Importance Despite the evidence supporting the relationship between socioeconomic status (SES) and severe mental disorders (SMD), the directionality of the associations between income or education and mental disorders is still poorly understood. Objective To investigate the potential bidirectional causal relationships between genetic liability to the two main components of SES (i.e., income and educational attainment (EA)) on three SMD: schizophrenia (SCZ), bipolar disorder (BD) and major depressive disorder (MDD) using multivariable Mendelian randomisation. Design, Setting, and Participants We performed a bidirectional, two-sample multivariable Mendelian randomisation (MVMR) study using summary statistics from the latest genome-wide associations studies (GWAS) of the Psychiatric Genomics Consortium (PGC), the UK biobank and the Social Science Genetic Association Consortium (SSGAC) to dissect the potential associations of income and EA with schizophrenia, BD and MDD including data for participants of European ancestry only. Data covered September 2022 to January 2023. Main Outcomes and Measures Socioeconomic status phenotypes (household income from the UK biobank, n=397,751, and EA based on the largest meta-analysis GWAS of individuals of European ancestry, n=766,345) and SMD from the PGC (schizophrenia, n=127,906; BD, n=51,710; and MDD n=500,119). Follow-up analyses were performed using the intelligence meta-analytical phenotype, n=269,867. Results GWAS data were derived from participants of European ancestry included in the UK Biobank, SSGAC and PGC consortiums (samples ranging from 766,345 to 51,710 individuals). Univariable MR (UVMR) showed that genetic liability to household income was associated with decreased risk of schizophrenia (IVW-ORSCZ=0.58 per-one-SD increase; P-value= 0.016) and MDD (IVW-ORMDD:0.66 P-value: 9.81e-08), with a smaller reverse effect of schizophrenia and MDD on income (IVW-βSCZ= -0.1147, P-value=7.55E-08; IVW-βDEP=-0.0356, P-value=8.70E-04). Direct effects (MVMR) after adjusting for EA were comparable. UVMR showed that genetic liability to EA was associated with lower risk of MDD and higher risk of BD, with no significant effects on schizophrenia. After accounting for income in the MMVR analyses, direct effects of genetic liability to EA was associated with increased risk of BD (MVMR-ORBIP: 2.69 per SD increase, P-value= 0.0000113) and schizophrenia (MVMR-ORSCZ: 2.09 per-SD-increase, P-value= 0.00108), but not with MDD. Effects for the association of EA with schizophrenia and BD were larger when including genetic liability to intelligence in the MVMR model, thus suggesting that they might reflect a non-cognitive component. Conclusions and Relevance The findings of this multivariable Mendelian randomisation analysis suggest an heterogenous pattern of causal links between SES and SMD. We found evidence for a negative bidirectional association between genetic liability to household income and the risk of schizophrenia and MDD. On the contrary, we found a positive bidirectional relationship of genetic liability to EA with schizophrenia and bipolar disorder, which only becomes apparent after adjusting for income. These findings shed light on the directional mechanisms between social determinants and mental disorders and may help to guide public mental health strategies addressing inequality and economic disadvantages.