Construction of dynamic factor models for forecasting of economic systems evolution

Olga Katunina
PhD in Economics, Associate Professor, Department of Economics and Mathematical Modeling, Kyiv National Economic University named after Vadym Hetman, Ukraine, e-mail: prommet@ukr.net, ORCID ID: https://orcid.org/0000-0001-7584-0037

Cite this article:

Katunina, O. (2019). Construction of dynamic factor models for forecasting of economic systems evolution. Ed.: M. Zvieriakov (ed.-in-ch.) and others [Pobudova dynamichnykh faktornykh modelei dlia prohnozuvannia rozvytku ekonomichnykh system; za red.: М. I. Zvieriakova (gol. red.) ta in.], Socio-economic research bulletin; Vìsnik socìal’no-ekonomìčnih doslìdžen’ (ISSN 2313-4569), Odessa National Economic University, Odessa, No. 1 (69), pp. 118–127.

Abstract

The simulation of dynamic economic systems whose evolution is described by a system of observable variables is considered in the article. Methodology of dynamic factor modeling was used; mathematical model that combines approaches of classical factor and autoregressive analysis was built. It has been established that the systems of dynamic factors describe the general dynamics of the selected group of economic indicators. An algorithm for constructing a dynamic model, in which dynamic factors are determined sequentially in solving special problems of nonlinear programming, is proposed. The first factor describes the movement of entire system as a whole and characterizes the general trend, since a linear combination of original time series is used to find it. Other factors, which built on the basis of residual series take into account the deviations of individual indicators from their regression estimates and describe fluctuations of time series. The main calculated relationships of the constructed model of dynamic factor analysis are given. For estimate the prediction error, the ex-post forecast method was used. Directions for investigating of the forecast quality by the number of factors taken into account, the lag length, and considered nonlinear programming problems parameters are proposed. It was determined that the choice of these parameters in the constructed algorithm allows minimizing the prediction error for a specific time series and obtaining several possible options for the system development. The developed model of dynamic factor analysis can have a wide practical use, because it opens the possibility to evaluate the impact of forcing a change in the predicted values of one or several indicators on the entire system dynamics. The directions for constructing a controlled multidimensional forecasting model for evolution analyzing of economic dynamic systems of various natures are substantiated. Obtaining a multivariate forecast of economic systems evolution is particularly important in strategic planning both at the macroeconomic level and for individual industries and enterprises. The proposed method of dynamic factor analysis of time series systems has certain universality and, together with other econometric methods, can be used, for example, in ecology, medicine, physics and other fields of science.

Keywords

econometrics; forecasting; physical economic; dynamic factor models; time series; dynamic factor analysis.

JEL classification: С530; E270; DOI: https://doi.org/10.33987/vsed.1(69).2019.118-127

UD classification: 658.14/17:338.24

This work is licensed under a Creative Commons Attribution 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

References

1. Lukyanenko, I. G., Vit, D., Primerova, O. K. at al. (2017). System analysis of the state policy formation in macroeconomic destabilization conditions [Systemnyi analiz formuvannia derzhavnoi polityky v umovakh makroekonomichnoi destabilizacii; za red. I. G. Lukyanenko], Natsionalnyi universytet «Kyievo-Mohilianska akademiia», Kyiv, 463 s., available at: http://ekmair.ukma.edu.ua/handle/123456789/12348 [in Ukrainian]
2. Vitlinsky, V. V. (2003). Economics simulation [Modeliuvannia ekonomiky], KNEU, Kyiv, 408 s. [in Ukrainian]
3. Peña, D., Poncela, P. (2006). Nonstationary dynamic factor analysis. Journal of Statistical Planning and Inference, No. 136, pp. 1237–1257. DOI: https://doi.org/10.1016/j.jspi.2004.08.020.
4. Bai, J., Ng, S. (2008). Large dimensional factor analysis. Foundations and Trends in Econometrics, Vol. 3, No. 2, pp. 89–163. DOI: https://doi.org/10.1561/0800000002.
5. Ajevskis, V., Dāvidsons, G. (2008). Dynamic factor models in forecasting Latvia’s Gross domestic product. Department of the Bank of Latvia, Vol. 2, 24 p.
6. Derbentsev, V. D., Serdyuk, O. A., Solovyov, V. M., Sharapov, O. D. (2010). Synergetic and econophysical methods for studying the dynamic and structural characteristics of economic systems [Sinerhetychni ta ekonofizychni metody doslidzhennia dynamichnykh ta strukturnykh kharakterystyk ekonomichnykh system], Brama-Ukraina, Cherkasy, 287 p. [in Ukrainian]
7. Chernavsky, D. S., Starkov, N. I., Malkov, S. Yu., Kosse Yu. V., Shcherbakov, A. V. (2011). On the econophysics and its place in the modern theoretical economics [Ob ekonomfizike i eye meste v sovremennoy teoreticheskoy ekonomike], Uspekhi Fizicheskikh Nauk, No. 181, s. 767–773, DOI: https://doi.org/10.3367/UFNr.0181.201107i.0767 [in Russian]
8. Stock, J. H., Watson, M. W. (2006). Forecasting with many predictors. Ch. 6 in Handbook of Economic Forecasting. Ed. by Graham Elliott, Clive W. J. Granger and Allan Timmermann Elsevier, pp. 515–554. DOI: https://doi.org/10.1016/S1574-0706(05)01010-4.
9. Lipovina-Božović, M. (2013). A comparison of the VAR model and the PC factor model in forecasting inflation in Montenegro,
10. Economic annals, Vol. LVIII, No. 198, pp. 115–136. DOI: https://doi.org/10.2298/EKA1398115L.
11. Ye, Hua (2011). Macroeconomic forecasting using large vector auto regressive model. Master Thesis in partial fulfillment of the requirements for the degree of Master of Economics and Management Science, Berlin, July 15th, 40 p.
12. Stock, J. H., Watson, M. W. (2002). Macroeconomic forecasting using diffusion indexes, Journal of Business & Economic Statistics, American Statistical Association, Vol. 20, No. 2, pp. 147-162. DOI: https://doi.org/10.1198/073500102317351921.
13. Katunina, O. S. (2014). Forecasting of market saturation processes based on dynamic factor models. [Prohnozuvannia protsesiv nasychennia rynku na bazi dynamichnykh faktornykh modelei], Modeliuvannia ta informatsiini systemy v ekonomitsi, KNEU, Kyiv, Vyp. 90, s. 106–125 [in Ukrainian]
14. Katunina, O. S. (2017). Modeling the dynamics of the world stock indices [Modelirovanie dinamiki mirovykh fondovykh indeksov], Biznes Inform, No. 11, s. 197–202 [in Russian]