The present study entitled “Yield forecasting of groundnut in Bihar through Auto-
Regressive Integrated Moving Average (ARIMA) models” is based on the ARIMA models
for forecasting groundnut yield in Bihar. The secondary data on groundnut yield were
collected from the year 1980 to 2018 from the Directorate of Groundnut Research,
Directorate of Oilseeds Development and India Agril. Stat. The data from 1980 to 2016 were
used for analysis of forecasting groundnut yield and the data for 2017 to 2018 were kept for
model validation. Instead of conventional or econometric methods, the ARIMA models were
used to forecast the productivity of groundnut in Bihar. The time series data of 37 years from
1980 to 2016 were used for the study. Models ARIMA (0,1,1), ARIMA (0,1,2), ARIMA
(0,0,1), ARIMA (1,0,0), ARIMA (1,0,1), ARIMA (1,1,1), ARIMA (2,0,0) and ARIMA
(2,0,1) were built. The parameters of all these models were computed and tested for their
significance. Various statistics were also computed for selecting the adequate and
parsimonious model i.e., t-test and chi-square test. This is supported by low values of MAPE,
MAE, RMSE and BIC for forecasting of groundnut yield in Bihar. Using the selected
ARIMA models ARIMA (1,0,1) the yield values were forecasted for five five-year period
ahead i.e. from 2017 to 2021 in Bihar. The forecasted values of Bihar are 1026.72 kg/ ha,
1028.74 kg/ ha, 913.00 kg/ ha, 913.00 kg/ ha and 913.00 kg/ ha, respectively for 2017, 2018,
2019, 2020 and 2021. The forecasted values of Bihar exhibit an increasing trend, for 2017
and 2018, in the yield of groundnut. These yield values were presented along with their lower
and upper limits with 95% confidence interval. Using the mathematically sound ARIMA
models, the groundnut yield values were forecasted with 0.90 percent of one step ahead
forecast errors for Bihar. The two steps ahead forecast errors are 1.69 per cent for Bihar. All
the 8 models were subjected to critical examination. Among them ARIMA (1,0,1) model was
chosen as it is stationary, invertible, parsimonious, stable and has minimum error. Thus, the
forecast model for groundnut productivity in Bihar is,

Z t – Z t-1 = 6.879 + 0.855 (z t-1 – z t-2 ) – 0.354 (a t-1 – a t-2 ) + a t