Classic panel data modelling has large cross section data (N) and small time series data (T). The aim of the study was to develop a panel data type T > N by adding space-time effect to a panel ARDL model. The basic idea was to combine between the AR (Autoregressive), DL (Distributed Lag), and ST (space-time) effect. The model was applied to paddy producer price at the farmer level in Java from January 2016 to December 2019 where the explanatory variable was the Farmers’ Terms of Trade. Both variables were stationary in the first-difference I (1). The results showed that the ST-ARDL model was good for T > N panel data types. The ST-ARDL model with reparameterization of explanatory variables was able to overcome the problem of multicollinearity. The ST-ARDL model was able to improve the performance of the classic panel data model which was able to reduce the RMSE and increased R2-adj. The linear combination of this model was cointegrated or had a long-term equilibrium relationship. Another result of the study was the ST-ARDL model provided better estimation performance than the AR (p), ARDL (p, q) and GSTAR (p, λ) models with the smaller MAPE values. For further research, the ST-ARDL model can be developed by adding the effect of space-time interaction.