Are CDS spreads predictable during the Covid-19 pandemic? Forecasting based on SVM, GMDH, LSTM and Markov switching autoregression

Darko B Vukovic; Kirill Romanyuk; Sergey Ivashchenko; Elena M Grigorieva

doi:10.1016/j.eswa.2022.116553

Are CDS spreads predictable during the Covid-19 pandemic? Forecasting based on SVM, GMDH, LSTM and Markov switching autoregression

Expert Syst Appl. 2022 May 15:194:116553. doi: 10.1016/j.eswa.2022.116553. Epub 2022 Jan 22.

Authors

Darko B Vukovic^{1

2}, Kirill Romanyuk³, Sergey Ivashchenko^{4

5

6}, Elena M Grigorieva^{1

7}

Affiliations

¹ International Laboratory for Finance and Financial Markets, Faculty of Economics, People's Friendship University of Russia (RUDN University), 117198, Miklukho-Maklaya str. 6, Moscow, Russia.
² Geographical Institute "Jovan Cvijic" SASA, Djure Jaksica 9, 11000 Belgrade, Serbia.
³ Higher School of Economics University, 101000, Myasnitskaya str. 6, Moscow, Russia.
⁴ North-Western Main Branch of the Bank of Russia, 191038, Saint Petersburg Fontanka River Embankment, 68 (107016, Moscow, 12 Neglinnaya Street), Russia.
⁵ Macroeconomic Analysis Division, Financial Research Institute, 127006, Moscow, Nastasyinsky Lane, 3, p. 2, Russia.
⁶ Theoretical Economics Laboratory, The Institute of Regional Economy Studies (Russian Academy of Sciences), 190013, Saint Petersburg, Serpukhovskaya Street 36-38, Russia.
⁷ Finance and Credit Department, Faculty of Economics, People's Friendship University of Russia (RUDN University), 117198, Miklukho-Maklaya str.6, Moscow, Russia.

Abstract

This paper investigates the forecasting performance for credit default swap (CDS) spreads by Support Vector Machines (SVM), Group Method of Data Handling (GMDH), Long Short-Term Memory (LSTM) and Markov switching autoregression (MSA) for daily CDS spreads of the 513 leading US companies, in the period 2009-2020. The goal of this study is to test the forecasting performance of these methods before and during the Covid-19 pandemic and to check whether there are changes in the market efficiency. MSA outperforms all other methods most frequently. GMDH breaks the efficient market hypothesis more frequently (75%) than other methods. The change of the relative predictability during Covid-19 is small with some increase of the advantage of the investigated methods over a benchmark. We find that the market has been less efficient during Covid-19, however, there are no huge differences in prediction performances before and during the Covid-19 period.

Keywords: CDS spreads; Covid-19; GMDH; LSTM; Markov switching autoregression; SVM.