1-2 MINERVA BANCARIA RIVISTA BANCARIA - U-PAD Unimc
←
→
Trascrizione del contenuto della pagina
Se il tuo browser non visualizza correttamente la pagina, ti preghiamo di leggere il contenuto della pagina quaggiù
RIVISTA BANCARIA MINERVA BANCARIA www.rivistabancaria.it ISTITUTO DI CULTURA BANCARIA «FRANCESCO PARRILLO» Gennaio-Aprile 2020 1-2 Tariffa Regime Libero:-Poste Italiane S.p.a.-Spedizione in abbonamento Postale-70%-DCB Roma
RIVISTA BANCARIA MINERVA BANCARIA COMITATO SCIENTIFICO (Editorial board) PRESIDENTE (Editor): GIORGIO DI GIORGIO, Università LUISS Guido Carli, Roma MEMBRI DEL COMITATO (Associate Editors): PAOLO ANGELINI, Banca d’Italia GIOVANNI FERRI, Università LUMSA MASSIMO BELCREDI, Università Cattolica del S.C. FRANCO FIORDELISI, Università degli Studi “Roma Tre” - co Editor EMILIA BONACCORSI DI PATTI, Banca d’Italia LUCA FIORITO, Università degli Studi di Palermo CONCETTA BRESCIA MORRA, Università degli Studi “Roma Tre” FABIO FORTUNA, Università Niccolò Cusano FRANCESCO CANNATA, Banca d’Italia EUGENIO GAIOTTI, Banca d’Italia ALESSANDRO CARRETTA, Università degli Studi di Roma “Tor Vergata” GUR HUBERMAN, Columbia University ENRICO MARIA CERVELLATI, Università di Bologna AMIN N. KHALAF, Ernst & Young RICCARDO CESARI, Università di Bologna e IVASS MARIO LA TORRE, Sapienza - Università di Roma - co Editor NICOLA CETORELLI, New York Federal Reserve Bank RAFFAELE LENER, Università degli Studi di Roma “Tor Vergata” SRIS CHATTERJEE, Fordham University NADIA LINCIANO, CONSOB N.K. CHIDAMBARAN, Fordham University PINA MURÉ, Sapienza - Università di Roma LAURENT CLERC, Banque de France FABIO PANETTA, Banca Centrale Europea MARIO COMANA, LUISS Guido Carli ALBERTO FRANCO POZZOLO, Università degli Studi “Roma Tre GIANNI DE NICOLÒ, International Monetary Fund ZENO ROTONDI, Unicredit Group RITA D’ECCLESIA, Sapienza - Università di Roma ANDREA SIRONI, Università Bocconi GIOVANNI DELL’ARICCIA, International Monetary Fund MARIO STELLA RICHTER, Università degli Studi di Roma “Tor Vergata” STEFANO DELL’ATTI, Università degli Studi di Foggia - co Editor MARTI SUBRAHMANYAM, New York University CARMINE DI NOIA, CONSOB ALBERTO ZAZZARO, Università degli Studi di Napoli “Federico II” LUCA ENRIQUES, University of Oxford Comitato Accettazione Saggi e Contributi: GIORGIO DI GIORGIO (editor in chief) - Domenico Curcio (co-editor) Alberto Pozzolo (co-editor) - Mario Stella Richter (co-editor) Direttore Responsabile: Giovanni Parrillo Comitato di Redazione: Francesco Baldi, Peter Cincinelli, Alfonso Del Giudice, Vincenzo Formisano, Stefano Marzioni, Federico Nucera, Biancamaria Raganelli, Stefania Sylos Labini, Giuseppe Zito. ISTITUTO DI CULTURA BANCARIA «FRANCESCO PARRILLO» PRESIDENTE CLAUDIO CHIACCHIERINI VICE PRESIDENTI MARIO CATALDO - GIOVANNI PARRILLO CONSIGLIO TANCREDI BIANCHI, FABRIZIO D’ASCENZO, GIAN GIACOMO FAVERIO, ANTONIO FAZIO, GIUSEPPE GUARINO, PAOLA LEONE, ANTONIO MARZANO, FRANCESCO MINOTTI, PINA MURÈ, FULVIO MILANO, ERCOLE P. PELLICANO’, CARLO SALVATORI, MARIO SARCINELLI, FRANCO VARETTO In copertina: “Un banchiere e sua moglie” (1514) di Quentin Metsys (Lovanio, 1466 - Anversa, 1530), Museo del Louvre - Parigi.
RIVISTA BANCARIA MINERVA BANCARIA ANNO LXXVI (NUOVA SERIE) GENNAIO-APRILE 2020 N. 1-2 SOMMARIO Editoriale G. DI GIORGIO La politica monetaria nella pandemia. Ci salverà l’elicottero? ... 3 Saggi R. COCOZZA D. CURCIO A. PACIFICO Do Global Markets Imply Common Fear? ............................. 7 Le piccole e medie imprese italiane nella riflessione G. GAROFALO di Francesco Parrillo e nel contesto dell’economia italiana G. GUARINI del III millennio: un percorso di ricerca sul capitalismo C. CHIACCHIERINI italiano con una verifica empirica .......................................... 47 Saggi - Sezione Giovani G. CHIORAZZO Dismissione di NPL tramite iniziative di sistema (bad-bank): un modello contabile per la simulazione degli effetti sui bilanci delle banche italiane ........................... 77 E. GABRIELE Tips for Financial Risk Managers During QE Enforcement: Evidence from EU-Core Countries ......... 129 Rubriche Valutazioni giuridiche sull’impatto della Mifid II (A. Fittante) .................................................................................................................... 171 L’incertezza frena le imprese. Nel 2019 calano le richieste di credito (E. Mazzotti) .................................................................................................................. 183 Bankpedia ...................................................................................................................... 189 Assicurazione dei depositi in Italia e nella UE (C. Oldani) La nuova Via della Seta - Belt and Road Initiative (C. Oldani) Recensioni F. Capriglione, A. Sacco Ginevri, Metamorfosi della governance bancaria (Marina Brogi)................................................................................................................ 203 A. Dell’Atti, F. Miglietta e A. P. Iannuzzi, Il sistema bancario e la crisi finanziaria (S. Sylos Labini) .............................................................................................................. 206 ISSN: 1594-7556 La Rivista è accreditata AIDEA e SIE Econ.Lit
RIVISTA BANCARIA - MINERVA BANCARIA Rivista Bancaria - Minerva Bancaria è sorta nel 1936 dalla fusione fra le precedenti Rivista Bancaria e Minerva Bancaria. Dal 1945 - rinnovata completamente - la Rivista ha proseguito senza interruzioni l’attività di pubbli- cazione di saggi e articoli in tema di intermediazione bancaria e finanziaria, funzionamento e regolamentazione del sistema finanziario, economia e politica monetaria, mercati mobiliari e finanza in senso lato. Particolare attenzione è dedicata a studi relativi al mercato finanziario italiano ed europeo. La Rivista pubblica 6 numeri l’anno, con possibilità di avere numeri doppi. Note per i collaboratori: Gli articoli ordinari possono essere presentati in italiano o in inglese e devono essere frutto di ricerche originali e inedite. Ogni articolo viene sottoposto alla valutazione anonima di due referee selezionati dal Comitato Scientifico, ed eventualmente da un membro dello stesso. Gli articoli accettati sono pubblicamente scaricabili (fino alla pubblicazione cartacea) sul sito della rivista: www.rivistabancaria.it Gli articoli dovranno essere corredati da una sintesi in italiano e in inglese, di massimo 150 parole. Per maggiori indicazioni sui criteri redazionali si rinvia al sito della Rivista. La Rivista ospita anche, periodicamente, interventi pubblici, atti di convegni patrocinati dalla Rivista stessa, dibattiti, saggi ad invito e rubriche dedicate. Questi lavori appaiono in formato diverso dagli articoli ordinari. La responsabilità di quanto pubblicato è solo degli autori. Gli autori riceveranno in omaggio tre copie della Rivista Gli articoli possono essere sottomessi inviando una email al seguente indirizzo: redazione@rivistabancaria.it Istituto di Cultura Bancaria “Francesco Parrillo” L’Istituto di Cultura Bancaria è un’associazione senza finalità di lucro fondata a Milano nel 1948 dalle maggiori banche dell’epoca allo scopo di diffondere la cultura bancaria e di provvedere alla pubblicazione della Rivista. La Rivista è stata diretta dal 1945 al 1974 da Ernesto d’Albergo e poi per un altro trentennio da Francesco Parrillo, fino al 2003. In questo secondo periodo, accanto alla trat- tazione scientifica dei problemi finanziari e monetari, la rivista ha rafforzato il suo ruolo di osserva- torio attento e indipendente della complessa evoluzione economica e finanziaria del Paese. Giuseppe Murè, subentrato come direttore dal 2003 al 2008, ha posto particolare accento anche sui problemi organizzativi e sull’evoluzione strategica delle banche. Nel 2003, l’Istituto di Cultura Bancaria è stato dedicato alla memoria di Francesco Parrillo, alla cui eredità culturale esso si ispira. Editrice Minerva Bancaria srl DIREZIONE E REDAZIONE Largo Luigi Antonelli, 27 – 00145 Roma redazione@rivistabancaria.it AMMINISTRAZIONE EDITRICE MINERVA BANCARIA S.r.l. presso P&B Gestioni Srl, Viale di Villa Massimo, 29 - 00161 - Roma - Fax +39 06 83700502 amministrazione@editriceminervabancaria.it Spedizione in abbonamento postale - Pubblicazione bimestrale - 70% - Roma Finito di stampare nel mese di aprile 2020 presso Press Up, Roma Segui Editrice Minerva Bancaria su:
DO GLOBAL MARKETS IMPLY COMMON FEAR? ROSA COCOZZA DOMENICO CURCIO ANTONIO PACIFICO Abstract This research investigates the relations of the US VIX and the European VSTOXX, on the one hand, and their main determinants, on the other hand. In line with prior studies, US and European stock and volatility markets ex- hibit a contemporaneous negative relationship. Furthermore, the increase of VIX and VSTOXX associated with a fall in stock prices is larger than the decrease that the two indices experience when stock prices rise. During the great financial crisis, the negative relation between volatility indices and stock market indices weakens for both financial markets; the asymmetric reaction to changes in stock market returns becomes stronger for the VSTOXX index and decreases for the VIX. The analysis of the dynamic interaction between volatility indices and main financial and macroeconomic variables based on a Structural Panel Bayesian VAR shows that the US monetary policy rate is an important driver of both the US and European volatility indices’ behavior, thus hugely affecting not only the domestic financial market, but also the European one. Moreover, the presence of consistent co-movements and interdependencies among volatil- ity indices prove that international spillovers are mainly linked via financial markets. Department of Economics, Management, Institutions, University of Naples “Federico II”, Via Cinthia, Com- plesso Monte S. Angelo, Napoli (NA) 80126, Italy, E-mail: rosa.cocozza@unina.it Corresponding author. Department of Economics, Management, Institutions, University of Naples “Federico II” Via Cinthia, Complesso Monte S. Angelo, Napoli (NA) 80126, Italy, E-mail: domenico.curcio@unina.it Department of Political Science, LUISS Guido Carli, Viale Romania 32, Roma (RM) 00197, E-mail: apacifi- co@luiss.it RIVISTA BANCARIA - MINERVA BANCARIA N. 1 - 2 / 2020 7
ROSA COCOZZA, DOMENICO CURCIO, ANTONIO PACIFICO Mercati globali implicano comune paura? – Sintesi Questa ricerca analizza le relazioni tra gli indici di volatilità VIX e VSTOXX, da una parte, e le loro principali determinanti, dall’altra. In linea con la prece- dente letteratura, gli indici dei mercati azionari statunitensi ed europei mostrano una relazione negativa con i corrispondenti indici di volatilità. Inoltre, l’aumento del VIX e del VSTOXX, associato ad un calo dei prezzi azionari nei rispettivi mercati, è maggiore della riduzione che i due indici di volatilità subiscono quan- do, invece, i prezzi aumentano. Durante la grande crisi finanziaria, la relazione negativa tra gli indici di volatilità e gli indici azionari si indebolisce per entrambi i mercati finanziari; la reazione asimmetrica alle variazioni dei rendimenti azio- nari diventa più forte per il VSTOXX e si attenua per il VIX. L’analisi dell’interazione dinamica tra gli indici di volatilità e le principali variabili finanziarie e macroeconomiche, basata su un modello di tipo Structural Panel Bayesian VAR, mostra che il tasso di politica monetaria degli Stati Uniti è un driver importante del comportamento di entrambi gli indici di volatilità, in grado di influenzare, quindi, in maniera significativa, non solo il mercato finan- ziario domestico, ma anche quello europeo. Inoltre, lo studio mostra la presenza di co-movimenti e interdipendenze tra gli indici di volatilità, che dimostrano che fenomeni di spillover a livello internazionale sono interrelati per il tramite dei mercati finanziari. Parole chiave: Indici di Volatilità; Indici Azionari; Politica Monetaria; Tassi di Interesse. Codici JEL: E44; F30; F44. Keywords: Volatility Indices; Stock Indices; Monetary Policy; Interest Rates 8 SAGGI
DO GLOBAL MARKETS IMPLY COMMON FEAR? 1. Introduction According to a growing body of literature, risky asset prices around the globe and international capital flows do have a strong common component (see, among the others, Calvo et al., 1996, Bekaert et al., 2012, Miranda Agrippino and Rey, 2012). Cycles in the real interest rates and in the growth rate of advanced economies have been identified as important “push” factors for capital flows. Several studies have found that movements in the VIX, the Chicago Board Options Exchange (CBOE) market volatility index, and in other indices of “market fear” (Forbes and Warnock, 2012) are strongly and negatively associated with capital flows. Bruno and Shin (2015) emphasize the surge in international capital flows caused by the lowering of the VIX. Furthermore, even financial intermediaries’ leverage and leverage growth are negatively correlated with the VIX (Rey, 2016). Within such a financially globalized world, monetary conditions of the center country (the US) have an impact on changes in aggregate risk aversion and volatility, capital flows and the leverage of the financial sector in many parts of the international financial system (Miranda Agrippino and Rey, 2012). While seeing a lot of co-movement in asset prices and capital flows worldwide may just reflect financial markets integration, the fact that these co-movements are to some extent caused by US monetary policy is important and we believe that shedding more light on the role played by volatility indi- ces is crucial to get a comprehensive knowledge of how financial globalization works. On the other hand, empirical evidence shows a strong and persistent correlation among main volatility indices, which is much higher than the correlation among the corresponding stock market indices (Aussenegg et al., 2013). Global financial market integration may not be the only reason to ex- plain such strong and persistent co-movements in the volatility indices, given the structural differences between the reference stock markets. By focusing on the VIX and the VSTOXX volatility indices, we observe a stable and very high positive correlation since their introduction. The correla- tion of the two volatility indices is much higher than the correlation between the two corresponding stock market indices, the S&P 500 and Euro Stoxx 50, that are different because of a number of factors, among which the liquidity RIVISTA BANCARIA - MINERVA BANCARIA N. 1 - 2 / 2020 9
ROSA COCOZZA, DOMENICO CURCIO, ANTONIO PACIFICO of their reference markets, the economic sectors respectively represented in the two stock indices and the leverage of the companies included in each of them. By accounting for these issues and the post-crisis financial markets con- ditions, detecting the factors determining market volatility indices is a timely and key research area. Since, to the best of our knowledge, there is no prior literature specifically investigating the joint dynamics of volatility indices, this research aims to fill this gap and investigate the causes of what we define “common fear” of global markets. In particular, in order to address the issues associated with the main determinants of the volatility indices, this research contributes to prior literature by: i) shedding light on the link between the volatility indices and the risk-free interest rates that are used in their calculation; ii) adding new evi- dence, particularly referred to the great financial crisis period, on the relation- ship between the volatility indices and the corresponding stock indices, with a special focus on the great financial crisis; and iii) providing new insights about the importance of monetary policy actions in determining market implied volatility from an international standpoint. From a methodological perspective, we first examine how market fear in- dices are built in order to detect potential drawbacks in their construction models. Based on an analytic approach, we show how stock market indices and interest rates included in the VIX and VSTOXX formulas contribute to the level and dynamics of the two fear indices themselves. Then, based on the analysis of the time series of daily market data, in order to further investigate the determinants of these two indices, we study their sensitivity to both global and local financial market variables, such as US and Euro Area monetary pol- icy rates and stock market indices. Finally, the adoption of a Structural Panel Bayesian Vector Autoregressive (SPBVAR) framework allows us to compre- hensively capture the interdependencies among the fear indices and the other financial variables included in our analysis as well as their dynamic behavior. Our main findings are: i) US and European stock and volatility markets exhibit a contemporaneous negative relationship, which is slightly less pro- nounced for the European market; ii) the reaction of VIX and VSTOXX to changes in stock prices is asymmetric in the sense that the increase of volatil- ity indices when stock prices go down is larger than the decrease caused by a 10 SAGGI
DO GLOBAL MARKETS IMPLY COMMON FEAR? rise in stock prices; iii) during the great financial crisis, the negative relation between volatility indices and stock market indices weakens for both financial markets, whereas the asymmetric reaction to changes in stock market returns becomes stronger for the VSTOXX index and less intense for the VIX; iv) the US monetary policy rate appears to be an important driver of the US and European volatility indices’ behavior. The rest of the paper is organized as follows: section 2 presents the fear indices we are interested in and focuses on their calculation methodology, interpretation and potential benefits for investors; in section 3 we discuss the relationship between the volatility indices and their main determinants, namely the underlying stock market indices and interest rates; in section 4 we present the empirical evidence resulting from static OLS regression models and from a Structural Panel Bayesian Vector Autoregressive (SPBVAR) model analysis; section 5 contains some concluding remarks. 2. The VIX and VSTOXX indices: calculation methodology, interpreta- tion and benefits for investors The Chicago Board Options Exchange (CBOE) volatility index (VIX) is a forward-looking measure of the future market volatility that investors expect to see over the next 30 calendar days and is implied by the current prices of the S&P500 index (SPX index) options. In particular, to calculate the VIX, near- and next-term put and call options with more than 23 days and less than 37 days to expiration are taken into account. The VIX was introduced in 1993 in order to provide, on the one hand, a benchmark for expected short- term market volatility and, on the other hand, an index upon which futures and options contracts could be written. The VIX was initially designed to measure the market’s expectation of 30- day volatility implied by at-the-money S&P100 index (OEX index) option prices. Later, in 2003, the CBOE introduced a more detailed methodology to calculate the VIX, which is based on the S&P500 index and estimated expect- ed volatility by averaging the weighted prices of S&P500 puts and calls over a RIVISTA BANCARIA - MINERVA BANCARIA N. 1 - 2 / 2020 11
ROSA COCOZZA, DOMENICO CURCIO, ANTONIO PACIFICO wide range of strike prices1. Finally, in 2014, the CBOE enhanced the VIX to include series of SPX Weeklys, allowing to calculate the index with S&P500 index option series that most precisely match the 30-day target timeframe for expected volatility that the VIX aims to represent2. VIX calculation employs rules for selecting component options and the following, generalized formula to calculate the index values: 2 | 3K i RT 1: F D 2 VIX = 100 # T i K 2i e Q ( K i) - T K0 - 1 (1) where: T = Time to expiration; F = forward SPX level desired from index option prices. It is determined by identifying the strike price at which absolute difference between the call and put prices is smallest; K0 = First strike below the forward index level F; K i = Strike price for the i th out-of-the-money option; a call if K i > K 0 , and a put if K i < K 0 , both put and call if K i = K 0; 3K i = interval between strike prices, half the difference between the strike Ki+1 - Ki-1 on either side of K i: 3 K i = 2 ; Q (K i) = the midpoint of the bid-ask spread for each option with stike ; R = risk free interest rate to expiration. The VIX is quoted in percentage points and represents the expected range of movement in the S&P500 index over the next year, at a 68% confidence level (i.e. one standard deviation of the normal probability curve). For exam- ple, a VIX equal to 15 represents an expected annualized change, with a 68% probability, of less than 15% up or down. Based on the same methodology, European stock exchanges also created volatility indices for several European stock markets. In particular, Eurex in- troduced the Euro STOXX 50 volatility index (VSTOXX) to measure the im- 1 In particular, the new methodology is based on Carr and Madan (1998) and Demeterfi et al. (1999) work on pricing variance swaps. 2 Using SPX options with more than 23 days and less than 37 days to expiration ensures that the VIX will always reflect an interpolation of two points along the S&P500 volatility term structure. 12 SAGGI
DO GLOBAL MARKETS IMPLY COMMON FEAR? plied volatility of the EURO STOXX 50 index (ESTOXX), i.e., to indicate, in percentage points, which volatility is to be expected for the EURO STOXX 50 index over the coming 30 days. The index is based on a basket of EURO STOXX 50 index options quoted at-the-money or out-the-money. VIX and VSTOXX are measures of the overall market perceived volatility, both upwards and downwards. Therefore, high values of the index do not necessarily represent negative predictions for stocks movements, since inves- tors can become nervous even during market rallies. For example, Panel A of Figure 1 shows that the VIX spikes during periods of market turmoil, as it happened right after the Lehman Brothers’ collapse on September 2008, but the volatility index might also rise when the level of the S&P 500 index increases. Figure 1 - Volatility and stock market indices (January 1st, 2002 – December 31st, 2015; daily observations) Panel A - VIX (black line and right scale in % per year) and S&P 500 (grey line and left scale in points) RIVISTA BANCARIA - MINERVA BANCARIA N. 1 - 2 / 2020 13
ROSA COCOZZA, DOMENICO CURCIO, ANTONIO PACIFICO Panel B - VSTOXX (black line and right scale in % per year) and Euro Stoxx 50 (grey line and left scale in points) Source: our elaborations on data from DatastreamTM. As concerns the interpretation of the values and trends of the volatility indices, if investors expect great upside volatility, they will sell upside stock call options only if they obtain a large premium. On the other hand, option buyers will pay such high premiums only if they similarly anticipate a large upside movement. Due to the resulting aggregate of increase in upside stock call option prices, the volatility index will raise. Similarly, the index will raise in the case of an aggregate growth in downside stock put option premiums that would occur when both option buyers and sellers expect a sharp down- ward movement. When investors perceive neither significant downside risk nor significant upside potential, the volatility index will be low. By focusing on stock prices, the behavior of volatility indices can be explained as follows: if expected market volatility increases (decreases), investors demand higher (lower) rates of return on stocks, so stock prices fall (rise) (Whaley, 2009). Prior literature has investigated both the contribution of VIX and other volatility indices to investors’ portfolio management and how well the im- plied volatility forecasts future realized volatility. From a portfolio manage- ment perspective, since they are tradeable via swaps, futures, options and exchange traded notes, volatility indices can help to get a desired level of portfolio diversification and can provide investors with easy access to strate- gies involving hedging of equity risk by taking positions in implied volatility. 14 SAGGI
DO GLOBAL MARKETS IMPLY COMMON FEAR? For example, when equity markets become highly volatile and the portfolio tracking error and the rebalancing costs increase, using volatility futures helps to hedge against these frictional costs. On the other hand, investors may also use volatility index derivatives to enrich the set of potential speculative di- rectional positions. Szado (2009) and Rhoads (2011) highlight the potential benefits of adding volatility derivatives to equity index portfolios. Similarly, Chen et al. (2011) demonstrate that adding VIX futures contracts can im- prove the mean-variance investment frontier, while Daigler and Rossi (2006) find significant diversification benefits from adding a long VIX position to an S&P500 portfolio. The question of how well the implied volatility forecasts future realized volatility has received a great deal of attention in the financial literature. The general conclusion of prior studies is that implied volatility outperforms the well-known volatility measures based on historical data: see, among the oth- ers, Blair et al. (2001), Corrado and Miller (2005) as well as Carr and Wu (2006), who show that VIX outperforms GARCH volatility estimated from the S&P500 index returns. However, Becker et al. (2006) find that VIX is not an efficient forecaster of future realized volatility, and that other volatility estimates based on historical data can be superior to VIX alone. 3. Analyzing the determinants of volatility indices: stock market indices and interest rates Equation (1) shows that volatility indices are functions of a wide set of variables. In order to address the issues associated with the main determinants of the volatility indices, this research contributes to prior literature by: i) add- ing new evidence on the relationship between the volatility indices (VIt) and stock indices (SIt), that, in addition to prior literature, we examine during the great financial crisis, ii) shedding light on the link between the volatility indi- ces, the risk-free interest rates (IRt) that are used to calculate the US VIX and European VSTOXX, and iii) providing new insights about the importance of monetary policy actions in determining market implied volatility. RIVISTA BANCARIA - MINERVA BANCARIA N. 1 - 2 / 2020 15
ROSA COCOZZA, DOMENICO CURCIO, ANTONIO PACIFICO 3.1. Volatility and stock indices According to prior literature, the negative relationship between stock re- turn and implied volatility can be justified by either firm fundamentals or by the behavior of market participants. Within the former group of studies, ac- cording to the “leverage hypothesis”, the increase in leverage, which is caused by a decrease in stock prices, determines the increase in the equity volatil- ity and the associated risk of equity holders (Black, 1976; Christie, 1982; Schwert, 1989). Based on the “feedback hypothesis”, due to a rise in volatility, investors raise the future rates of stock returns they ask for, thus determining a fall in stock prices (French et al., 1987; Bekaert and Wu, 2000; Wu, 2001; Kim et al., 2004). As concerns the studies accounting for the behavior of mar- ket participants, investors believe that good investments are characterized by low risk and high return and associate larger negative (positive) returns with larger (smaller) volatility (Hibbert et al., 2008; Low, 2004). From a methodological perspective, we start our analysis by regressing the daily log returns of all sample stock indices on the corresponding volatility index log returns, based on the following regression model: 3VI t = c + b 1 3 SI t + f t (2) where: 3VI t is the log return of the i - th volatility index (either the VIX or the VSTOXX), 3SI t is the log return of the corresponding i - th stock mar- ket index (either the S&P500 or the EURO STOXX 50) and f t represents a normally distributed error term. Early studies also investigate for a potential asymmetric relation between volatility and stock market indices, but evidence is not conclusive. For ex- ample, Giot (2005) finds not only that the relation between the S&P100’s returns and the changes in VXO (the “old VIX”) is asymmetric, but also that it tends to differ in periods of low and high volatility. The increase in implied volatility (when negative stock index returns occur) is lower in high-volatility periods than in low-volatility periods. Fleming et al. (1995) also find a statis- tically significant asymmetric relation between the old VIX and S&P returns, 16 SAGGI
DO GLOBAL MARKETS IMPLY COMMON FEAR? whereas Simon (2003)’s results show a statistically significant asymmetric re- lation between VIX, VNX (the technology sector based on the Nasdaq 100 implied volatility) and their respective equity indices. Siriopoulos and Fassas (2009) and Whaley (2000) do not find evidence for strong asymmetric re- lation between VIX, RVX (the CBOE Russell 2000 Volatility Index), VNX and their respective stock market indices. Alexander (2008) and Siriopoulos and Fassas (2012) both show an asymmetric negative relation for VFTSE (the FTSE volatility index) and FTSE whilst Gonzalez and Novales (2009) report lack of asymmetric negative relation between VDAX (the German DAX vola- tility index), VSMI (the Swiss SMI volatility index), and their corresponding stock market indices. In order to detect any potential asymmetric relation between each of the two volatility indices and the underlying stock indices, following Whaley (2009), we use a slightly modified version of the previous model, that can be written as follows: 3VI t = b 0 + b 1 3 SI t + b 2 3 SI -t + f t (3) where 3SI -t is the log return of the stock market index conditional on the market going down and 0 otherwise. Should both the slope coefficients be negative and statistically sgnificant, we would find support not only to the in- verse relation between stock market index and volatility, but also to its asym- metric nature. 3.2. Volatility indices and interest rates VIX and VSTOXX are calculated as risk-neutral expectations of volatility for near- and next-term options via a linear interpolation. The risk-free in- terest rates for the expectations referred to the near- and next-term options, respectively R 1 and R 2, are the bond-equivalent yields of the US T-bill matur- ing closest to the expiration dates of relevant SPX options for the US volatility index and the corresponding Euribor rate for the European one. The risk-free interest rate is also relevant to determine the option price Q (K i) and the forward level of stock index, F, as shown by the following RIVISTA BANCARIA - MINERVA BANCARIA N. 1 - 2 / 2020 17
ROSA COCOZZA, DOMENICO CURCIO, ANTONIO PACIFICO equation (CBOE, 2009): F = Strike Price + e RT # ^ Call Price - Put Price h (4) Therefore, based on equation (4), volatility indices react to the risk-free interest rate according to the following sensitivity index, which is obtained by calculating the derivative of the volatility function with respect to the risk-free rate: 2R = 50 E 6 v 2@ # ' T | Q (K i) + T | 2 i e RT t i + 1 2VI 1 2 3K i RT 2 3K 2 Te i Ki i Ki - T : K - 1D; K Te RT E 1 2 F ( c - p) (5) 0 0 where t i measures the sensitivity of an option to a change in interest rate, c is the call price and p is the put price. The derivative of the volatility index with respect to the risk-free rate shown in equation (5) is not always positive. The sign of this sensitivity index depends on the difference in brackets and on the of the options included in the portfolio taken into account to calculate the volatility index. In particular, the sign is positive if: ^c - ph 2 ti $ T # ( :F D 2 3K i 3K i | i Ki K 0 # K 0 - 1 - | i K 2i Q (K i) (6) As a result, it is first the construction mechanism of the volatility indices to determine the final relationship of VIX and VSTOXX with interest rates. In particular, the sign of the t of the options portfolio in equation 6 depends on the relative weight of the call and put options included in the portfolio itself. The impact of a rise in interest rates on the value of a generic option de- pends on the combination of the two following effects: due to an increase in interest rates, on the one hand, the expected stock return required by inves- tors tends to rise, and, on the other hand, the present value of any future cash flow received by the holder of a certain option decreases. Ceteris paribus, the combination of these two effects leads to an increase in the value of a generic call option t call > 0 and to a decrease in the value of a generic put option 18 SAGGI
DO GLOBAL MARKETS IMPLY COMMON FEAR? (t put < 0). Finally, there is one last channel to consider to get a comprehensive view of the impact of interest rates movements on volatility indices. Consider the companies whose stocks are included in the stock index underlying the op- tions used to calculate the volatility index. If these companies pay dividends, by affecting their stock prices, interest rate movements also affect the value of the stock index and, eventually, the value of the options included in the volatility index. By accounting also for the risk-free interest rates, equation (3) can be writ- ten as follows: 3VI t = b 0 + b 1 3 SI t + b 2 3 SI -t + b 3 3 IR t + f t (7) where 3IR t is the percentage change in the risk-free rate used in the formula of each of the two volatility indices. This research also aims to shed light on the relationship between the volatility indices and the monetary policy decisions taken by the two most important central banks, namely the US Federal Reserve and the European Central Bank. By changing their policy rates, monetary authorities affect in- vestors’ decisions and expectations and can contribute to explain volatility indices behavior. In order to test for the impact of monetary policy actions on volatility indices, we introduce a new variable 3MPR t , i.e., the change in the mone- tary policy rate, in equation (7). Therefore, our final regression model can be formalized as follows: 3VI t = b 0 + b 1 3 SI t + b 2 3 SI -t + b 3 3 IR t + b 4 3 MPR t + f t (8) The global financial turmoil erupted in June 2007 occurred during the time-span we examine and the Chow test confirms the presence of a structur- al break in our time series. Therefore, to detect whether the events of those years of financial turbulences do affect the relationships we are interested in, we introduce into equations (2), (3), (7) and (8) a dummy variable CRISIS, that takes the value of 1 for the observations referred to the crisis period, and RIVISTA BANCARIA - MINERVA BANCARIA N. 1 - 2 / 2020 19
ROSA COCOZZA, DOMENICO CURCIO, ANTONIO PACIFICO 0 otherwise, together with its interaction with the regressors specified in each of the models previously defined. The final version of the regression equations that we estimate in the empir- ical section are as follows: 3VI t = b 0 + b 1 3 SI t + b 2 CRISIS + b 3 CRISIS ) 3SI t + f t (9) 3VI t = b 0 + b 1 3 SI t + b 2 CRISIS + b 3 CRISIS ) 3SI t + +b 4 3 SI -t + b 5 CRISIS ) 3SI -t + f t (10) 3VI t = b 0 + b 1 3 SI t + b 2 CRISIS + b 3 CRISIS ) 3SI t + +b 4 3 SI -t + b 5 CRISIS ) 3SI -t + b 6 3 IR t + +b 7 CRISIS ) 3IR t + f t (11) 3VI t = b 0 + b 1 3 SI t + b 2 CRISIS + b 3 CRISIS ) 3SI t + +b 4 3 SI -t + b 5 CRISIS ) 3SI -t + b 6 3 IR t + +b 7 CRISIS ) 3IR t + b 8 3 MPR t + +b 9 CRISIS ) 3MPR t + f t (12) 4. Empirical analysis 4.1. Data The data used in our empirical analysis are taken from DatastreamTM and refer to the period ranging from January 1st, 2002 to December 31st, 2015, for a total of 3,653 daily observations. The main descriptive statistics of the volatility indices time series are presented in Panel A of Table 1. Our sample is characterized by wide ranging levels of volatility: the highest single value, equal to 87.5, was reached by VSTOXX on October 16th, 2008 while the minimum of 9.9 was recorded by the VIX on January 24th, 2007. The daily mean and median volatility log returns are close to zero, reflecting the absence of a deterministic growth trend in volatility, and confirming that volatili- 20 SAGGI
DO GLOBAL MARKETS IMPLY COMMON FEAR? ty follows a mean-reverting process (Allen et al., 2006). We find significant positive skewness in both the levels and the log returns of the indices. The Shapiro-Francia statistics reject the hypothesis of a normal distribution in all cases at the 1% significance level, which implies a higher probability of extreme movements. The negative relationship between equity and volatility markets is con- firmed by the correlation matrix presented in Panel B of Table 1. The daily log returns of stock markets and volatility indices are negatively correlated at the 1% level: we observe a negative correlation of 74.5% between the VIX and S&P500 and of 74.6% between the VSTOXX and the EURO STOXX 50 index. The correlation between the levels of VIX and VSTOXX stands at an impressive value of 91.9%. The correlation between S&P500 and EURO STOXX 50 is also positive, but it is much lower than the correlation between the two volatility indices, being equal to 42.4%. We calculate the correlations between the volatility indices, on the one hand, and the stock market indices, on the other hand, for three different sub-periods, namely the last 10 years, the last 5 years and the last year of our sample period, respectively (Figure 2). Correlation between VIX and VS- TOXX ranges from 95%, if calculated over the last 10 years of the time hori- zon under analysis, to 80%, if referred to the 2015 year only. The correlation between SPX and ESTOXX stands at 36.1% if calculated over the 2006-2015 period, whereas it raises to 88.3% over the last five years of the sample period and 77.9% for the last 12 months. Figure 2 - Evolution of the correlation between VIX and VSTOXX and between SPX and Euro Stoxx 50 (daily observations). RIVISTA BANCARIA - MINERVA BANCARIA N. 1 - 2 / 2020 21
ROSA COCOZZA, DOMENICO CURCIO, ANTONIO PACIFICO Source: our elaborations on data from DatastreamTM. The time horizon we investigate covers periods with different stock market conditions. For example, Panel B of Figure 1 depicts the evolution of the EuroStoxx 50 and VSTOXX. After a minor market correction in the first half of 2004 a tight range of volatility lasted until early 2006 when the closure of a majority of Ameriquest’s branches announced the imminent credit crises3. Although low-volatility conditions lasted for several months, implied volatil- ity rose to 35% in the last quarter of 2007, when some of Bear Stearns hedge funds bankrupted. In order to detect the relation between the volatility indices and risk-free interest rates used in equation (1), we use the 4-week T-Bill interest rate for the US market and the 1-month Euribor for the European market. Panel B of Table 1 includes the pairwise correlation coefficients between the changes in these interest rates and the volatility indices. In particular, the log return of the VIX index is negatively correlated with the percentage change in the 1-month T-Bill rate at the 1% confidence level, whereas the correlation co- efficient between the log return of the VSTOXX index and the percentage change in the 1-month Euribor is positive but neither economically, nor sta- tistically significant. As concerns the impact of monetary policy on volatility indices, we rep- resent the monetary policy stance through the percentage changes in the Fed Funds rate (FFR) for the VIX and in the Euro Overnight Deposit rate (EU- RDEPRATE) for the VSTOXX. The change in FFR (∆FFR) is positively and statistically correlated with both the level of the VIX index and its log returns at the 1% confidence level, with two correlation coefficients approximately equal to 6% and 4%. The correlation coefficients of the change in the Euro Overnight Deposit rate (∆EURDEPRATE) with the level of the VSTOXX index and its log returns are -1.1% and 0.6%, respectively. Nevertheless, nei- ther of them is statistically significant. 3 Ameriquest was one of the largest mortgage lenders in North America. 22 SAGGI
DO GLOBAL MARKETS IMPLY COMMON FEAR? Table 1 - Volatility indices, stock market indices and interest rates (January 1st, 2002 – December 31st, 2015; daily observations). VIX is the CBOE volatility index, based on the S&P 500 index; VSTOXX is the Euro Stoxx 50 volatility index; ∆VIX is the log return of the VIX index; ∆VSTOXX is the log return of the VSTOXX index; SPX is the S&P 500 stock market index; ESTOXX is the Euro Stoxx 50 stock market index; ∆SPX is the log return of the S&P 500 stock market index; ∆ESTOXX is the log return of the Euro Stoxx 50 stock index; ∆TBILL is the percentage change in the 1-month T-Bill interest rate; ∆EURI- BOR is the percentage change in the 1-month Euribor rate; ∆FFR is the percentage change in the US Fed Funds Rate used as a proxy of the Federal Reserve’s monetary policy stance; ∆EURDEPRATE is the percentage change in the Euro Overnight Deposit rate, taken as a proxy of the monetary policy rate of the European Central Bank. RIVISTA BANCARIA - MINERVA BANCARIA N. 1 - 2 / 2020 23
SAGGI Panel A - Descriptive statistics of VIX and VSTOXX Mean Median Maximum Minimum Std. Dev Skewness Kurtosis Shapiro Francial normality test VIX 0.201 0.175 0.809 0.099 0.092 2.21** 9.92** 0.798*** VSTOXX 0.248 0.222 0.875 0.116 0.103 1.63** 6.10** 0.856*** ΔVIX 0.000 -0.002 0.496 -0.351 0.066 0.70** 7.25** 0.948*** ΔVSTOXX 0.000 -0.003 0.328 -0.249 0.058 0.71** 5.81** 0.961*** Panel B - Correlation matrix of volatility indices, stock indices and interest rates ROSA COCOZZA, DOMENICO CURCIO, ANTONIO PACIFICO VIX VSTOXX SPX ESTOXX ΔVIX ΔVSTOXX ΔSPX ΔESTOXX ΔTBILL ΔEURIBOR ΔFFR ΔEURDEPRATE VIX 1 VSTOXX 0.919*** 1 SPX -0.507*** -0.474*** 1 ESTOXX -0.430*** -0.520*** 0.424*** 1 ΔVIX 0.084*** 0.022 -0.004 0.01 1 ΔVSTOXX 0.071*** 0.080*** -0.004 -0.006 0.534*** 1 ΔSPX -0.130*** -0.062*** 0.029* 0.001 -0.745*** -0.445*** 1 ΔESTOXX -0.115*** -0.115*** 0.029* 0.033** -0.469*** -0.746*** 0.597*** 1 ΔTBILL 0.015 0.005 0.030* -0.008 -0.054*** -0.029* 0.071*** 0.043*** 1 ΔEURIBOR -0.041** -0.037** 0.019 0.048*** 0.014 0.005 -0.019 -0.01 0.030* 1 ΔFFR 0.060*** 0.037** -0.001 -0.003 0.039*** 0.006 -0.078*** -0.029* -0.02 -0.005 1 ΔEURDEPRATE -0.016 -0.011 0.004 -0.023 -0.016 0.006 0.012 0.009 0.004 0.006 -0.026 1 24 ***, ** and * denote significance at the 1%, 5% and 10% level, respectively. Source: our elaborations on data from DatastreamTM.
DO GLOBAL MARKETS IMPLY COMMON FEAR? 4.2. The determinants of the volatility indices: results of the OLS regressions Table 2 presents the results of our OLS estimates of equations (9)-(12), re- ferred to both the VIX (columns 1-4 in Panel A) and VSTOXX (columns 5-8 in Panel B). The negative and statistically significant coefficients of the rate of change in the stock indices 3SI t confirm, for both the European and US mar- ket, the negative correlations between changes in implied volatility indices and underlying stock market returns that we have discussed in the previous section, and are consistent with, among the others, Whaley (2000) and Giot (2005). Positive stock returns correspond to declining implied volatility levels, while neg- ative returns are associated with increasing implied volatility levels. The negative relationship appears much less pronounced for the European market. During the crisis period, the negative relationship between stock indices returns and implied volatility seems to be less negative, since the coefficients of interaction terms CRISIS ) 3SI t are positive and statistically significant in all the specifications reported in Table 2. Since the coefficients of the variable 3SI -t are negative for both the VIX and VSTOXX regressions, our empirical evidence supports the hypothesis of an asymmetric relation between the two volatility indices and the respective stock market indices. VIX and VSTOXX show wider changes when stock prices go down, relative to the case in which stock prices rise. For example (see column 1 in Panel A), if, during the non-crisis years, the S&P500 index rises by 100 basis points, the VIX will fall by almost 4.9%. Should the S&P500 fall by 100 basis points, the VIX will experience an increase of circa 5.16%. Consistently with prior literature, the volatility indices appear to be measures of investors’ fear of the downside risk more than indicators of investors’ excite- ment in a market rally i.e., downward volatility seems to be higher than the up- ward one. According to Giot (2005), the increase in implied volatility is lower in high-volatility periods than in low-volatility periods. In line with this result, we find that, during the years of the great financial crisis, when volatility is higher than that referred to ordinary market conditions, the asymmetric nature of the relation is weaker for the US market: the coefficients of the interaction terms CRISIS ) 3SI -t are all positive and statistically significant at the 1% confidence level. Results are opposite for the European market, where we observe that the RIVISTA BANCARIA - MINERVA BANCARIA N. 1 - 2 / 2020 25
ROSA COCOZZA, DOMENICO CURCIO, ANTONIO PACIFICO coefficients of the interaction terms CRISIS ) 3SI -t are always negative and sta- tistically significant at the 5% confidence level, thus supporting the hypothesis that the asymmetric behavior of the VSTOXX is even stronger during the finan- cial turmoil. The coefficients of the changes in risk-free interest rates are positive for the American market and negative for the European one, respectively, but they are not statistically significant in none of the two cases. The effect of risk-free rates becomes statistically significant during the crisis period with a negative sign for the VIX-S&P 500 relation (see column 4 in Panel A), and a positive sign for the European market (see column 8 in Panel B). As to the impact of monetary policy stance, the coefficient of 3MPR t is neg- ative and statistically significant at the 1% confidence level in the case of the VIX regression (see column 4 in Panel A), which means that VIX decreases when the monetary policy rate increases. Therefore, a restrictive monetary policy is associ- ated with a contemporaneous decrease in implied volatility for the US market. This negative relation, which is consistent with what we would expect to see during periods of ordinary economic and financial conditions, becomes much weaker during the crisis period since the interaction term CRISIS ) 3MPR t is positive and statistically significant at the 1% confidence level. On the contrary, the coefficient of the change in the Euro overnight deposit rate is positive but neither statistically nor economically significant (see column 8 in Panel B), and its interaction term with the dummy CRISIS is negative but significant at only 10% confidence level. 26 SAGGI
Table 2 - OLS regression results This table presents the results of OLS regressions of volatility indices returns on returns of their respective stock market indices, risk-free interest and money market rates. Columns labeled with 1, 2, 3 and 4 in Panel A exhibit results for equations (9), (10), (11) and (12), respectively, when the dependent variable is the log return of the VIX index ); columns labeled with 5, 6, 7 and 8 in Panel B exhibit results for equations (9), (10), (11) and (12), respectively, when the dependent variable is the log return of the VSTOXX index ). is the log return of the stock market indices (S&P 500 for Panel A and EuroStoxx50 for CRISIS Panel B, respectively) and ) 3SI -t is the log return of stock market indices conditional on the market going down andCRISIS 0 otherwise. ) 3IR t is the percentage change in the 4-week T-Bill interest rate for regressions in Panel A, and the percentage change in the 1-month Euribor for regressions in Panel B. 3MPR t is the percentage change in the Fed Funds Rate for regressions in Panel A, and the percentage change in the European overnight deposit rate for regressions in Panel B. CRISIS )is 3 a dummy SI -t variable taking the value of 1 for the observations referred to the crisis years (i.e., from June 2007 onward). Panel A: Dependent variable: 3VIX t Panel B: Dependent variable: 3VSTOXX t 1 2 3 4 5 6 7 8 0.0014* -0.0048*** -0.0048*** -0.0047*** 0.0002 -0.0056*** -0.0056*** -0.0056*** Intercept (1.92) (-4.62) (-4.64) (-4.59) (0.29) (-5.91) (-5.92) (-5.96) -5.0216*** -4.0987*** -4.0986*** -4.0905*** -3.0548*** -2.4451*** -2.4450*** -2.4448*** CRISIS ) 3SI t (-65.80) (-31.38) (-31.39) (-31.32) (-61.08) (-28.65) (-28.65) (-28.66) -0.0028 -0.0009 -0.0012 -0.0012 -0.0019 -0.0077*** -0.0074** -0.0073*** CRISIS ) 3SI -t (-1.42) (-0.34) (-0.44) (-0.44) (-1.05) (-3.22) (-3.06) (-3.03) 2.2389*** 1.6404*** 1.7172*** 1.7031*** 0.5787*** 0.8430*** 0.8696*** 0.8855*** CRISIS ) 3SI t (19.74) (8.37) (8.66) (8.58) (6.11) (5.23) (5.37) (5.46) - -1.8163*** -1.8180*** -1.8284*** - -1.2034*** -1.2043*** -1.2056*** CRISIS ) 3SI -t - ( -8.66) (-8.67) (-8.72) - (-8.74) (-8.74) (-8.76) RIVISTA BANCARIA - MINERVA BANCARIA N. 1 - 2 / 2020 - 1.2098*** 1.1174*** 1.1567*** - -0.4793** -0.5265** -0.5385** CRISIS ) 3SI -t - (3.94) (3.61) (3.71) - (-1.88) (-2.05) (-2.10) - - 0.0008 0.0010 - - -0.0064 -0.0065 CRISIS ) 3IR t - - (0.44) (0.50) - - (-0.47) (-0.48) - - -0.0059** -0.0059** - - 0.2316 0.2746** CRISIS ) 3IR t - - (-2.10) (-2.10) - - (1.77) (2.06) - - - -0.0062*** - - - 0.0004 3MPR t - - - (-2.88) - - - (1.29) - - - 0.0056** - - - -0.0301* CRISIS ) 3MPR t - - - (2.42) - - - (-1.72) Adjusted R2 0.5978 0.6064 0.6069 0.6071 0.5614 0.5772 0.5773 0.5776 F-statistics 1810.00*** 1126.33*** 806.34*** 627.93*** 1559.33*** 997.97*** 713.54*** 555.86 # of obs. 3,653 3,653 3,653 3,653 3,653 3,653 3,653 3,653 T-statistics are reported in parentheses. 27 DO GLOBAL MARKETS IMPLY COMMON FEAR? *** and * denote significance at the 1% and 5% level, respectively. Source: our elaborations on data from DatastreamTM.
ROSA COCOZZA, DOMENICO CURCIO, ANTONIO PACIFICO 5. The determinants of the volatility indices: a SPBVAR evidence 5.1. Econometric Methodology and Specifications The above OLS approach is highly sensitive to departures from normal dis- tribution and suffers from significant limits in jointly analyzing the dynamic of potentially endogenous variables, like the financial variables we are inter- ested in. Therefore, to better investigate the interaction between VIX, VS- TOXX and their drivers, here we use a vector autoregressive model (VAR)4. VAR models are among the best suited approaches to analyze multivariate time series and are typically used to describe the dynamic behavior of eco- nomic and financial variables. A general VAR model has many parameters that may be difficult to inter- pret due to the complex interactions and feedback between the variables in the model. Consequently, in this section, we propose a time-varying Structur- al Panel Bayesian VAR (SPBVAR) model of the form: Y mi,t = U i,0 + | j = 1 U mit,l (B) Y mi,t - j + f i,t p (13) where the subscripts i, l = 1, 2, ..., N are country indices, t = 1, 2, ..., T denotes time, B stands for the lag operator, U i,0 is an NM ) 1 vector of intercepts for each i, U it,l is an NM ) NM matrix of coefficients for each pair of countries (i, l) for a given m, Yi,t - j is an NM ) 1 vector of lagged variables of inter- est for each i for a given m, and f i,t + i.i.d.N (0, | t) is an NM ) 1 vector of disturbance terms. The subscripts j = 1, 2, ..., p are lags for each of the m = 1, 2, ..., M endogenous variables. The model (13) would correspond to a simplified version of the time-vary- ing multicountry SPBVAR model proposed in Pacifico (2019a,b,c). To be more precise, in this analysis, we only consider two country indices (N = 2) and a set of five lagged endogenous variables ( M = 5). Moreover, in this study, in order to examin the dynamic interaction between volatility indices and main financial and macroeconomic variables in US and European stock and volatil- ity markets, we assume the variance-covariance matrix of the vector of errors 4 For applications of VAR models to financial data, see Hamilton (1994) and Campbell, Lo and MacKinlay (1997). 28 SAGGI
DO GLOBAL MARKETS IMPLY COMMON FEAR? | diag (exp (t t ), exp (t 2t)), where t it = (t 1t, t 2t) denotes the time-varying 1t log-volatilities evolving over time according to the following random walk: t t = t t - 1 + r t with r + N (0, | t) (14) where | t is a block diagonal matrix and t 0 denotes the initial conditions to be estimated. The random-walk assumption in (14) is very common in the time-varying VAR literature and has the advantage of focusing on permanent shifts and reducing the number of parameters in the estimation procedure. Moreover, with the hierarchical strategy used to construct the time-varying SPBVAR in (13), one would be able to investigate any type of coefficient factors via their interactions. The variance in f i,t is allowed to be time-variant and it is a way of modelling time-varying conditional second moments to provide an alternative to the stochastic volatility specification. The main difference is that volatility changes are replaced by coefficient changes and the computational costs involved in using this specification are moderate since the high dimen- sionality is avoided via Bayesian inference and Markov Chain Monte Carlo (MCMC) integrations. The variables included in our SPBVAR model are: (i) the natural logarithm of the volatility indices, respectively labeled as logVIX and logVSTOXX, (ii) the US monetary policy interest rate (FFR), (iii) the EU monetary policy interest rate (EURDEPRATE), (iv) the natural logarithm of the underlying stock indices, labeled as logSPX and logESTOXX50, (v) the 18-month for- ward exchange rate. Here, all variables in the system are endogenous and time-varying. The estimation sample covers daily data for the period 2002-2015 and all time-series have persistent memory, but the partial autocorrelation func- tion of each variable is not statistically different from zero over lag 4. Based on conventional Akaike (AIC), Schwarz-Bayesian (BIC) and Hannan-Quinn (HQ) information criteria, we use one lag in our VAR models both for VIX and VSTOXX. The gray shaded areas correspond to 95 percent confidence interval. Generally, the volatility indices have same dynamic reactions to the RIVISTA BANCARIA - MINERVA BANCARIA N. 1 - 2 / 2020 29
ROSA COCOZZA, DOMENICO CURCIO, ANTONIO PACIFICO shocks of all the variables taken into account. In equation (13), the dynamic relationships are allowed to be unit-spe- cific, and all coefficients vary over time. Moreover, if the elements of U t (L) are stacked over i, it is possible to obtain a matrix that is not block-diagonal for at least some p. Thus, cross-unit lagged interdependencies matter, and dynamic feedback and interactions among volatility indices and macroeco- nomic-financial variables are possible. Nevertheless, even if this feature adds flexibility to the specification, it is very costly since the number of coefficients is increased by 6^ N ) M h ) p@ factors. To avoid the curse of dimensionality, by following the framework of Pacifico (2019a,b), a 1 ) k vector X t = (I, Y l mi,t - 1, Y l mi,t - 2, ..., Y l mi,t - p)l can be defined con- taining all lagged variables in the system for each i, where k = 6^ N ) M h ) p@ is the number of all matrix coefficients in each equation of the SPBVAR model for the two country indices. Then, we define an 6 NMk ) 1@ vector c itk ,l = vec (g itk ,l) containing all columns, stacked into a vector5, of the matrix U t (L) for the two country indices for a given k, with g itk ,l = ^ Uli,0 , Ul it1 ,l, Ul it2 ,l, ..., Ul itM,l hl , and c t = ^ cl1t , cl2t hl denoting the time-varying coefficient vectors, stacked for i, for each country-variable pair. With these specifications, we are able to express the model (13) in a simultaneous-equation form: Yt = ^ I NM 7 X t h c t + E t (15) where Yt = (Y l 1mt, Y l 2mt)l and E t = (fl1t , fl2t )l are [NM ) 1] vectors containing the observable variables of interest and the random disturbances of the model for each i for a given m, respectively. In model (15), there is no subscript i since all lagged variables in the system are stacked in X t . Now, because the coefficient vectors in c t vary in different time periods for each country-variable pair and there are more coefficients than data points, it is impossible to eliminate c t . To be more precise, each equation of the time-varying SPBVAR in equation (13) has k = (2 ) 5) p = 10 coefficients, and there are 4,400 equations in the system. Thus, to cope with the problem of dimensionality, we assume that c t has the following factor structure: 5 The vec operator transforms a matrix into a vectorby stacking the columns of the matrix, one underneath the other. 30 SAGGI
Puoi anche leggere