Cotton is an important commercial crop in India. The present study focuses on measurement of variability pattern of cotton yield and use of principal component analysis for developing cotton yield forecast model for Hisar district of Haryana (India). Instability index has been observed to study the variability behavior of cotton yield in the district. Time series data on cotton yield and fortnightly data of five weather variables for the crop season for 38 years (1980-91 to 2017-18) have been used. In all, three models have been developed by using direct weather variables, PC scores and components with higher loading as regressors and developed models have been used to forecast yield for four subsequent years 2014-15 to 2017-18 (which were not included in model development). The model with PC scores was found to be most appropriate to provide reliable yield forecast.