This paper presents evidence of the relationship between the disparity in the academic performance of boys and girls in Colombia and the country's excessively high school dropout rates. By using the OLS and trimming for bounds techniques, and based on data derived from the PISA 2009 database, the presented findings show that the vast majority of this gender-related performance gap is explained by selection problems in the group of low-skilled and poor male students. In particular, the high dropout rate overestimates male performance means, creating a selection bias in the regular OLS estimation. In order to overcome this issue, unobservable male students are simulated and bounding procedures used. The results of this analysis suggest that low-income men are vulnerable to dropping out of school in the country, which leads to overestimating the actual performance levels of Colombian men.