Non-bank financial intermediation – a focus on South Africa

Kemp, Esti

Non-bank financial intermediation – a focus on South Africa

Doctoral Thesis

2022

Abstract

We measure the non-bank financial intermediation (NBFI) sector of South Africa over time, and how it is connected to the banking system. While the growth of the NBFI-sector has outpaced that of banks - driven mainly by collective investment schemes, banks continue to hold the largest share of financial assets. We show relatively high levels of interconnectedness between banks and non-banks, specifically investment funds. We also show high levels of portfolio overlap, or indirect interconnectedness, for money market funds (MMFs) registered in South Africa. Given the limited academic work measuring interconnectedness beyond banks, as a second part of the work, a novel dataset is used to analyse the interconnectedness in South Africa in more detail. We propose a method to compute losses on the financial system as a result of a failure of the bank based on look-through exposures - i.e. those beyond the direct and indirect balance sheet links. Specifically, we show that the exposures of financial institutions in the SA financial system to the default of one one of its "big six" banks may be severely underestimated when only considering direct and indirect exposures. The default of one of the big six banks causes financial distress to spread throughout the system. Consequently, additional losses accumulate to institutions over time that are not covered by the direct and indirect exposures. We introduce the higher-order share of exposure (HSE), which expresses what percentage of an exposure is overlooked when only considering direct and indirect exposures. We show that the HSE is close to 100% for a substantial part of the South African financial system, and that in other parts the HSE rises steeply during times of financial distress, when exposures matter most. We show that these higher order losses depend strongly on the network structure of the SA financial system and the robustness of its institutions. In a new domain of estimating exposures, we confirm an earlier established result, which finds that jointly including multiple asset classes and multiple types of financial exposures is requisite to avoid underestimating losses. This highlights the importance of granular data, and network-based modeling approaches that take advantage of these data to properly estimate exposures. The third part of the work focuses on identifying and measuring the financial cycle in South Africa, using three different methodologies. The financial cycle is calculated using credit, house prices and equity prices as indicators, and estimated using traditional turning-point analysis, frequency-based filters and an unobserved components model-based approach. We then consider the financial cycle's main characteristics and examine its relationships with the business cycle. We confirm the presence of a financial cycle in South Africa that has a longer duration and a larger amplitude than the traditional business cycle. Developments in measures of credit and house prices are important indicators of the financial cycle, although the case for including equity prices in the measures is less certain. Periods where financial conditions are stressed are associated with peaks in the financial cycle, suggesting that the estimated financial cycle may have similar leading indicator properties to financial conditions or stress indices. To determine the role of the NBFI sector in the financial cycle, in the final part of the work we also estimate the non-bank credit cycle. This methodology is applied to estimate the non-banking cycle of several economies, to gain insights into differences with the bank credit cycle. We find that the cyclical properties of non-bank credit cycles differ from those of bank credit: while the duration is similar, the amplitude of non-bank credit is relatively larger. The relationship between bank and non-bank credit is not stable and differs among jurisdictions, at a global level this relationship becomes less synchronised in the period leading up to the 2008 financial crisis. We argue that monitoring non-bank credit can bring additional information as a leading indicator for periods of financial instability, in particular currency crises. We complement the existing literature on leading indicators for financial crises by showing that bank credit is a useful indicator for systemic banking crises, while non-bank credit is helpful to predict currency crises, but not vice versa.

Keywords

economics

Reference:

Collections

PhD / Doctoral

Full item page