Prediction of Agricultural GSDP of Assam using Cobb Douglas, Constant Elasticity of Substitution and Multiple Linear Regression models

Abstract

This study predicts the agricultural Gross State Domestic Product (GSDP) of Assam, using statistical and econometric models for policy formulation in agricultural economics. We explore various models, including Cobb Douglas model, Constant Elasticity of Substitution model and Multiple Linear Regression model and found that the Multiple Linear Regression offers the best fit based on higher

????², lower MSE, and AIC, along with statistically significant t-values.

Our findings highlight the critical role of agricultural productivity in driving economic growth, enhancing the Gross State Domestic Product (GSDP), and supporting food security and employment. By integrating robust econometric models such as the Cobb-Douglas production function, the Constant Elasticity of Substitution (CES) model, and Multiple Linear Regression (MLR), this study provides valuable empirical insights.

Objectives:

1. To predict the agricultural economy in the state using Machine Learning models.
2. To study the performance of selected model to assess the agricultural economy in the state.

Methods: The methods and models we have used here are Ordinary Least Squares (OLS) and Non- linear Curve Fitting methods are used to estimate the parameters of the models. OLS is applied in the Cobb-Douglas and Multiple Linear Regression models, while Non-linear Curve Fitting is used for the Constant Elasticity of Substitution (CES) model.

Results: Among the employed models, area and labour were identified as significant determinants of Assam's agricultural economic growth. Among the predictive models, the Multiple Linear Regression (MLR) model demonstrated the best fit, as indicated by its highest R² value, lowest Mean Squared Error (MSE), lowest root mean squared error (RMSE), lowest mean absolute percentage error (MAPE) and lowest Akaike Information Criterion (AIC).

Conclusions: This study employed Ordinary Least Squares (OLS) and Non-linear Curve Fitting techniques to estimate parameters for three production function models, Cobb-Douglas, Constant Elasticity of Substitution (CES), and Multiple Linear Regression (MLR). OLS was applied to the Cobb- Douglas and MLR models, while the CES model was estimated using Non-linear Curve Fitting due to its structural complexity.

The analysis identified area and labour as key contributors to Assam's economic growth. Among the models, the Multiple Linear Regression (MLR) model outperformed the others in terms of predictive accuracy, as demonstrated by its highest R², lowest Mean Squared Error (MSE), lowest root mean squared error (RMSE), lowest mean absolute percentage error (MAPE) and lowest Akaike Information Criterion (AIC) values. These results highlight the effectiveness of MLR in capturing the relationship between agricultural inputs and economic output in Assam, making it the most suitable model for policy formulation and future forecasting efforts.