Supply chain finance has become an interesting research topic which attracts lots of attention from scholars recently, particularly after the global financial crisis. However, only few studies have examined the causal relationship between supply chain finance and firm performance. More specially, there is a big research gap when almost none of existing research has analysed the nonlinear impact of supply chain finance on firm performance. With this aim, this paper succeeds in giving first empirical evidence on the U-shaped nonlinear relationship between supply chain finance and the performance of seafood firms in Vietnam. Specifically, a bad performance of supply chain finance (the increase in cash conversion cycle - CCC) causes a lower firm performance (FP). Nevertheless, if any decrease in firm performance reaches its minimum (CCC*), the restructuring of the firm will gradually improve it. In addition, firm performance is significantly influenced by controlled variables of firm-specific, firm size (SIZE) and capital structure (CAP), and macroeconomic, economic growth (EG), factors. The findings are valuable for the management as well as scholars in bringing a more comprehensive perspective on the causal relationship between supply chain finance and firm performance.

One time series model developed to predict economic data is Spatial Vector Autoregressive with Exogenous Variables (SpVARX). This model can accommodate simultaneously the interrelationships between variables, the impact of exogenous variables, and Inter-regional linkages. However, this model has not adjusted the nonlinearity relationships between variables. The relationship between economic variables is usually not linear. Therefore, we introduce the Threshold Spatial Vector Autoregressive with exogenous variables (TSpVARX). This paper assesses the forecasting performance of TSpVARX and compares it with SpVARX models. We conducted a simulation study by generating 100 times the simulation data with twelve scenarios. We found that the forecasting performance of the TSpVARX model is better than SpVARX when there is a nonlinear relationship between variables. In addition, we find that the forecasting performance of TSpVARX models will improve as the sample size increases.