Hi all,

I am trying to estimate the effects of a conditional cash transfer program in some labor market outcomes like employment and wages using an RD design. The data I'm using has approx. 11 million observations. My running variable is the per capita income by family, and it only has discrete values (integer values).

I have generated some results using different bandwidths values, but I am having problems in estimating the RD with the optimal bandwidth procedure created by Calonico, Cattaneo, and Titiunik (rdrobust). The command I am using for this is:

rdrobust `var' ${running_var}, fuzzy(${treatment}) kernel(triangular) p(1) c(70)

And I got the following message “Invertibility problem in the computation of preliminary bandwidth below the threshold Invertibility problem in the computation of preliminary bandwidth above the threshold Invertibility problem in the computation of bias bandwidth (b) below the threshold Invertibility problem in the computation of bias bandwidth (b) above the threshold Invertibility problem in the computation of loc. poly. bandwidth (h) below the threshold Invertibility problem in the computation of loc. poly. bandwidth (h) above the threshold”.

I am not pretty sure why I am getting this mistake. Could it be explained by the discrete nature of the running variable or the number of observations?

I would be very thankful for any help or advice.

Best regards,
Sebastian