Digital economy era: the role of the telecommunications sector in frequency-dependent default risk connectedness

Abstract

We use frequency-dependent connectedness measures to study the role played by the telecommunications (telecoms) sector in sectoral default risk connectedness at three frequency bands, i.e. the short-, medium-, and long-term financial cycles. We extend credit risk spillovers analysis from the time domain to the frequency domain. Our findings indicate that investors in the global CDS sector index market have different investment horizons, but they prefer to process default risk information mainly within one week. In the within-region analysis, except for North America, the telecoms sector plays a significant role in transmitting net credit risk to the other sectors, especially in the short-term financial cycle. In the cross-region analysis, the European telecoms sector is the major net default risk transmitter on all three frequency bands. Our study has noteworthy empirical implications for consumption-based asset pricing models, cross-sector credit risk connectivity, and regional financial stability.

Keywords:

• Telecommunications sector
• CDS sector index
• Frequency-dependent connectedness
• Frequency domain
• Systemic risk

JEL Classification:

• G01
• G15

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 ‘EnviroTech’ refers to Environmental Technology; ‘FinTech’ refers to Financial Technology; ‘MedTech’ refers to Medical Technology; and ‘EdTech’ refers to Education Technology.

2 The R software package ‘frequencyConnectedness’ is kindly provided by Barunik and Krehlik, and it is available at CRAN or https://github.com/tomaskrehlik/frequencyConnectedness.

3 Time domain and frequency domain (or spectral analysis) are the two major methods used in a time series analysis.

4 For the same purpose, other data cleaning approaches could be adopted. For example, Eichengreen et al. (2012) use the weekly averaged CDS data to smooth out sharp daily fluctuations and irregular trading.

5 Using the generalised variance decompositions method ensures that variance decompositions is invariant to the ordering of variables in the VAR model.

6 On the frequency band d, Barunik and Krehlik (2018) define a within connectedness measure as: $C_d^W=\left(\left(\sum _{k,j=1,k\ne j}^N{\left({\tilde{\theta }}_d\right)}_{j,k}\right)/\left(\sum _{k,j=1}^N{\left({\tilde{\theta }}_d\right)}_{j,k}\right)\right)\times 100$. The relation between $C_d^W$ and $C_d^F$ is: $C_d^F=C_d^W\left(\right(\sum _{k,j=1}^N{\left({\tilde{\theta }}_d\right)}_{j,k}\left)/\right(\sum _{k,j=1}^N{\left({\tilde{\theta }}_{\mathrm{\infty }}\right)}_{j,k}\left)\right)$. When $d=\left(-\pi ,\pi \right)$, $C_d^W=C_d^F=C_{\mathrm{\infty }}$ and it is the Diebold and Yilmaz’s (2012) case; when $d\subseteq \left(-\pi ,\pi \right)$, $C_d^W\ge C_d^F$. $C_d^W$ is weighted by the power of one asset in the system at the given time-frequency, and it represents the risk connectedness within that time-frequency. $C_d^F$ disintegrates the total risk connectedness, $C_{\mathrm{\infty }}$, into different components, e.g., short- and long-term parts.

7 The levels of risk connectedness are not significantly affected by the choices of order of VAR, which is in line with the findings of Diebold and Yilmaz (2012) and Barunik and Krehlik (2018).

8 As mentioned in footnote 6, we also measure ‘within connectedness’ within each time-frequency. However, for the purpose of interpretation and brevity, we only report the results of ‘frequency connectedness’. The results of ‘within connectedness’ are available upon request.

9 Due to data availability, we could not conduct quantitative analysis to identify which factor(s) drive the sectoral default risk transmission within each region.