| Field data are used to illustrate that, other things constant, regression results using Ordinary Least Squares (OLS) converge to Tobit estimates as the number of zeros in the regressand decrease. Tobit estimates are theoretically superior to OLS estimates when using censored data. However, if little difference exists between OLS and Tobit estimates, OLS may be operationally acceptable. OLS may even be optimal in a bounded rationality sense because the extra cost of using Tobit analysis may be less than the extra benefits from a very slight increase in accuracy. |
