Taro Mieno - https://agecon.unl.edu/faculty/taro-mieno
Measurement errors are often ignored as a source of estimation bias in many of the fields of economics despite its ubiquity. One such field is climate econometric analysis. Commonly used weather datasets (e.g., PRISM, Daymet) in climate change analysis have measurement errors because they are products of spatial interpolation/modeling of weather point data observed at individual weather stations. The degree of measurement errors in these products and also its econometric consequences are not well understood.
Another under-appreciated econometric problem often encountered in climate change analysis is aggregation bias. In this presentation, first, I will present some basic theory of the impacts of measurement errors and aggregation bias when the variable of interest has a non-linear relationship with the dependent variable. One of the findings is that measurement errors can bias the estimation of where the threshold (where the marginal impact of a variable changes the sign) is.
I will also show that aggregation can counter measurement error problems under a certain condition.
Finally, I will present simple numerical examples of the potential degree of estimation bias associated with measurement errors in precipitation and aggregation after characterizing measurement errors observed in PRISM precipitation data.