Synthetic Conditional
Data Envelopment Analysis

Synthetic Conditional Data Envelopment Analysis

The objective of this project was to estimate the efficiency of education in different Brazilian municipalities in terms of the available student potential (measured by their social economic status, SES), financial resources, the results in terms of number of students served and average grades.

There are different options in terms of which variables should be considered environmental variables (the conditional part) and which should be considered input or output variables.

For instance, we could set SES as an environmental variable, financial resources as an input and grades and quantity of students as an output, or we could set the number of students as an environmental variable, SES and expenditure as input variables and grades as outputs.

The problem with the data envelopment method is that since it uses a benchmark concept if there are isolated data points where there few other comparable units then they tend to have very high efficiency evaluation because they face little competition.

Our solution was to artificially create new data points, which is why we added the “synthetic” to the name of the method. The idea is to estimate what would be the performance of the most efficient municipalities if they had different characteristics.

The most efficient municipalities tended to be poor small municipalities so the idea was to estimate what would be their results in terms of expenditure and grades if they had more students and a better student pool.

Not only the final result are more reasonable and have several technical improvements, they were more stable across different specifications which meant that we could have a more reliable way to estimate efficiency that is not as dependent of technicalities.