Utility of Smoothing Techniques in Yield Curve Modeling for the Asian Pacific Frontier Capital Market

Traditional yield curve models, such as the Nelson–Siegel parsimonious model, excel in accurately depicting the interest rate term structure when applied to smoothly evolving yield curve data. Nelson–Siegel's model has garnered widespread international adoption due to the meaningful economic interpretations it provides for its parameters. In the context of developed capital markets, where abrupt economic fluctuations are infrequent, the smoothness of yield curve data presents no significant hurdles in the modeling of interest rate curves. However, Asian Frontier countries that include Sri Lanka, Pakistan, Bangladesh, and Vietnam used to face economic downturns frequently and the impact of such conditions influenced the country substantially. Adopting the most flexible and interpretable yield curve model such as the aynamic Nelson–Siegel (NS) model is a challenge under such conditions. The yield curve data from January 2010 to 2022 was examined and clustered into steady-state and non-steady-state data based on inflation and exchange rate movement for each country. The accuracy of the Nelson–Siegel model was observed to decline in non-steady-state conditions compared to steady-state scenarios, as indicated by lower R-squared values and higher mean absolute deviation (MAD). Furthermore, the data exhibited varying degrees of smoothness between these two states, with a higher degree of smoothness observed in steady-state conditions based on the autocorrelation function. Several smoothing techniques, including spline smoothing, super smoothing, and LOWESS smoothing, were applied to both steady-state and non-steady-state datasets. LOWESS smoothing and spline smoothing appear to excel in achieving smoother results for steady-state data, whereas super smoothing and spline smoothing prove more effective in non-steady-state situations. Notably, spline smoothing outperformed other smoothing techniques in model accuracy tests, such as R-squared and MAD, suggesting that smoothing factor alone cannot be used to decide the selection of suitable smoothing tools for NS modeling. The paper proposed an algorithm to use in modeling yield curves for Asia–Pacific Frontier markets.