Establishing balance for important covariates is often critical in causal inference and clinical trials. Covariate-adaptive randomization (CAR) and stratified permuted block (STR-PB) are commonly implemented to achieve this goal. While the balance properties of these methods have been extensively studied for observed covariates, their properties for balancing unobserved covariates have less been understood from the theoretical aspect, and have been subjected to criticism in the literature. In this presentation, we will introduce a framework for assessing the theoretical properties of unobserved covariate imbalance. This versatile framework allows for the analysis and the comparison for the balance properties of complete randomization (CR), STR-PB, and other CAR procedures in relation to the unobserved covariates. Our findings highlight the advantages of utilizing CAR or STR-PB (especially when the number of strata is relatively small compared to the sample size) for balancing unobserved covariates. Numerical studies are also presented to demonstrate finite sample properties of our theoretical results. These findings not only serve as a foundation for the future research on the impact of unobserved covariates for covariate-adaptive randomized trials but also open up possibilities for various other applications.