Missing data is a problem that occurs frequently in survey data. Missing data results in biased estimates and reduced efficiency for regression estimates. The objective of the current study is to analyze the impact of missing-data imputation, using multiple-imputation methods, on regression estimates for agricultural household surveys. The current study also analyzes the impact of multiple-imputation on regression results, when all the variables in the regression have missing observations. Finally, the current study compares the impact of univariate multiple imputation with multivariate normal multiple imputation, when some of the missing variables have discrete distribution. The results of the current study show that multivariate-normal multiple imputation performs better than univariate multiple imputation model, and overall both methods improve the efficiency of regression estimates.