As a phenomenon in the process of urbanization, the relationship between urban shrinkage and carbon emissions is complicated. There are few studies to explore the driving factors of carbon emissions from the perspective of multi-dimensional urban shrinkage. Thus, we constructed an urban shrinkage evaluation system from three dimensions of population, society and economy, and used the panel data of 50 cities in four provinces of the middle reaches of the Yellow River from 2012 to 2020 in this paper. The impact of urban shrinkage on spatio-temporal differences of carbon emissions is explored by the entropy method, night lighting index, and geographic detector. The results verify that,first the shrinkage phenomenon of resource-based cities along the middle reaches of the Yellow River is the most serious one, which shows the characteristics of "large dispersion and small gathering". The proportion of population shrinkage is the largest one, and the comprehensive shrinkage is the least sector. Second, the total carbon emissions increased from 2327 Mt to 3040 Mt during the study period, and the hot spots are distributed in Ordos, Wuhai, Hohhot, Xinzhou, and Shuozhou. The cold spots gradually changed from scattered distribution in Shaanxi and Henan to agglomerate in Henan Province. The total carbon emissions are polarized and the number of high-value areas is small, but the value is large, and the increase rate is larger and larger; the number of low-value areas is large, but the value is low, and the trend is decreasing year by year. Growing and mature resource cities are the hot pots of carbon emissions, and urban shrinkage aggravates carbon emissions; Declining and renewable resource cities become the cold spots of carbon emissions, inhibitory effect is strengthened. Third, the impact of urban shrinkage on carbon emissions is complex, and the influencing factors of economic contraction have the strongest explanatory power on the spatial differentiation of carbon emissions. Conversely, population shrinkage has the least explanatory force, and the spatial superposition of most influencing factors will produce a two-factor enhancement. And the dominant factors have changed from population and economic contraction to social and economic contraction. Overall, there are differences in the spatial and temporal patterns of urban carbon emissions and their influencing factors along the middle reaches of the Yellow River. Therefore, these differences should be considered so that the energy-saving and emission-reduction countermeasures can be put forward and adapted to the needs of regional development.