Yield curve and business cycle dynamics in South Africa: new evidence from a Markov switching model
- Authors: Rotich, Mercyline Chepkemoi
- Date: 2024-04-03
- Subjects: Uncatalogued
- Language: English
- Type: Academic theses , Master's theses , text
- Identifier: http://hdl.handle.net/10962/434739 , vital:73101
- Description: Globally, several empirical studies have demonstrated the ability of the yield spread to predict a recession in a country. In South Africa, previous studies have not only shown the yield curve's predictive power but have further demonstrated that it outperforms other commonly used variables, such as the growth rate of real money supply, changes in stock prices, and the index of leading economic indicators. However, some recent studies have shown that the yield spread (the spread between 10-year bonds and 3-month Treasury bills) gave false signals of recession. In this study, we explore the possible reasons for the false signals of the yield spread by addressing the following questions. Does the yield spread used matter? Does the measure of the business cycle used matter? And do the estimation techniques used matter? To address the first question, unlike the previous studies, this paper uses four different yield spreads- depicting short-term, medium-term, and long-term government bonds against the backdrop of a changing structure of bond holding, which reflects the increasing risk eversion of investors in South Africa. Second, the paper used different measures of business cycles, namely industrial production index, lagging, coincident, and leading economic indicators. The empirical models were estimated using both univariate and multivariate Markov switching models. As economic theory suggests, the univariate Markov switching model was used to determine if each variable exhibits a significant regime switching. The multivariate Markov switching model was estimated for each business cycle and yield spread variable, with each of the other variables serving as a non-switching explanatory variable, thereby addressing potential endogeneity concerns and the predictive power of the explanatory variable. Finally, the multivariate Markov switching model was estimated for three monthly sample periods, a full sample for 1986 to 2022, and two sub-samples – 1986 to 2009 and 2010 to 2022. This analysis consistently reveals significant regime-switching behavior across all the series thus, affirming the superiority of the regime switching model over the standard model used in previous studies. By analyzing the transition probabilities and the expected durations between these regimes, we find that including the spreads in the business cycle model improves the models’ predictability, with the medium-term bonds spread performing better than the usual long-term spread. The smoothed regime probability of the best-performing models is compared with the SARB recession dates; the two closely resemble each other, proving that the Markov switching model can help predict the turning points in the business cycle in South Africa. , Thesis (MCom) -- Faculty of Commerce, Economics and Economic History, 2024
- Full Text:
- Date Issued: 2024-04-03
- Authors: Rotich, Mercyline Chepkemoi
- Date: 2024-04-03
- Subjects: Uncatalogued
- Language: English
- Type: Academic theses , Master's theses , text
- Identifier: http://hdl.handle.net/10962/434739 , vital:73101
- Description: Globally, several empirical studies have demonstrated the ability of the yield spread to predict a recession in a country. In South Africa, previous studies have not only shown the yield curve's predictive power but have further demonstrated that it outperforms other commonly used variables, such as the growth rate of real money supply, changes in stock prices, and the index of leading economic indicators. However, some recent studies have shown that the yield spread (the spread between 10-year bonds and 3-month Treasury bills) gave false signals of recession. In this study, we explore the possible reasons for the false signals of the yield spread by addressing the following questions. Does the yield spread used matter? Does the measure of the business cycle used matter? And do the estimation techniques used matter? To address the first question, unlike the previous studies, this paper uses four different yield spreads- depicting short-term, medium-term, and long-term government bonds against the backdrop of a changing structure of bond holding, which reflects the increasing risk eversion of investors in South Africa. Second, the paper used different measures of business cycles, namely industrial production index, lagging, coincident, and leading economic indicators. The empirical models were estimated using both univariate and multivariate Markov switching models. As economic theory suggests, the univariate Markov switching model was used to determine if each variable exhibits a significant regime switching. The multivariate Markov switching model was estimated for each business cycle and yield spread variable, with each of the other variables serving as a non-switching explanatory variable, thereby addressing potential endogeneity concerns and the predictive power of the explanatory variable. Finally, the multivariate Markov switching model was estimated for three monthly sample periods, a full sample for 1986 to 2022, and two sub-samples – 1986 to 2009 and 2010 to 2022. This analysis consistently reveals significant regime-switching behavior across all the series thus, affirming the superiority of the regime switching model over the standard model used in previous studies. By analyzing the transition probabilities and the expected durations between these regimes, we find that including the spreads in the business cycle model improves the models’ predictability, with the medium-term bonds spread performing better than the usual long-term spread. The smoothed regime probability of the best-performing models is compared with the SARB recession dates; the two closely resemble each other, proving that the Markov switching model can help predict the turning points in the business cycle in South Africa. , Thesis (MCom) -- Faculty of Commerce, Economics and Economic History, 2024
- Full Text:
- Date Issued: 2024-04-03
Systematic effects and mitigation strategies in observations of cosmic re-ionisation with the Hydrogen Epoch of Reionization Array
- Authors: Charles, Ntsikelelo
- Date: 2024
- Subjects: Uncatalogued
- Language: English
- Type: Academic theses , Doctoral theses , text
- Identifier: http://hdl.handle.net/10962/432605 , vital:72886 , DOI 10.21504/10962/432605
- Description: The 21 cm transition from neutral Hydrogen promises to be the best observational probe of the Epoch of Reionisation (EoR). It has driven the construction of the new generation of lowfrequency radio interferometric arrays, including the Hydrogen Epoch of Reionization Array (HERA). The main difficulty in measuring the 21 cm signal is the presence of bright foregrounds that require very accurate interferometric calibration. However, the non-smooth instrumental response of the antenna as a result of mutual coupling complicates the calibration process by introducing non-smooth calibration errors. Additionally, incomplete sky models are typically used in calibration due to the limited depth and resolution of current source catalogues. Combined with the instrumental response, the use of incomplete sky models during calibration can result in non-smooth calibration errors. These, overall, impart spectral structure on smooth foregrounds, leading to foreground power leakage into the EoR window. In this thesis we explored the use of fringe rate filters (Parsons et al., 2016) as a mean to mitigate calibration errors resulting from the effects of mutual coupling and the use of an incomplete sky model during calibration. We found that the use of a simple notch filter mitigates calibration errors reducing the foreground power leakage into the EoR window by a factor of ∼ 102. Thyagarajan et al. (2018) proposed the use of closure phase quantities as a means to detect the 21 cm signal, which has the advantage of being independent (to first order) from calibration errors and, therefore, bypasses the need for accurate calibration. In this thesis, we explore the impact of primary beam patterns affected by mutual coupling on the closure phase. We found that primary beams affected by mutual coupling lead to a leakage of foreground power into the EoR window, which can be up to ∼ 104 times and is mainly caused by the unsmooth spectral structure primary of primary beam sidelobes affected by mutual coupling. This power leakage was confined to k < 0.3 pseudo h Mpc−1. Lastly, we also proposed and demonstrated an analysis technique that can be used to derive a flux scale correction in post-calibrated HERA data. We found that after applying flux scale correction to calibrated HERA data, the bandpass error reduces significantly, with an improvement of 6%. The derived flux scale correction was antenna-independent, and it can be applied to fix the overall visibility spectrum scale of H4C data post-calibration in a fashion similar to Jacobs et al. (2013). , Thesis (PhD) -- Faculty of Science, Physics and Electronics, 2024
- Full Text:
- Date Issued: 2024
- Authors: Charles, Ntsikelelo
- Date: 2024
- Subjects: Uncatalogued
- Language: English
- Type: Academic theses , Doctoral theses , text
- Identifier: http://hdl.handle.net/10962/432605 , vital:72886 , DOI 10.21504/10962/432605
- Description: The 21 cm transition from neutral Hydrogen promises to be the best observational probe of the Epoch of Reionisation (EoR). It has driven the construction of the new generation of lowfrequency radio interferometric arrays, including the Hydrogen Epoch of Reionization Array (HERA). The main difficulty in measuring the 21 cm signal is the presence of bright foregrounds that require very accurate interferometric calibration. However, the non-smooth instrumental response of the antenna as a result of mutual coupling complicates the calibration process by introducing non-smooth calibration errors. Additionally, incomplete sky models are typically used in calibration due to the limited depth and resolution of current source catalogues. Combined with the instrumental response, the use of incomplete sky models during calibration can result in non-smooth calibration errors. These, overall, impart spectral structure on smooth foregrounds, leading to foreground power leakage into the EoR window. In this thesis we explored the use of fringe rate filters (Parsons et al., 2016) as a mean to mitigate calibration errors resulting from the effects of mutual coupling and the use of an incomplete sky model during calibration. We found that the use of a simple notch filter mitigates calibration errors reducing the foreground power leakage into the EoR window by a factor of ∼ 102. Thyagarajan et al. (2018) proposed the use of closure phase quantities as a means to detect the 21 cm signal, which has the advantage of being independent (to first order) from calibration errors and, therefore, bypasses the need for accurate calibration. In this thesis, we explore the impact of primary beam patterns affected by mutual coupling on the closure phase. We found that primary beams affected by mutual coupling lead to a leakage of foreground power into the EoR window, which can be up to ∼ 104 times and is mainly caused by the unsmooth spectral structure primary of primary beam sidelobes affected by mutual coupling. This power leakage was confined to k < 0.3 pseudo h Mpc−1. Lastly, we also proposed and demonstrated an analysis technique that can be used to derive a flux scale correction in post-calibrated HERA data. We found that after applying flux scale correction to calibrated HERA data, the bandpass error reduces significantly, with an improvement of 6%. The derived flux scale correction was antenna-independent, and it can be applied to fix the overall visibility spectrum scale of H4C data post-calibration in a fashion similar to Jacobs et al. (2013). , Thesis (PhD) -- Faculty of Science, Physics and Electronics, 2024
- Full Text:
- Date Issued: 2024