return factors into above equation in order to control for possible spurious depend- ence between returns and volatility measures. Price range estimators are used as volatility estimators that are computed on the basis of full publicly available set of information about financial in- strument prices what make them more efficient than volatility estimators calculated as the squared return (Fiszeder and Perczak, 2013; Molnár, 2012; Sharma et al., 2014):  daily volatility estimator created by Parkinson (1980): ,)]ln()[ln(361.01 2 itit PLPHV  (2) where: PHit – highest price of i-nth financial instrument on t day, PLit – lowest price of i-nth financial instrument on t day;  daily volatility estimator proposed by Garman and Klass (1980): ,)]ln()][ln(12ln2[)]ln()[ln(5.02 22 itititit POPCPLPHV  (3) where: PCit – closing price of i-nth financial instrument on t day, POit – opening price of i-nth financial instrument on t day;  daily volatility estimator defined by Rogers and Satchell (1991): )].ln())][ln(ln()[ln( )]ln())][ln(ln()[ln(3 itititit itititit PCPLPOPL PCPHPOPHV   (4) Originally, the Parkinson, Garman-Klass, Rogers-Satchell volatility measures of financial instruments prices are the estimators of uncondition- ally volatility for the geometric Brownian motion determined for time inter- val [0, T]. Therefore, all of analysed here estimators are derived under the assumptions of continuous sampling, no bid-ask spread and constant volatil- ity. Moreover, in the case of the Parkinson and Garman-Klass estimators an additional assumption refers to zero drift. It is worth attracting attention on Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 92 some properties of range-based volatility estimators, especially in the case when one day is used as a unit of time. It was proved that the Garman-Klass estimator characterized by the highest efficiency in the case when the as- sumption of zero drift was met. The efficiency of range- based volatility estimators for zero drift are as follow: the Parkinson – 4.9, the Garman-Klass – 7.4, the Rogers-Satchell – 6.0 and they are significantly higher than the efficiency of simply volatility estimator (based on squared daily returns) equals by definition 1 (Molnár, 2012). For daily financial time series, mean return is often much smaller than its standard deviation, which is in line with the assumption of zero drift. Fiszeder and Perczak (2013) derived the ex- pected values of the Parkinson and Garman-Klass estimators for the arithme- tic Brownian motion with non-zero drift and they proved the unbiasedness of the Parkinson and Garmann-Klass estimators for the process with a zero drift and of the Rogers-Satchell estimator for any drift. It is worth stressing that the square root of any range-based volatility estimators is a biased estimator of standard deviation of a Brownian motion, but this bias is rather small (3– 4%) compared to bias of the square root of the simply volatility estimator (25%). The questionable issue is the use of the range-based volatility estima- tors for the standardization of the returns. It was shown that returns normal- ized by means of the Garman-Klass estimator were approximately normally distributed. Another disadvantage concerning the use of range estimators in practice is failure to comply with the assumptions of continuous sampling. Prices are observed at discrete points in time (vector of POt, PHt, PLt, PCt for each day) and it causes that the observed high price is below the true high price and observed low price is above the true low price. Bid-spread effect works in the opposite direction. These two effects are small for liquid stocks (Molnár, 2012). A debatable question is also the adjustment of range-based volatility estimators for the opening jumps, resulting from the situation that most of the assets are not traded continuously for 24h a day (Molnár, 2012; Fiszeder and Perczak, 2013). Sharma et al. (2014) limited their empirical studies to modelling the firm volatility shocks ( error term) by using GARCH(1,1) model with the con- ditional normal distribution of innovations. In the above equation the authors used full available on a given day information on the market portfolio (open- ing price, the lowest and the highest price, closing price), which will allow us to eliminate the effect of influencing the research results with one type of data only – closing price. For each analyzed series of share prices volatility of the given enterprises the ARFIMA(P,D,Q)-FIGARCH(p,d,q) models were chosen individually in accordance with the following stages: Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 93  specification of the equation describing the relation between market volatility and firm volatility (1), adjusted for possible autocorrelation de- pendencies occurring in the company share volatility series;  specification of the conditional variance equation describing the ARCH effect in the series of residuals from the equation (1) and the selection of the form of innovation density function. The fractionally integrated autoregressive moving average process (ARFIMA) is more flexible econometric tool for modelling the conditional mean process than ARMA model in the situation when it exhibits long memory properties. Similarly, the fractionallly integrated generalized auto- regressive conditional heteroskedasticity process, constituting a generaliza- tion of the GARCH model, enables for capturing the persistence in the con- ditional variance process (Bollerslev and Mikkelsen, 1996). However, mar- ket shocks have a simultaneous impact on the conditional mean and condi- tional variance. Therefore, some recent empirical studies have focused on the analyzing the relationship between the conditional mean and the condi- tional variance of the process that simultaneously exhibits long memory properties (Beine et al., 2002; Fiszeder, 2009; Kang and Yoon, 2012). The ARFIMA(P,D,Q)-FIGARCH(p,d,q) model is defined in the following way (Arouri et al., 2012): , )](1[)1)(( )()()1)(( 2 2 ttt tt d ttt ttt D h LLL h LyLL         (5) where: L denotes a lag operator;               00 )()1( )( )1()1( j jjj j D L Dj Dj L j D L – differential filter of D order, where, )( denotes gamma function; ,...1)( 1 P P LLL   Q Q LLL   ...1)( 1 are respectively lags polynomials for the autoregressive part of P order and moving average of Q order, whose all roots lie outside the unit circle, i, i are the model parame- ters;               00 )()1( )( )1()1( s sss s d L ds ds L s d L – differential filter of d order; ,...1)( 1 q q LLL   p p LLL   ...1)( 1 are respectively Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 94 lags polynomials for the ARCH part of q order and GARCH part of p order, all the roots of )](1[),( LL   lie outside the unit circle, αi, βi are the model parameters; t  is an innovations series, ).1,0(..~ diit On the basis of D parameter value one can identify the memory type of the process (Hosking,1981):  means that the process exhibits negative dependencies between distant observations (anti-persistence, but stationary process),  corresponds to the long-memory stationary process,  means reduction of the ARFIMA(P,D,Q) to the stationary ARMA(P,Q) model (short memory process),  indicates a non-stationary process (in particular for the process follows a unit root process – ARIMA(P,1,Q) model). For the FIGARCH model the influence of current shocks for volatility fore- casts decreases to zero, but at a slower rate than for GARCH processes. Moreover, the autocorrelation function of FIGARCH squared residuals de- creases at a hyperbolic rate to zero, which indicates long memory in volatil- ity of series described by this class models. The existence of long memory in volatility process may be recognized on the basis of estimated value of d parameter (Fiszeder, 2009):  means that volatility process exhibits long memory property,  means reduction of the FIGARCH(p,d,q) to the GARCH(p,q) process for which the influence of current volatility on forecasts of con- ditional variance decays at fast rate (also short memory process in sense of “influence of current volatility for the true conditional variance proc- ess”),  means reduction of the FIGARCH(p,d,q) to the IGARCH(p,q) process for which current volatility has permanent impact on forecasts of conditional variance (short memory in above explained sense). This class of models is estimated by using the quasi-maximum likelihood (QML) estimation method, based on the following log-likelihood function (under the assumption about Gaussian distribution of innovations) (Fiszeder, 2009; Kang and Yoon, 2012): ,])[ln( 2 1 )2ln( 2 1 1 2    T t ttGaussian hTLL  (6) where T denotes the number of observation. It is worth stressing that one of stylized facts of high – frequency time series of financial prices is excess kurtosis and skewness of returns distributions (Fiszeder, 2009; Laurent, 2013; Włodarczyk, 2010). In order to capture the Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 95 excess kurtosis and fat-tailed effect in the residuals, the Student t-distribution may be used as the innovation distribution in ARFIMA-FIGARCH model, and then the log-likelihood function is defined as follows (Kang and Yoon, 2012; Laurent, 2013 ): , 2 1ln)1()ln( 2 1 )]2(ln[ 2 1 2 ln 2 1 ln 1 2                                                  T t t t Student h TLL      (7) where Parameter measures the degree of fat tails of the density function and the lower its values is, the fatter tails of distribution are. The latest specification of ARFIMA-FIGARCH models, which be considered in this work, allows for both the excess skewness and kurtosis of residuals by using the skewed Student t-distribution. In this case the log-likelihood function is as follows (Kang and Yoon, 2012; Laurent, 2013 ):   , 2 )( 1ln)1()ln( 2 1 ln 1 2 ln)]2(ln[ 2 1 2 ln 2 1 ln 1 2 2                                                                        T t IStS t SentSkewedStud tkh k k TLL      (8) where: , / if 1 / if 1 t t       SS SS tI   (9) is the indicator function, which is determined on the basis of the mean (S) and standard deviation (σS) of the skewed Student t-distribution (Kang and Yoon, 2012; Laurent, 2013 ): , 1 2 2 2 1                        k kS      (10) Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 96 .1 1 2 2 22 SS k k         (11) The value of the asymmetry parameter ln(k) determined the type of skew- ness, that is, if 0)l n ( k , the density is right skewed and if 0)ln( k , the density is left skewed. 3. Data Description Empirical research were conducted for the chosen companies from con- struction and IT sector listed on the main market of the Warsaw Stock Ex- change whereas restrictive provisions which these companies had to meet in order to be able to exist on it. To the research 17 companies from the con- struction sector were chosen in the stable period and during a turbulent three more. The IT sector study involved 10 companies during the stable period and 20 in a turbulent period. The authors excluded from the research sample enterprises which in the given sub-period withdrew from the market or their debuts took place in the course of the given sub-period. Data were obtained from the Notoria Service and consist of 2013 daily observations of high, low, open and close prices for analyzed companies from January 2, 2004 to December 30, 2011. The starting point is determining the breakthrough on the basis of which it is possible to distinguish two sub-periods: the stable period which corre- sponds to low stock market volatility and the turbulent period characterized by high market volatility. First symptoms of growing volatility of main in- dexes of American market (DJIA, NASDAQ, S&P500) can be observed in July and August of 2007. They are accompanied by bankruptcy of two hedg- ing funds of Bear Stearns bank connected with the mortgage market in the USA (July 2007), as well as the insolvency of three funds of the French bank BNP Paribas investing in the bonds market secured with mortgages sub- prime (9 August 2007) (Otola, 2013; Burzała, 2013; Dungay et al., 2011). The analyzed period was divided into two sub-samples:  the stable period (02.01.2004–25.07.2007) – 899 observations;  the turbulent period (26.07.2007–31.12.2011) – 1114 observations. Due to the prolonging period of instability on the European financial market, caused not only by the occurrence of the subprime crisis, but also the debt crisis in the peripheral countries of the EU zone, the turbulent period in the present analysis was lengthened to the end of the year 2011. Selection of sectors was conducted on the following basis. The construc- tion sector is most connected with the country's economy and most suscepti- Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 97 ble to GDP changes. It is also most strongly represented sector on the Polish stock market. Moreover, since 2004 Poland as a member of the European Union has been receiving financial resources within European funds to im- prove the living conditions of inhabitants. In the course of the second pro- gramming period (2007–2013) Poland received about 242,3 billion PLN, the part of which was destined for public infrastructural investments, both in the scope of road construction as well as other public facilities. About 48% of allocated within structural funds resources were used to execute the Infra- structure and Environment Programme. Additionally, the decision of UEFA in the II quarter of 2007 to give the right to organize the European Football Championship – Euro 2012 in Poland and Ukraine was also connected with the necessity to complete numerous construction investments in short time. According to the Central Statistical Office since the year 2005 the construc- tion sector was the fastest growing sector of Polish economy. Moreover, this sector is the most abundantly represented sector on the Warsaw Stock Ex- change. We claim that there is no significant increase in the fraction of firms, belonging to the construction sector, whose volatility is shaped by market volatility in the turbulent period compared to the stable period. Thus, one can expect that external shocks transmitted on the stock market in Poland in the subprime crisis period and the debt crisis in the Euro zone should not be noticeable by the enterprises of this sector. Claessens and Forbes (2004) emphasised the importance of investor reactions in the contagion process in the liquidity risk context. Referring to this theory one can expect that inves- tors having in mind good development perspectives resulting from a substan- tial financing of this sector with structural funds should not withdraw capital located in the stocks of construction companies. The second select sector, namely IT, may be considered as the one based on knowledge, most intellectual and innovative one. The IT sector may be divided into three groups due to the form of conducted activity:  production of IT equipment (among others Comp, Elzab, Novitus);  software related services (among others Asseco Poland, Comarch, Syg- nity);  distribution of IT solutions (among others Arcus, B3System, NNT Sys- tem). Comparing it with the construction sector one should notice that in the analyzed period a lot of enterprises from the IT sector put on acquiring capi- tal to conduct activity through emission of shares (see table 1). The resources acquired from the shareholders for development of this sector are used not only on tangible investments but also investments in intellectual capital. Moreover, development of new and growth of awareness of needs in the Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 98 scope of IT also in the sector of Small and Medium Enterprises, the use of structural funds to liquidate disproportions in access to tele-informatic tech- nologies in rural areas are only some of the factors conditioning continuity of operations of this sector companies. The IT sector is perceived in the world as one of the most perspective ones, the one which creates new inno- vative products, such as for example mobile product applications or devel- opment of cloud services. Customers of products and services of the IT sec- tor are primarily enterprises from the public administration sector that to a large extent use for this purpose the resources from the European funds. More important customers of services and products from the IT sector are enterprises operating in the banking and tele-communication sectors. Such a perception of the IT sector by the investors should contribute to the fact that the enterprises listed on the stock market should show resistance to shocks coming from the market. Thus, in our opinion commonality in vola- tility did not increase significantly in the turbulent period compared to the stable period in the IT sector on the Polish Capital Market. Table 1. Descriptive statistics for volatility measures – stable period Variable Minimum Mean Maximum Standard deviation Skewness Kurtosis V1_WIG 0.0004 0.008 0.157 0.012 5.577 [0.000] 47.622 [0.000] V2_WIG 0.0003 0.008 0.195 0.011 7.348 [0.000] 85.857 [0.000] V3_WIG 0 0.008 0.269 0.013 9.764 [0.000] 147.92 [0.000] V1_Construction 0.013 0.136 1.303 0.130 3.042 [0.000] 14.837 [0.000] V2_Construction 0.011 0.131 1.786 0.146 4.248 [0.000] 28.550 [0.000] V3_Construction 0.006 0.158 3.506 0.241 5.418 [0.000] 47.534 [0.000] V1_IT 0.009 0.116 1.046 0.119 2.759 [0.000] 10.424 [0.000] V2_IT 0.010 0.084 1.397 0.081 6.675 [0.000] 78.417 [0.000] V3_IT 0.011 0.092 2.179 0.103 10.018 [0.000] 165.52 [0.000] Note: all volatility series were scaled by 100, p-value in brackets. On the basis of daily information on four price categories (open, close, high and low) of the WIG stock index and shares of particular companies the authors determined volatility series in the stable and turbulent period accord- ing to the price range volatility measures (2)–(4). For each sector the authors estimated series of average firm volatility in each of the analyzed sub- Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 99 periods and then on their basis they determined descriptive statistics (Tables 1–2). Table 2. Descriptive statistics for volatility measures – turbulent period Variable Minimum Mean Maximum Standard deviation Skewness Kurtosis V1_WIG 0.0002 0.015 0.373 0.028 6.125 [0.000] 53.180 [0.000] V2_WIG 0.0002 0.013 0.305 0.027 6.340 [0.000] 50.279 [0.000] V3_WIG 0 0.013 0.357 0.029 6.651 [0.000] 55.075 [0.000] V1_Construction 0.012 0.205 2.481 0.351 2.176 [0.000] 3.837 [0.000] V2_Construction 0.010 0.120 3.394 0.195 7.707 [0.000] 103.24 [0.000] V3_Construction 0.010 0.100 3.623 0.253 10.149 [0.000] 112.55 [0.000] V1_IT 0.061 0.289 1.224 0.170 1.861 [0.000] 4.802 [0.000] V2_IT 0.011 0.092 0.571 0.075 2.919 [0.000] 11.721 [0.000] V3_IT 0.010 0.104 0.697 0.084 2.626 [0.000] 9.510 [0.000] Note: all volatility series were scaled by 100, p-value in brackets. One can observe that the average volatility in the IT sector was relatively lower than in the construction sector for all three measures. A similar de- pendence can be observed for the maximum value of each volatility meas- ures evaluated for the firms in both sectors. It is worth emphasizing that in the stable period market volatility was substantially lower than volatility of firms in both sectors, comparing volatility range determined in accordance with the three analyzed measures. Also determined skewness and kurtosis indicate sector differentiation of the volatility process, which may indicate heterogeneity of the construction and IT sectors on the Warsaw Stock Ex- change. Similar conclusions were formulated for the American stock ex- change in the paper of Sharma, Narayan and Zheng (2014). The determined average statistics for the series of firm and market volatility took higher values in the turbulent period in comparison with the stable pe- riod. Also in this period market volatility was substantially lower than the firm volatility in both sectors. Due to high kurtosis and skewness of the vola- tility series determined in accordance with the Rogers and Satchel formula only volatility calculated according to the Parkinson method has been in- cluded in the further part of the analysis. Shaping of daily volatility for the Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 100 Figure 1. Shaping of daily volatility for the companies from the construction sector in the stable period 1 Figure 2. Shaping of daily volatility for the companies from the construction sector in the turbulent period 2 1 Explanations of used abbreviations for listed companies can be found in Table 5. 2 Explanations of used abbreviations for listed companies can be found in Table 6. Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 101 Figure 3. Shaping of daily volatility for the companies from the IT sector in the stable period 3 Figure 4. Shaping of daily volatility for the companies from the IT sector in the turbulent period 4 3 Explanations of used abbreviations for listed companies can be found in Table 7. 4 Explanations of used abbreviations for listed companies can be found in Table 8. SGN 2008 2010 2 4 SGN CTC 2008 2010 2 4 CTC ELZ 2008 2010 2.5 7.5 ELZ ACP 2008 2010 2.5 7.5 ACP CMR 2008 2010 2 4 CMR MCL 2008 2010 2 4 MCL TLX 2008 2010 2 4 TLX SME 2008 2010 2 4 SME WAS 2008 2010 2 4 WAS CDR 2008 2010 2 4 CDR BCM 2008 2010 2 4 BCM ATM 2008 2010 2 4 ATM CMP 2008 2010 2 4 CMP PCG 2008 2010 5 15 PCG QSM 2008 2010 2 4 QSM U2K 2008 2010 2 4 U2K LSI 2008 2010 2 4 LSI IMX 2008 2010 2 4 IMX PRD 2008 2010 2 4 PRD NTT 2008 2010 2 4 NTT Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 102 companies from the construction and IT sectors in the stable and turbulent periods show Figures 1–4. In the stable period the biggest range of daily share price volatility fell for Budopol SA company. High daily spread of volatility was also characteristic for such companies as: Awbud SA, Energopol Południe SA, Ulma Construc- cion SA. In the turbulent period two companies Elkop SA and Erbud SA were characterized by very high daily volatility spread. In the stable period the highest spread in daily volatility was observed for the companies: Simple SA, Wasko SA and Calatrava Capital SA. In the turbu- lent period a high daily spread of volatility was observed in the companies: ZUK Elzab, Asseco Poland SA, CD Project Red SA, Calatrava Capital SA, Simple SA, PC Guard SA, LSI Software SA, NTT System SA. The presented above descriptive statistics for enterprises listed on the Warsaw Stock Exchange from the construction and IT sectors let us notice differences both between sectors as well as the stable and turbulent periods. 4. Verifying the Hypothesis about the Commonality in Volatility in Stable and Turbulent Period Time series volatility V1 of share prices of enterprises from the construc- tion and IT sectors in the stable period i.e. 02.01.2004.–25.07.2007 and the turbulent one 26.07.2007–31.12.2011 were subject to modelling in order to verify the occurrence of significant relationships between firm volatility and WIG index volatility. Due to the characteristic properties of volatility time series, which were the subject of modelling: skewness, high kurtosis, fat- tailed distribution, autocorrelation dependencies (Box-Pierce test), occur- rence of long memory effect (GPH test), occurrence of volatility clustering effect (Engle test, McLeod and Li test) (see Table 3), for each volatility se- ries individual specification of the model in the form (5)–(8) were selected with t-Student distribution or skewed t-Student distribution (Laurent, 2013). Additionally, in the conditional mean equation is assumed that the firm volatility is dependent on previous, present and future market volatility and also expected daily change of the WIG index and the squared WIG index return. The choice of the final form of the firm volatility model has been performed on the basis of :  evaluating significance of the model parameters estimations;  minimizing information criteria of Akaike’a (AIC) and Bayesian Schwarz (SC); Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 103  test results of standardized models residuals which verified the correct- ness of the model dynamic specification: Box-Pierce test, Engle test, McLeod and Li test;  Nyblom test for checking the constancy of model parameters over time;  Engle and Ng set of tests for existing leverage effect and investigating possible misspecification of the conditional variance equation (FI- GARCH vs. FIEPARCH), with in particularly covered Negative Sign Bias test and Positive Sign Bias test for verifying the effect of negative and positive shocks on the conditional variance (Laurent, 2013). Table 3. Descriptive statistics and diagnostic tests for volatility series of MSW in stable period Parameter/Test Value of statistics Mean 0.868 Standard deviation 1.410 Minimum 0 Maximum 14.537 Skewness 5.239 [0.000] Kurtosis 38.493 [0.000] Jarque-Bera statistics 59616 [0.000] ADF statistics –6.588 D-GPH statistics 0.259 [0.000] Q Box-Pierce statistics Q(5) 71.1948 [0.000] Q(10) 120.783 [0.000] Q(20) 145.085 [0.000] Q(50) 172.831 [0.000] Q McLeod-Li statistics Q(5) 117.096 [0.000] Q(10) 182.624 [0.000] Q(20) 204.178 [0.000] Q(50) 214.615 [0.000] Engle LM statistics ARCH(1–1) 46.606 [0.000] ARCH(1–2) 25.877 [0.000] ARCH(1–5) 23.281 [0.000] ARCH(1–10) 15.902 [0.000] ARCH(1–20) 9.590 [0.000] Note: p-value in brackets. Below we present the detailed results of the estimation of ARFIMAX- FIGARCH model with t-Student innovations for the chosen enterprise from the construction sector – Mostostal Warszawa SA (MSW; see Table 4a–4b). Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 104 Table 4a. Estimation results of ARMAX(1,2)-GARCH(1,1) for MSW in stable period Parameter Parameter estimates Cst(M) 0.223 [0.000] Vm,t_WIG 0.369 [0.374] AR(1) 0.858 [0.000] MA(1) –0.712 [0.000] MA(2) –0.049 [0.044] Cst(V) 0.221 [0.025] ARCH1 0.165 [0.003] GARCH1 0.757 [0.000] Df-Student 2.258 [0.000] AIC 2.029 SC 2.072 Note: p-value in brackets. Table 4b. Diagnostic residuals tests for MSW Test Estimates of the test statistics Q Box-Pierce statistics for standardized residuals Q(5) 8.183 [0.037] Q(10) 11.812 [0.107] Q(20) 19.714 [0.289] Q(50) 54.719 [0.205] Q Box-Pierce statistics for squared standardized residuals Q(5) 0.660 [0.883] Q(10) 0.916 [0.999] Q(20) 2.187 [0.999] Q(50) 3.857 [0.999] Engle LM statistics ARCH(1–2) 0.035 [0.965] ARCH(1–5) 0.128 [0.986] ARCH(1–10) 0.088 [0.999] Jarque-Bery for normality 27151 [0.000] Joint Nyblom test of stability 1.2021 Nyblom for Cst(M) 0.1501 Nyblom for Vm,t_WIG 0.0541 Nyblom for AR(1) 0.0577 Nyblom for MA(1) 0.0505 Nyblom for MA(2) 0.0803 Nyblom for Cst(V) 0.1432 Nyblom for ARCH1 0.2187 Nyblom for GARCH1 0.2149 Nyblom for Df-Student 0.0712 Sign Bias Test 0.899 [0.369] Negative Sign Bias Test 0.381 [0.703] Positive Sign Bias Test 0.275 [0.783] Jointly Engle–Ng test 2.227 [0.527] Note: p-value in brackets, p-value (5%) = 0.47 for individual Nyblom test. Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 105 It is worth emphasizing that for each estimated model for the Mostostal Warszawa company, the parameters indicating the strength and direction of dependencies between enterprise volatility and market volatility was statisti- cally insignificant. The results summarizing estimation of ARFIMAX-FIGARCH class models with included variables describing volatility of the stock market for particular companies from the construction sector in the stable period and turbulent period have been presented in Tables 5–6. The analysis of data included in Table 5 allows us to formulate the following conclusions. In case of Energopol-Południe and Budopol-Wrocław enter- prises the relationship between the market volatility and volatility of their share prices was statistically insignificant. For the remaining enterprises the relationship between the WIG index volatility and firms volatility was statis- tically significant and positive. All additional variables describing market volatility (Vm,t-1_WIG, Rm,t, Rm,t 2 ) turned out to be statistically insignificant in case of all estimated models. Table 5. Results on commonality in volatility for construction sector in stable period Company Model specification Estimate of Vm,t_WIG parameter Significance of the influence of others market variables (Vm,t-1_WIG, Rm,t, Rm,t2) on firm volatility Mostostal Export (MSX) ARMAX(1,1)- FIGARCH(1,0.54,1), t-Student distribution 2.069 [0.000] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Prochem (PRM) ARFIMAX(1,0.121,1)- GARCH(1, 1), t-Student distribution 0.390 [0.005] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Mostostal Zabrze Hol- ding (MSZ) ARFIMAX(1,0.049,1)- GARCH(1, 1), t-Student distribution 1.191 [0.013] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Budimex (BDX) ARFIMAX(1,0.021,1)- GARCH(1, 1), skewed t-Student distribution 0.090 [0.006] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Elektrobudo- wa (ELB) ARFIMAX(1,0.105,1)- GARCH(1, 1), t-Student distribution 0.677 [0.010] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Energoapara- tura (ENP) ARFIMAX(1,0.086,0)- ARCH(1), t-Student distribution 0.539 [0.022] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Energopol- Południe (EPL) ARFIMAX(1,0.185,1)- GARCH(1, 1), t-Student distribution 0.075 [0.410] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 106 Table 5. continued Company Model specification Estimate of Vm,t_WIG parameter Significance of the influence of others market variables (Vm,t-1_WIG, Rm,t, Rm,t2) on firm volatility Projprzem (PJP) ARFIMAX(1,0.027,0)- ARCH(1), t-Student distribution 0.167 [0.003] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Elkop (EKP) ARMAX(1,1)- GARCH(1, 1), t-Student distribution 0.023 [0.018] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Ulma Construccion Polska (ULM) ARFIMAX(0,0.06,1)- FIGARCH(1,0.57,1), t-Student distribution 0.041 [0.355] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Energomon- taż-Południe (EDP) ARFIMAX(1,0.042,1)- GARCH(1,1), t-Student distribution 0.815 [0.045] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Mostostal Płock (MSP) ARX(2)-GARCH(1,1), t-Student distribution 3.642 [0.000] Rm,t: 0.081 [0.027] Rm,t2: –0.069 [0.001] Budopol- Wrocław (BDL) ARFIMAX(0,0.06,1)- FIGARCH(1,0.79,1), t-Student distribution 0.024 [0.283] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Awbud (AWB) ARFIMAX(0,0.06,1)- GARCH(1,1), t-Student distribution 0.921 [0.005] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Pemug (PMG) ARX(2)-GARCH(1,1), t-Student distribution 0.748 [0.099] All additional market variables have statisti- cally insignificant impact on the firm volatili- ty – elimination from the model Instal Kraków (INK) ARMAX(1,1)- GARCH(1,1), t-Student distribution 2.618 [0.000] Rm,t2: –0.076 [0.001] Note: p-value in brackets. In the turbulent period for five enterprises: Awbud, Projprzem, Prochem, Elkop, Ulma Construccion the relationship between the WIG index volatility and firm volatility was statistically insignificant. For the remaining 15 enter- prises statistically significant influence of market volatility on firm volatility was confirmed, while the direction of this dependence was positive. Addi- tionally, share price volatility of Mostostal Warszawa was significantly shaped by market volatility observed on the previous day. In case of the remaining enterprises additional variables did not significantly influence share price volatility. The conducted commonality in volatility analysis for the enterprises listed on the Warsaw Stock Exchange shows that in the stable period 82% of analyzed enterprises were sensitive to the WIG index volatili- ty, the appropriate ratio in the turbulent period reached the value of 75%. Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 107 Table 6. Results on commonality in volatility for construction sector in turbulent period Company Model specification Estimate of Vm,t_WIG parameter Significance of the influence of others market variables (Vm,t-1_WIG, Rm,t, Rm,t2) on firm volatility Mostostal Export (MSX) ARFIMAX(1,0.07,1)- GARCH(1,1), t-Student distribution 0.949 [0.007] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Mostostal Warszawa (MSW) ARX(2)-GARCH(1,1), t-Student distribution 0.171 [0.006] Vm,t-1_WIG: 0.176 [0.040] Prochem (PRM) ARMAX(1,1)- FIGARCH(1,0.68,1), t-Student distribution 0.138 [0.483] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Budimex (BDX) ARFIMAX(1,0.103,1)- GARCH(1,1), t-Student distribution 0.206 [0.000] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Mostostal Zabrze Holding (MSZ) ARFIMAX(1,0.076,0) - GARCH(1,1), t-Student distribution 0.867 [0.000] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Elektrobudowa (ELB) ARFIMAX(1,0.121,1)- GARCH(1,1), t-Student distribution 0.122 [0.082] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Ulma Construccion Polska (ULM) ARFIMAX(1,0.041,1)- GARCH(1,1), t-Student distribution 0.227 [0.454] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Energoaparatura (ENP) ARFIMAX(1,0.111,0)- GARCH(1,1), t-Student distribution 0.508 [0.005] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Polimex- Mostostal (PXM) ARMAX(1,1)- GARCH(1,1), t-Student distribution 1.639 [0.000] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Mostostal Płock (MSP) ARMAX(1,1)- ARCH(1), t-Student distribution 0.401 [0.000] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Energopol- Południe (EPL) ARFIMAX(1,0.045,1)- GARCH(1,1), t-Student distribution 0.521 [0.012] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Instal Kraków (INK) ARFIMAX(0,0.12,1)- FIGARCH(1,0.75,1), t-Student distribution 0.541 [0.000] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Projprzem (PJP) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.065 [0.562] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Elkop (EKP) ARFIMAX(1,0.999,0)- ARCH(1), t-Student distribution 0.054 [0.189] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 108 Table 6. continued Company Model specification Estimate of Vm,t_WIG parameter Significance of the influence of others market variables (Vm,t-1_WIG, Rm,t, Rm,t2) on firm volatility PBG (PBG) ARFIMAX(1,0.236,1)- GARCH(1,1), t-Student distribution 1.041 [0.000] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Herkules (HRS) ARFIMAX(1,0.063,1)- GARCH(1,1), t-Student distribution 1.016 [0.000] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Erbud (ERB) ARFIMAX(1,0.04,1)- GARCH(1,1), t-Student distribution 0.269 [0.020] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Energomontaż- Południe (EDP) ARFIMAX(1,0.14,1)- GARCH(1,1), t-Student distribution 0.457 [0.009] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Budopol- Wrocław (BDL) ARFIMAX(1,0.087,1)- GARCH(1,1), t-Student distribution 0.972 [0.000] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Awbud (AWB) ARMAX(1,1)- GARCH(1,1), t-Student distribution –0.018 [0.735] All additional market variables have statistically insignificant impact on the firm volatility – elimination from the model Note: p-value in brackets. Estimation of volatility models for the IT sector enterprises in both ana- lyzed periods was conducted in the same manner as in case of the construc- tion sector enterprises (Tables 7–8). The estimated models indicate a statistically insignificant relationship be- tween market volatility and share price volatility of IT sector companies only in case of Macrologic and Talex. Moreover, included in the modelling pro- cess the additional variables describing market volatility, did not have a sta- tistically significant impact on shares volatility of the analyzed enterprises. In the turbulent period for the seven analyzed enterprises from the IT sector: Comarch, Talex, Simple, Wasko, PC Guard, Unima 2000, Infovide-Matrix the relationship between the WIG index volatility and share price volatility was statistically insignificant at 10% significance level. In addition, for three enterprises: Comarch, Procad and NTT System market volatility from the previous day significantly affected share price volatility of these enterprises. On the basis of empirical studies results for IT sector one may draw the con- clusion that ARFIMAX-FIGARCH specification has been chosen more fre- quently for the post-crisis period compared to pre-crisis period. Therefore, long memory property is not an inherent feature of firm volatility process and the structural change connected with the subprime crisis may influence Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 109 to the assessment of the long memory property in the volatility 5 . The con- ducted commonality in volatility analysis for the IT sector enterprises listed Table 7. Results on commonality in volatility for IT sector in stable period Company Model specification Estimate of Vm,t_WIG parameter Significance of the influence of others market variables (Vm,t-1_WIG, Rm,t, Rm,t2) on firm volatility Calatrava Capital (CTC) ARFIMAX(1,0.104,1)- GARCH(1,1), t-Student distribution 1.186 [0.002] All additional market variables have statisti- cally insignificant impact on the firm volatility – elimination from the model ZUK Elzab (ELZ) ARFIMAX(1,0.044,1)- GARCH(1,1), t-Student distribution 0.362 [0.030] All additional market variables have statisti- cally insignificant impact on the firm volatility – elimination from the model Asseco Po- land (ACP) ARFIMAX(1,0.108,1)- GARCH(1,1), t-Student distribution 0.536 [0.000] All additional market variables have statisti- cally insignificant impact on the firm volatility – elimination from the model Comarch (CMR) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.349 [0.003] All additional market variables have statisti- cally insignificant impact on the firm volatility – elimination from the model Macrologic (MCL) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.392 [0.133] All additional market variables have statisti- cally insignificant impact on the firm volatility – elimination from the model Talex (TLX) ARMAX(1,1)- GARCH(1,1) with t- Student distribution 0.362 [0.155] All additional market variables have statisti- cally insignificant impact on the firm volatility – elimination from the model Simple (SME) ARMAX(1,1)- GARCH(1,1), t-Student distribution 1.414 [0.061] All additional market variables have statisti- cally insignificant impact on the firm volatility – elimination from the model CD Project Red (CDR) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.949 [0.001] All additional market variables have statisti- cally insignificant impact on the firm volatility – elimination from the model Sygnity (SGN) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.349 [0.086] All additional market variables have statisti- cally insignificant impact on the firm volatility – elimination from the model Wasko (WAS) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.719 [0.056] All additional market variables have statisti- cally insignificant impact on the firm volatility – elimination from the model Note: p-value in brackets. 5 It is worth underlining that for both the construction and IT sectors there were no rea- sons for changing the basic specification of firm volatility models to Exponential GARCH model, which enables capturing the asymmetric effect of negative and positive shocks for conditional variance (results of Sign Bias tests). But for deepening the studies over the link- age between market volatility and firm volatility on the Polish Capital Market we will use the asymmetric multivariate GARCH structure in near future. Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 110 on the Warsaw Stock Exchange shows that in the stable period 80% of ana- lyzed enterprises were sensitive to the WIG index volatility, while in the turbulent period the fraction of these enterprises dropped to 65%. Table 8. Results on commonality in volatility for IT sector in turbulent period Company Model specification Estimate of Vm,t_WIG parameter Significance of the influence of others market variables (Vm,t-1_WIG, Rm,t, Rm,t2) on firm volatility Sygnity (SGN) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.633 [0.005] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model Calatrava Capital (CTC) ARFIMAX(1,0.049,1)- GARCH(1,1), t-Student distribution 1.401 [0.013] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model ZUK Elzab (ELZ) ARMAX(1,1)- FIGARCH(1,0.39,1), t-Student distribution 1.494 [0.036] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model Asseco Poland (ACP) ARFIMAX(1,0.64,1)- GARCH(1,1),skewed t-Student distribution 1.381 [0.000] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model Comarch (CMR) ARFIMAX(1,0.07,1)- FIGARCH(1,0.31,1), t-Student distribution 0.085 [0.410] Rm,t2: 0.011 [0.005] Macrologic (MCL) ARFIMAX(1,0.08,1)- FIGARCH(1,0.7,1), t-Student distribution 0.020 [0.000] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model Talex (TLX) ARFIMAX(2,0.044,0)- GARCH(1,1), t-Student distribution –0.008 [0.942] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model Simple (SME) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.365 [0.203] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model Wasko (WAS) ARMAX(1,1)- GARCH(1,1), skewed t-Student distribution 0.073 [0.326] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model CD Project Red (CDR) ARMAX(1,1)- GARCH(1,1), t-Student distribution 2.180 [0.000] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model Betacom (BCM) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.706 [0.000] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model ATM (ATM) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.554 [0.049] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model Comp (CMP) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.188 [0.080] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 111 Table 8. continued Company Model specification Estimate of Vm,t_WIG parameter Significance of the influence of others market variables (Vm,t-1_WIG, Rm,t, Rm,t2) on firm volatility PC Guard (PCG) ARFIMAX(1,0.002,1)- GARCH(1,1), t-Student distribution 0.003 [0.731] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model, problem with parameters’ stability Qumak- Sekom (QSM) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.271 [0.021] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model Unima 2000 SystemyTele- informatycz- ne (U2K) ARMAX(1,1)- GARCH(1,1), t-Student distribution –0.009 [0.661] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model LSI Software (LSI) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.807 [0.016] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model Infovide- Matrix (IMX) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.043 [0.568] All additional market variables have statis- tically insignificant impact on the firm volatility – elimination from the model Procad (PRD) ARMAX(1,1)- FIGARCH(1,0.29,1), t-Student distribution 0.896 [0.004] Vm,t-1_WIG : 0.644 [0.005] NTT System (NTT) ARMAX(1,1)- GARCH(1,1), t-Student distribution 0.733 [0.000] Vm,t-1_WIG : 0.539 [0.000] Note: p-value in brackets. In order to verify the hypothesis assuming that within a given sector the percentage of enterprises in the stable and turbulent period, for which a sig- nificant positive dependence between their share price volatility and stock market volatility is the same, we used a statistical test (Krysicki et al. 2012): .: : 211 210 ppH ppH   The test statistics under the null hypothesis has a standardized normal distri- bution: ),1,0(arcsin2arcsin2 21 21 2 2 1 1 N nn nn n k n k U              (12) where: p1 – fraction of the firms in the given sector for which the commonal- ity in volatility effect was confirmed in stable period, p2 – fraction of the Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 112 firms in the given sector for which the commonality in volatility effect was confirmed in turbulent period, and – sample proportions corresponding to p1 and p2. Table 9. Statistics for similarity of commonality in volatility effect in stable and turbulent period U statistic (1.95 – 5% critical value) Construction sector IT sector 0.546 0.875 At 5% significance level there is no reasons for rejecting the hypothesis that commonality in volatility didn’t differ significantly in the turbulent period compared to the stable period in construction sector as well as in IT sector on the Polish Capital Market (Table 9). The conducted research show that the fractions of enterprises, in the sta- ble and turbulent period, for which the occurrence of significant relation- ships between their share price volatility and the WIG index volatility was proved, did not significantly differ one from the other. This may confirm that analyzed sectors were characterized by resistance to external shocks ab- sorbed by Polish stock market through the investor behaviour channel. Conclusions The analysis of relationships between market volatility and volatility of en- terprises from the construction and IT sectors has shown that there is not an increase in the fractions of firms, for which their share volatility in a signifi- cant and positive way was connected with stock market volatility in the pe- riod corresponding to the subprime crisis and the debt crisis in the Euro zone. Moreover, one can observe a different relation of these sectors to ex- ternal shocks coming to Polish market, namely in the IT sector the fraction of enterprises sensible to the WIG index volatility decreased, while in the construction sector it increased. However, the changes were not statistically significant. The financial situation of the enterprises from these sectors was conditioned not only by unforeseen factors connected with reaction of inves- tors to the information about crisis, but first of all economic factors and skills of managers in the scope of adjusting enterprise strategies to the changeable conditions of external environment, making the use of the oppor- tunities and threats filter in building and maintaining enterprise competitive position. On the basis of conducted empirical studies (Otola, 2013) concern- ing financial condition of enterprises from the construction and IT sectors we can draw the following conclusions. Majority of the analyzed enterprises Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 113 from the construction sector did not possess financial liquidity which is the determinant of the ability to pay back their liabilities. Moreover, in case of over 65% of the examined enterprises their rates of debt exceeded 70%. In addition, majority of the discussed enterprises achieved low sales profitabil- ity already at the level of gross profit from sales or operational profit. There were such enterprises among all the examined ones which were character- ized by high negative sales profitability. The situation was different in the IT sector. This sector enterprises were characterized by average debt level and maintained financial liquidity on the proper level (in accordance with the assumed norms of indexes), and a part of them possessed surplus of turnover assets with reference to the current liabilities, which indicates the excess of liquidity. Sales profitability indexes in the stable period were on a higher level than in the turbulent period, in which they can be considered as low and unsatisfactory for investors, nonetheless they were better than the ones from the construction sector. References Arouri, M., Hammoudeh, S., Lahiani, A., Nguyen, D. (2012), Long Memory and Structural Breaks in Modeling the Return and Volatility Dynamics of Precious Metals, Quarterly Review of Economics and Finance, 52, 207–218, DOI: http://dx.doi.org/10.1016/j.qref.2012.04.004. Baur, D. (2003), Testing for Contagion – Mean and Volatility Contagion, Journal of Multina- tional Financial Management, 13, 405–422, DOI: http://dx.doi.org/10.1016/S1042-444X(03)00018-5. Beine, M., Laurent, S., Lecourt, Ch. (2002), Accounting for Conditional Leptokurtosis and Closing Days Effects in FIGARCH Models of Daily Exchange Rates, Applied Finan- cial Economics, 12, 589–600, DOI: http://dx.doi.org/10.1080/09603100010014041. Bollerslev, T., Mikkelsen, H.O. (1996), Modeling and Pricing Long Memory in Stock Market Volatility, Journal of Econometrics, 73, 157–160, DOI: http://dx.doi.org/10.1016/0304-4076(95)01736-4. Burzała, M. (2013), Determination of the Time of Contagion in Capital Markets Based on the Switching Model, Dynamic Econometric Models, 13, 69–85, DOI: http://dx.doi.org/10.12775/DEM.2013.004. Campbell, J.Y., Lettau, M., Malkiel, B., Xu, Y. (2001), Have Individual Stocks Become More Volatile? An Empirical Exploration of Idiosyncratic Risk, Journal of Finance, 56, 1– 43, DOI: http://dx.doi.org/10.1111/0022-1082.00318. Cavaglia, S., Brightman, C., Aked, M. (2000), The Increasing Importance of Industry Factors, Financial Analysts Journal, 56, 41–54, DOI: http://dx.doi.org/10.2469/faj.v56.n5.2389. Chen, Y., Lai, K.K. (2013), Examination on the Relationship Between VHSI, HSI and Future Realized Volatility with Kalman Filter, Eurasian Business Review, 3, 200–216, DOI: http://dx.doi.org/10.14208/ebr.2013.03.02.005. http://dx.doi.org/10.1016/j.qref.2012.04.004 http://dx.doi.org/10.1016/S1042-444X(03)00018-5 http://dx.doi.org/10.1080/09603100010014041 http://dx.doi.org/10.1016/0304-4076(95)01736-4 http://dx.doi.org/10.12775/DEM.2013.004 http://dx.doi.org/10.2469/faj.v56.n5.2389 Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 114 Chuliá, H., Torró, H. (2011), Firm Size and Volatility Analysis in the Spanish Stock Market, The European Journal of Finance, 17, 695–715, DOI: http://dx.doi.org/10.1080/1351847X.2011.554286. Claessens, S., Forbes, K. (2004), International Financial Contagion: The Theory, Evidence and Policy Implication, https://pdfs.semanticscholar.org/0fda/b705fde6484961e3679cecd41e449721a132.pdf (2.12.2014). Corsetti, G., Pericoli, M., Sbracia, M. (2005), Some Contagion, Some Interdependence, More Pitfalls in Tests of Financial Contagion, Journal of International Money and Finance, 24, 1177–1199, DOI: http://dx.doi.org/10.1016/j.jimonfin.2005.08.012. Dungey, M., Fry, R., Gonzalez-Hermosillo, B., Martin, V.L., Tang, Ch. (2011), Contagion and the Transmission of Financial Crisis, in Kolb R.W. (ed.), Financial Contagion: The Viral Threat to the Wealth of Nations, John Wiley & Sons, Inc., Hoboken, New Jersey. Feurer, R., Chaharbaghi, K. (1994), Defining Competitiveness: A Holistic Approach, Man- agement Decision, 32, 49–58, DOI: http://dx.doi.org/10.1108/00251749410054819. Fiszeder, P. (2009), Modele klasy GARCH w empirycznych badaniach finansowych (The Class of GARCH Models in Empirical Finance Research), Wydawnictwo UMK, To- ruń. Fiszeder, P., Perczak, G. (2013), A New Look at Variance Estimation Based on Low, High and Closing Prices Taking Into Account the Drift, Statistica Neerlandica, 67, 456–481, DOI: http://dx.doi.org/10.1111/stan.12017. Giot, P. (2005), Relationships Between Implied Volatility Indexes and Stock Index Returns, The Journal of Portfolio Management, 31, 92–100, DOI: http://dx.doi.org/10.3905/jpm.2005.500363. Grabowska, M. (2013), Wartość rynkowa oznaką pozycji i przewagi konkurencyjnej przed- siębiorstw, (The Market Value as an Indication of the Position and Competitive Ad- vantage of Enterprises), Zeszyty Naukowe Uniwersytetu Szczecińskiego. Finanse, Rynki Finansowe, Ubezpieczenia (Scientific Journal of the University of Szczecin, Fi- nance, Financial Markets, Insurance), 786 (64/1), 147–155. Griffin, J.M., Karolyi, G.A. (1998), Another Look at the Role of the Industry Structure of Markets for International Diversification Strategies, Journal of Financial Economics, 50, 351–373, DOI: http://dx.doi.org/10.1016/S0304-405X(98)00041-5. Hosking, J.R.M. (1981), Fractional Differencing, Biometrika, 68, 165–176, DOI: https://doi.org/10.1093/biomet/68.1.165. Kang, S. H., Yoon, S. M. (2012), Dual Long Memory Properties with Skewed and Fat-Tail Distribution, International Journal of Business and Information, 7, 225–249. Karolyi, A. (2001), Why Stock Return Volatility Really Matters, Strategic Investor Relations, Institutional Investor Journals Series, http://bryongaskin.net/education/MBA%20track/Current/MBA611/Assignments/Proje ct/WhyVolatilityMatters.pdf (15.10.2014). Konopczak, M., Sieradzki, R., Wiernicki, M. (2010), Kryzys na światowych rynkach finansowych–wpływ na rynek finansowy w Polsce oraz implikacje dla sektora realnego (Global Financial Markets Crisis – Impact on the Polish Fnancial Market and Implications for the Real Sector of the Economy), Bank i Kredyt (Bank and Credit), 41, 45–70. Krysicki, W., Bartos, J., Dyczka, W., Królikowska, K., Wasilewski, M. (2012), Rachu- nek prawdopodobieństwa i statystyka matematyczna w zadaniach, (Probability and Mathematical Statistics in Exercises), Part 2, PWN, Warsaw. http://dx.doi.org/10.1080/1351847X.2011.554286 http://dx.doi.org/10.1016/j.jimonfin.2005.08.012 http://dx.doi.org/10.1108/00251749410054819 http://dx.doi.org/10.1016/S0304-405X(98)00041-5 https://doi.org/10.1093/biomet/68.1.165 Analysis of the Relationship between Market Volatility and Firms Volatility… DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 115 Laurent, S. (2013), Estimating and Forecasting ARCH Models Using G@RCH 7, Timberlake Consultants Ltd, London. Le, C., David, D. (2014), Asset Price Volatility and Financial Contagion Analysis Using the MS-VAR Framework, Eurasian Economic Review, 4,133–162, DOI: http://dx.doi.org/10.1007/s40822-014-0009-y. Molnár, P.(2012), Properties of Range-based Volatility Estimators, International Review of Financial Analysis, 23, 20–29, DOI: http://dx.doi.org/10.1016/j.irfa.2011.06.012. Otola, I. (2013), Procesy zarządzania przedsiębiorstwami a konkurencyjność w warunkach zarażonego rynku (The Management Processes of Enterprises and Competiveness Un- der Condition of Market Contagion), Wydawnictwo Politechniki Częstochowskiej, Częstochowa. Peng, Y., Ng, W. (2012), Analysing Financial Contagion and Asymmetric Market Depend- ence with Volatility Indices via Copulas, Annals of Finance, 8, 49–74, DOI: http://dx.doi.org/10.1007/s10436-011-0181-y. Phylatkis, K., Xia, L. (2009), Equity Market Comovement and Contagion: A Sectoral Per- spective, Financial Management, 38, 381–409, DOI: http://dx.doi.org/10.1111/j.1755-053X.2009.01040.x. Sadigue, S., Silvapulle, P. (2011), Long-term Memory in Stock Market Returns: International Evidence, International Journal of Finance and Economics, 6, 59–67, DOI: http://dx.doi.org/10.1002/ijfe.143. Sequeira, J.M., Lan, D. (2003), Does World-level Volatility Matter for the Average Firm in a Global Equity Market?, Journal of Multinational Financial Management, 13, 341– –357, DOI: http://dx.doi.org/10.1016/S1042-444X(03)00015-X. Sharma, S.S., Narayan, P.K., Zheng, X. (2011), An Analysis of Firm and Market Volatility, An Financial Econometrics Series, SWP 2011/02, Deakin University, Australia, https://core.ac.uk/download/pdf/6266325.pdf (12.06.2012). Sharma, S.S., Narayan, P.K., Zheng, X. (2014), An Analysis of Firm and Market Volatility, Economic Systems, 38, 205–220, DOI: http://dx.doi.org/10.1016/j.ecosys.2013.12.003. Shin, H.H., Stulz, R.M. (2000), Firm Value, Risk and Growth Opportunities, NBER Working Paper 7808, http://www.nber.org/papers/w7808.pdf (2.03.2015). Simon, D.P. (2003), The Nasdaq Volatility Index During and After the Bubble, The Journal of Derivatives,11, 9–24, DOI: http://dx.doi.org/10.3905/jod.2003.319213. Whaley, R.E. (2009), Understanding the VIX, Journal of Portfolio Management, 35, 98–105, DOI: http://dx.doi.org/10.3905/jpm.2009.35.3.098. Włodarczyk, A. (2010), Testowanie efektu długiej pamięci w zmienności cen metali szlachet- nych (Testing for Long Memory in Volatility of Precious Metal Prices), Roczniki Na- ukowe. Seria B, Nauki Ekonomiczne i Informatyka (Scientific Annals. Series B, Eco- nomic Science and Informatics), 1–2, 151–170. Analiza relacji między zmiennością rynku a zmiennością firm na polskim rynku kapitałowym Z a r y s t r e ś c i. W artykule analizowano kształtowanie się zależności między zmiennością rynku a zmiennością cen akcji w spółkach z wybranych sektorów Giełdy Papierów Warto- ściowych w Warszawie przed i po kryzysie subprime. Badania empiryczne dotyczą wybra- nych przedsiębiorstw z sektorów budownictwo i IT w latach 2004–2011. Miary zmienności zostały obliczone na podstawie najniższego i najwyższego kursu dziennego spółek należących http://dx.doi.org/10.1016/j.irfa.2011.06.012 http://dx.doi.org/10.1016/S1042-444X(03)00015-X http://dx.doi.org/10.1016/j.ecosys.2013.12.003 http://dx.doi.org/10.3905/jpm.2009.35.3.098 Włodarczyk, Otola DYNAMIC ECONOMETRIC MODELS 16 (2016) 87–116 116 do indeksu WIG. Dla każdej firmy został oszacowany, zarówno dla okresu stabilnego, jak i niestabilnego, model ARFIMAX-FIGARCH z dodatkowymi zmiennymi egzogenicznymi, odzwierciedlającymi zmienność rynku. Przeprowadzone badania empiryczne nie wykazały, że negatywne szoki płynące z rynku amerykańskiego przez kanał zachowań inwestorów przyczyniły się do wzrostu frakcji firm z sektorów budownictwa i IT notowanych na GPW, których zmienność jest kształtowana przez zmienność rynku. S ł o w a k l u c z o w e: ARFIMAX-FIGARCH, kryzys subprime, zmienność firm, zmien- ność rynku, Giełda Papierów Wartościowych w Warszawie.