A method has been proposed to form the optimal structure of regional state investments, which contributes to the strategic goals and objectives of the socio-economic development of a region through the advanced development of human capital. The dynamic model has been considered representing a mathematical programming problem, which describes in the form of recurrent dependences the chain of channels of influence: "the structure and volume of investments→the indicators of the regional human capital→the indicators for the socio-economic development of a region". The weighted average of degrees in achieving the target values of resulting indicators of socio-economic development of a region on the considered horizon of planning has been used as the objective function. Recurrent dependences are the lag econometric models of panel data involving the main components. To construct three types of models (the through models, those with the deterministic and random spatial effects) using the Best Subset method, the open-source software R was employed. The best models were chosen with the help of tests by Wald, Hausman, and Breusch–Pagan. The limitations within the model are a series of assumptions about the processes of the development of human capital and the socio-economic development considering the uncertainties. The optimization variables are shares of the distribution of investment resources based on the investment directions and years. Based on the results of modeling and numerical calculations on the example of several regions of Russia in dynamics over the years, the optimal investment structure has been proposed. A given structure enables making the maximum progress towards achieving the target values of strategic indicators of the development of a region through the development of human capital