In this work we analyze methodologies that are used in research to obtain estimates of policies efficiency and will show how Propensity Score Methods can be adopted to the given scarce data on the case of Russian loan guarantee program. We take number of workers employed in Russian SMEs as a target metric; an aim to increase this number to 25 ml workers which is claimed in Russia’s national development goals. The study suggests methods for matching PSM propensity scores that can be applied to obtain an assessment of the effectiveness of a specific government measure. These methods, along with two-stage Heckman regression and panel data models for multi-period Difference-in-Difference expansion, seem to be the most suitable for evaluating the effect of intervention and reducing sampling error. The estimates obtained on the basis of this methodology indicate a significant positive effect of issuing guarantees in terms of increasing employment. Also we highlighted industries with the most positive average treatment effect from the program. Those policy measures the efficiency of which was proved by quantitative research need to be promoted more actively than less efficient measures: authorities need to reduce obstacles that beneficiaries can face to receive government support, to increase financing, to improve SMEs’ awareness of such policy measures.