With the per capita GDP of counties in Guizhou Province from 2006 to 2017 as the research object, an analysis has been made of the dynamic evolution characteristics of spatial differences in economic at the counties in Guizhou Province by using ArcGIS and GeoDa software, with such spatial statistical methods as spatial auto-correlation analysis, spatio-temporal transition and other methods adopted. The results show that: during 2006—2017, the Global Moran , s I Index rose from 0.149 to 0.311, indicating that the county economy in Guizhou Province is characterized shows a significant spatial dependence, while the spatial dependence is increasing, the inter-county economic spatial differences keeps narrowing; the county economy of Guizhou Province is characterized with a spatio-temporal patternof the coexistence of“agglomeration”and“differentiation”. From 2006 to 2017, the number of H-H agglomeration counties decreased from 18 to 14, and the number of L-L agglomeration counties increased from 34 to 41, indicating that the degree of spatial agglomeration of counties with higher levels of economic development has weakened, and that of counties with lower levels of economic development agglomeration has increased. From 2006 to 2017, 55 counties experienced type IV transitions, accounting for 69.62% of all counties, indicating that the local spatial association structure and agglomeration characteristics of county economy in Guizhou Province have strong spatial stability, and the spatial spillover effect of county economy has less influence on the adjacent counties, i.e., the spatial pattern of county economy in Guizhou Province has certain spatial locking or path-dependent characteristics. From 2006 to 2017, the number of High-High agglomeration counties increased from 3 to 7, with the number of Low-Low agglomeration counties decreasing from 8 to 6. However, most counties in Guizhou Province failed to pass the significance test, indicating that most counties in Guizhou Province is basically in a state of isolated development with a degree of rather weak spatial correlation.