Subsidies for investing in energy efficiency measures: Applying a random forest model for unbalanced samples.

Investing in energy efficiency measures is a major challenge for SMEs, both for environmental and economic reasons. However, certain barriers often make it difficult to invest in such measures. Although public financial support helps to overcome economic barriers, public bodies face the challenge of identifying which SMEs display the greatest potential to invest in energy efficiency measures. By applying a random forest technique and by using sampling balancing techniques, this paper identifies the profile of industrial SMEs that might be potential beneficiaries of public aid, thereby helping public institutions to target their calls and direct their efforts towards this group of SMEs. Specifically, liquidity and indebtedness are found to be the most useful predictors for SMEs in the industrial sector. The results are robust and reveal that applying a random forest approach for unbalanced samples offers greater predictive capacity and statistical power than applying traditional estimation techniques. By identifying potentially benefiting firms, this work helps to boost the effectiveness of public subsidies and to improve the channeling of public funds, which ultimately favors investment in energy efficiency. • Public subsidies favor energy efficiency measures by reducing the up-front cost and making the investment more profitable. • Subsidy effectiveness depends on the capacity to identify which SMEs are potential beneficiaries of energy subsidies. • We evidence that applying a random forest approach for unbalanced samples offers greater predictive capacity and statistical power than traditional techniques. • The most useful predictors for SMEs in the industrial sector are related to liquidity and indebtedness. [ABSTRACT FROM AUTHOR]