![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
Abstract
We analyse the impact of news on five financial markets in the Czech Republic, Hungary and Poland using a newly constructed data set in a Generalized Autoregressive Conditional Heteroscedastic (GARCH) framework. Macroeconomic shocks (on Gross Domestic Product (GDP), inflation rate, current account and trade balance) are constructed as deviations from expected values. Economic and Monetary Union (EMU)-related political and fiscal news is captured as news dummies. Macroeconomic shocks significantly affect short-term interest rates and, to a lesser extent, other financial variables. Political and fiscal news has an impact on long-term bond yields and exchange rates. News displayed prominently in our media sources has a greater impact on financial markets than other news and, in addition, the sources of news themselves matter. We also discover asymmetric effects of news within markets. Finally, using a pooled GARCH model we find that macroeconomic shocks have the strongest impact on financial markets in Hungary, while political news has the largest influence in both Hungary and Poland.
Notes
1 As of July 2008, none of the CEEC-3 has entered the ERM II The ERM II, which is sometimes referred to as the ‘ante-chamber of euro adoption’, is a hybrid exchange rate regime, where currencies are allowed to fluctuate by 15% around a central parity against the euro. Before adopting the euro, a country has to remain in the ERM II for 2 years.
2 The figures on the CEEC-3 are taken from European Commission (Citation2006).
3 The Czech National Bank intervenes occasionally to reduce the volatility of the exchange rate. There is no explicit exchange rate target. For the exchange rate arrangements of the CEEC-3, see IMF (2006). For a more thorough treatment of exchange rate policy in the CEEC-3 (and Slovakia), see Kočenda and Valachy (2006).
4 This exchange rate regime was replaced by a free float on 26 February 2008.
5 For instance, in January 2003, the Hungarian central bank (MNB) had to defend the upper band of the exchange rate and later in the year the lower band, resulting in massive interest rate hikes.
6 The inflation rate fluctuated between 3.6% in 2004 and 1.3% in 2006.
7 Onder and Simga-Mugan (Citation2006) and Andritzky et al. (Citation2007) summarize the relevant literature on the effects of news for both mature and emerging financial markets.
8 In a recent study, Chaney (Citation2008) evaluates political events concerning Iraq using bond spreads versus US bonds.
9 Moreover, the authors do not include the first 15 and last 5 minutes of a trading day as those periods are associated with excess volatility. This practically excludes any effects of news published when markets are closed.
10 Another reason is that in these relatively young democracies, the political spectrum is far from settled. Changes in government are more frequent than in Western Europe and political parties are less oriented towards the political centre. Thus changes in governments have a larger impact on financial markets as they entail larger swings in the policy stance.
11 We performed Augmented Dickey–Fuller (ADF) tests and all but one time series (3 month Hungarian interest rates) had unit roots. These results are available upon request.
12 The Emerging Market Bond Index (EMBI) is calculated by JP Morgan and captures the spread between the yield on a dollar-denominated sovereign bond and the yield of a comparable bond issued by the US Treasury.
13 See and in the Appendix. Using a methodology applied by, inter alia, Joyce and Read (Citation1999), Gavin and Mandal (Citation2001) as well as Gravelle and Moessner (Citation2001), we confirm that our consensus forecasts are generally unbiased and efficient. That is, forecast errors are mean zero uncorrelated with their own past and future values as well as with actual and forecasted past and future values. Detailed results are available upon request. Note that for current account and trade balance data, the consensus forecasts are taken from the Interfax Business Reports. The base of these consensus forecasts is not consistent and therefore results should be treated with caution.
14 For the CEEC-3 this effect might be even larger as the so-called ‘convergence play’ increases the inflow of portfolio capital.
15 The following section discusses positive macroeconomic shocks, i.e. actual figures are above expected values and positive fiscal and political news. In the case of negative shocks and news, the effects outlined take the opposite direction.
16 Pearce and Roley (Citation1985) discuss other channels of how shocks to the inflation rate may affect stock markets.
17 However, in the medium term, this policy would also increase output and along with this imports.
18 Otherwise, we use OLS and add a maximum of 10 dummies capturing outliers to approximate a normal distribution. These dummies do not noticeably affect the estimates or significance of our variables of interest.
19 Depending on the outcome of the general-to-specific modelling process, in the reduced models we also apply OLS, Asymmetric GARCH (AGARCH), GARCH-in-Mean (GARCHM), Asymmetric GARCH-in-Mean (AGARCHM) models. See , 2, 3 and 6.
20 This applies to the GARCH specification.
21 In a few specifications, we had to ensure stationarity of the conditional variance by imposing the restriction α 1 + β 1 ≤ 1, leaving some of the parameters insignificant. These findings may also indicate the presence of integrated GARCH processes (see Nelson, Citation1990). On 6 February 2004, the Czech 3 month interest rate increased by almost 50 basis points without any apparent reason and on 14 January 2004 bond yields rose by 40 basis points in Poland. Since the convergence algorithms of our GARCH specifications were seriously affected by these outliers, we added additional impulse dummies for these days.
22 Note that in one case (3 month interbank interest rates in Poland) the asymmetry term is insignificant. It remains in the model as it cannot be eliminated in a consistent testing-down process.
23 Results omitted but available upon request. We also do not find any signs for heteroscedasticity in the OLS models. In some instances, we find evidence of nonnormality. We are able to remove this nonnormality by including dummy variables to capture a small number of outliers and can show that our estimation results are robust to these changes. Note that the estimates for the bond spread of the Czech Republic show mild signs of ARCH in the final model. Since these signs appeared only after removing nonnormality, we think that these ARCH effects are spurious and we choose to stick to the presented OLS model for reasons of estimation efficiency. In any case, the presented parameter estimates would not change in a noteworthy way.
24 See Tables for diagnostic statistics.
25 Note that the total number of news in differs from those in the previous tables since some news could not be clearly assigned to one of our source categories. See Tables for diagnostic statistics.
26 The Polish government bond spread also demonstrates the strongest asymmetric effects to fiscal and political news with around three basis points each. In other markets in the CEEC-3 asymmetric effects do not rise above one basis point (for spreads and interest rates) or one percentage point (stocks and exchange rates).
27 With: Chi2(1) = 4.9 (Polish zloty/euro exchange rate) and Chi2(1) = 4.5 (12 month interest rate).
28 The results are available upon request.
29 Note, however, that in this model the procedure differs from the estimation of Equation Equation1 insofar as the general-to-specific approach was applied to the whole model, i.e. including the news variables. Moreover, TB shocks are not included since they were not available for all countries. CA news is included only as news dummies for the same reason.