Abstract: We developed an evaluation system based on the Delphi method to assess the vulnerability of herder households under the framework of sustainable livelihoods in the grassland regions in northern China. The herder households’ livelihood capital conditions were measured in the meadow steppe, typical steppe, desert steppe, sandy steppe, and desert and mountain steppe regions of northern China. The characteristics of differentiation were also explored at household and regional scales. In the two rounds of expert opinion of the herders, the coefficients of coordination were 0.445 and 0.692 (P<0.01). The livelihood capital of the herders in northern China was lower, indicating that these herders were vulnerable (mean vulnerability index, 0.2990). The herders were clustered into two groups by the K-means clustering method. The data were analyzed by an F-test and the F-value was significant, confirming that the herders could be divided into two groups on the basis of their vulnerability index. The herders’ human capital, natural capital, physical capital, financial capital, and social capital stock were significantly lower in the more vulnerable group than in the less vulnerable group. On the whole, the householders’ vulnerability tended to decrease from west to east at the county or regional scales. The regions were ranked, from most vulnerable herder households to least vulnerable, as follows: mountain steppe>steppe desert>desert steppe>sandy steppe>meadow steppe>typical steppe.