Objectives This present study was aimed to examine the associations between childhood socioeconomic status (CSES) and type 2 diabetes mellitus (T2DM) among mid-late Chinese and disentangle the pathways using structural equation modelling (SEM). Methods Using cross-sectional data from China Health and Retirement Longitudinal Study (CHARLS), this study included 19767 participants aged 45 and over. SEM models were constructed to decompose the intricate relationships between CSES, childhood health history (CHH), adulthood socioeconomic status (ASES), health-related behaviors (HRB) and T2DM. Results The results showed that T2DM was significantly associated with CSES (sβ = -0.239; P = 0.001), CHH (sβ = -0.016; P = 0.005) and ASES (sβ = -0.180; P = 0.002) directly, While the indirect effect of CSES on T2DM was sβ = -0.111; P = 0.001with an acceptable goodness-of-fit. The model presented an acceptable goodness of fit: RMSA0.082, CFI 0.803, GFI 0.938, AGFI 0.904, and SRMR 0.060. Conclusions CSES had direct and indirect effects on later incidence of T2DM, which was mediated by ASES and CHH, supporting the life course theory, indicating that optimal interventions should be conducted in the early stages of life to narrow the socioeconomic and obtain maximal health benefits.