Binary logistic regression. missing values


I have a question about missing values. My dependent value is whether the respondend voted in the last election or not - N=2600 cases. I want to build a model with several independent variables (income, education, parrty identification etc.) but I have a problem. If these independent variables are included, the number of cases included in the analysis sinks to only abou N=1300. I run a missing values analysis and discovered that most missing values are in the variables level of education and income, but these are essential to this model and I can't ommit them so easily. Homser-Lemeshow is ok, R2 is high enough, the model is not overfitted, bud I donť know, wheter it is legit to build a model, where about 50% of cases are missing due tu item non-response.

Can someone help, please?