We develop a methodology that recovers an estimate of the average stock of vacancies using the information on aggregated hires. We show that our prediction of the vacancy stock is unbiased, and it captures well the level and the dynamics of the United States job opening positions reported in the Job Openings and Labor Turnover Survey. We use the methodology to predict vacancies in Colombia for formal and informal salaried workers; together with unemployment, we estimate Beveridge curves and matching functions by occupations, which allows us to study the nature of the efficiency, frictions, and mismatches for different occupations. We find that the formal labor market of technicians is the most inefficient of them all; this inefficiency comes from the mismatch between the abilities of the workers and the requirement of the vacancies. Reducing friction in this occupation will require education and job-oriented training policies. In contrast, the frictions in the market for unskilled workers come from informational lacks. The reductions of friction, in this case, will come from better intermediation and active search policies.