Abstract
In this paper we use a battery of various mixed-frequency data models to forecast Czech GDP growth. The models employed are mixed-frequency vector autoregressions, mixed-data sampling models, and the dynamic factor model. Using a dataset of historical vintages of unrevised macroeconomic and financial data, we evaluate the performance of these models over the 2005–2014 period and compare them with the Czech National Bank’s macroeconomic forecasts. The results suggest that for shorter forecasting horizons the CNB forecasts outperform forecasts based on the mixed-frequency data models. At longer horizons, mixed-frequency vector autoregressions and the dynamic factor model are able to perform similarly or slightly better than the CNB forecasts. Furthermore, moving away from point forecasts, we also explore the potential of density forecasts from Bayesian mixed-frequency vector autoregressions.
Similar content being viewed by others
Notes
Arnoštová et al. (2010) consider a naïve MA(4) model of GDP growth, the model based on prediction of the expenditure components of GDP used in the CNB for near-term output forecasting, a simple bivariate VAR model of quarterly GDP growth and an aggregated monthly indicator, and a bridge equations model that starts with forecasting of monthly indicators, which are then combined with the GDP series in a VAR model for quarterly data. Apart from these four models, Arnoštová et al. (2010) also examine several factor models based on static principal components and dynamic factors.
Regarding the relative forecasting performance of MIDAS and MF-VAR in output forecasting, Kuzin et al. (2011) show that in the case of nowcasting and forecasting of euro area GDP growth, MF-VAR performs better for longer horizons (5–9 months) while MIDAS outperforms MF-VAR at shorter horizons (1–4 months).
Since nowcasts are computed for several points in a quarter only, the MSE is computed at the same point in the previous four quarters. Next, note that for forecasts the mean squared errors used to compute the weights for quarter \( t_{q} \) are taken for quarters \( t_{q - 2} , \ldots ,t_{q - 5} \) to account for the GDP publication lag.
Two additional schemes for coefficients are considered: a normalized beta polynomial scheme with a zero last lag and a normalized exponential Almon lag polynomial scheme. The results are available upon request.
In our case, using the last vintage data instead of the data available at the time of the forecast does not consistently lower the forecast errors. There is some evidence of a decline in errors for the DFM at short horizons and also for MF-BVAR at medium horizons. Often, however, using last vintage data worsens the forecasting performance, although the differences in the RMSE are mostly rather small. The detailed results are available upon request.
As another robustness check, some coincident indicators—industrial production, construction, and the unemployment rate—were excluded from the set of monthly indicators for forecasting. While coincident indicators are important for nowcasting, their usefulness for forecasting could be questioned. The results of the forecasting performance exercise with the reduced set of monthly indicators are almost unchanged. Implicit weighting of the DFM or explicit weighting based on the MSE seems to be sufficient to deal with the different importance of different monthly indicators for nowcasting and forecasting.
References
Aastveit, K. A., Gerdrup, K. R., Jore, A. S., & Thorsrud, L. A. (2014). Nowcasting GDP in real-time: A density combination approach. Journal of Business & Economic Statistics, 32(1), 48–68.
Andreou, E., Ghysels, E., & Kourtellos, A. (2011). Forecasting with mixed-frequency data. In Clements, M., & Hendry, F. (Eds.), The Oxford handbook of economic forecasting. Oxford University Press, pp 225–245.
Andrle, M., Hlédik, T., Kameník, O., & Vlček, K. (2009). Implementing the new structural model of the Czech National Bank. Czech National Bank Working Paper 2/2009, Czech National Bank, Prague
Armesto, M., Engemann, K., & Owyang, M. (2010). Forecasting with mixed frequencies. The Federal Reserve Bank of St. Louis Review, 92(6), 521–536.
Arnoštová, K., Havrlant, D., Růžička, L., & Tóth, P. (2010). Short-term forecasting of Czech quarterly GDP using monthly indicators. Czech National Bank Working Papers, No. 12/2010, Czech National Bank, Prague
Bache, I. W., Brubakk, L., Jore, A. S., Maih, J., & Nicolaisen, J. (2010). Monetary policy analysis in practice—a conditional forecasting approach. Norges Bank Monetary Policy Staff Memo, No. 8/2010.
Banbura, M., Giannone, D., & Lenza, M. (2015). Conditional forecasts and scenario analysis with vector autoregressive models for large cross-sections. International Journal of Forecasting, 31(3), 739–756.
Banbura, M., & Modugno, M. (2014). Maximum likelihood estimation of factor models on datasets with arbitrary pattern of missing data. Journal of Applied Econometrics, 29(1), 133–160.
Brůha, J., Hlédik, T., Holub, T., Polanský, J., & Tonner, J. (2013). Incorporating judgments and dealing with data uncertainty in forecasting at the Czech National Bank. Czech National Bank Research and Policy Note, No. 2/2013, Czech National Bank, Prague
Camacho, M., & Perez-Quiros, G. (2010). Introducing the Euro-sting: Short-term indicator of Euro area growth. Journal of Applied Econometrics, 25(4), 663–694.
Canova, F. (2007). Methods for applied macroeconomic research. Princeton, NJ: Princeton University Press.
Chiu, C. W., Eraker, B., Foerster, A. T., Kim, T. B., & Seoane, H. D. (2012). Estimating VAR’s sampled at mixed or irregular spaced frequencies: A Bayesian approach. The Federal Reserve Bank of Kansas City Research Working Paper, No. 11-11, Federal Reserve Bank of Kansas City, Kansas City
Clements, M. P., & Galvão, A. B. (2008). Macroeconomic forecasting with mixed-frequency data: Forecasting output growth in the United States. Journal of Business & Economic Statistics, 26(4), 546–554.
Clements, M. P., & Galvão, A. B. (2009). Forecasting US output growth using leading indicators: An appraisal using MIDAS models. Journal of Applied Econometrics, 24, 1187–1206.
Coats, W., Laxton, D., & Rose, D. (2003). The Czech National Bank’s forecasting and policy analysis system. Prague: Czech National Bank.
Foroni, C., & Marcellino, M. (2013). A survey of econometrics methods for mixed-frequency data. Norges Bank Working Paper, No. 2013/06, Norges Bank, Oslo
Foroni, C., & Marcellino, M. (2014). A comparison of mixed frequency approaches for nowcasting Euro area macroeconomic aggregates. International Journal of Forecasting, 30, 554–568.
Foroni, C., Schumacher, C., & Marcellino, M. (2015). Unrestricted mixed data sampling (MIDAS): MIDAS regressions with unrestricted lag polynomials. Journal of the Royal Statistical Society: Series A (Statistics in Society), 178(1), 57–82.
Ghysels, E. (2016). Macroeconomics and the reality of mixed frequency data. Journal of Econometrics, 193(2), 294–314.
Ghysels, E., Santa-Clara, P., & Valkanov, R. (2004). The MIDAS touch: Mixed data sampling regressions. CIRANO Working Paper, No. 2004s-20, CIRANO, Montreal
Ghysels, E., Sinko, A., & Valkanov, R. (2007). MIDAS regressions: Further results and new directions. Econometric Reviews, 26(1), 53–90.
Giannone, D., Reichlin, L., & Small, D. (2008). Nowcasting: the real-time informational content of macroeconomic data. Journal of Monetary Economics, 55(4), 665–676.
Havranek, T., Horvath, R., & Mateju, J. (2012). Monetary transmission and the financial sector in the Czech Republic. Economic Change and Restructuring, 45, 135–155.
Kascha, C., & Ravazzolo, F. (2010). Combining inflation density forecasts. Journal of Forecasting, 29(1–2), 231–250.
Kuzin, V., Marcellino, M., & Schumacher, C. (2011). MIDAS vs. mixed-frequency VAR: Nowcasting GDP in the Euro area. International Journal of Forecasting, 27, 529–542.
Marcellino, M., Porqueddu, M., & Vendetti, F. (2013). Short-term GDP forecasting with a mixed frequency dynamic factor model with stochastic volatility. Journal of Business and Economic Statistics, 34(1), 118–127.
Marcellino, M., & Schumacher, C. (2010). Factor-MIDAS for now—and forecasting with ragged-edge data: A model comparison for German GDP. Oxford Bulletin of Economics and Statistics, 72, 518–550.
Mariano, R. S., & Murasawa, Y. (2003). A new coincident index of business cycles based on monthly and quarterly series. Journal of Applied Econometrics, 18, 427–443.
Mariano, R. S., & Murasawa, Y. (2010). A coincident index, common factors, and monthly real GDP. Oxford Bulletin of Economics and Statistics, 72(1), 27–46.
Mitchell, J., & Hall, S. G. (2005). Evaluating, comparing, and combining density forecasts using the KLIC with an application to the Bank of England and NIESR “fan” charts of inflation. Oxford Bulletin of Economics and Statistics, 67, 995–1033.
Rusnák, M. (2013a). Nowcasting Czech GDP in real time, Czech National Bank Working Papers, No. 6/13, Czech National Bank, Prague
Rusnák, M. (2013b). Revisions to the Czech National Accounts: Properties and predictability. Czech Journal of Economics and Finance (Finance a uver), 63(3), 244–261.
Schorfheide, F., & Song, D. (2013). Real-time forecasting with a mixed frequency VAR. NBER Working Paper, No. 19712.
Sims, C. A. (2002). The role of models and probabilities in the monetary policy process. Brookings Papers on Economic Activity, 33(2), 1–62.
Acknowledgments
We acknowledge support from the Czech National Bank (Project No. B6/13), Rusnák acknowledges support from Grant Agency of Charles University (#888413) and the Grant Agency of Czech Republic (Grant p402/12/G097). We thank Marta Bańbura and Eric Ghysels for sharing parts of their Matlab codes. We also thank Oxana Babecká Kucharčuková, Claudia Foroni, Ana Beatriz Galvão, seminar participants at the Czech National Bank and two anonymous referees for their helpful comments. The views expressed here are those of authors and not necessarily those of the Czech National Bank.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Franta, M., Havrlant, D. & Rusnák, M. Forecasting Czech GDP Using Mixed-Frequency Data Models. J Bus Cycle Res 12, 165–185 (2016). https://doi.org/10.1007/s41549-016-0008-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s41549-016-0008-z