Estimating bilateral relationships from aggregated data
Harald Badinger and Jesus Crespo-Cuaresma
presented at: Estimating Bilateral Relationships from Aggregated Data (01 Dec 2011)
Harald Badinger, WU, WIFO
with Jesus Crespo Cuaresma, WU, IIASA, WIC, WIFO
In many econometric applications, inference about bilateral relationships across economic units is desired, but only aggregated data exist for the variable of interest. In particular, often bilateral explanatory variables are available but the dependent variable is collected at an aggregated level. Such settings are common in econometric applications related to international trade (where certain variables may be observed at the country level but inference on bilateral country relationships is of interest) and regional economics (where variables of interest may be observed at a lower level of geographical disaggregation). This paper proposes a method to overcome this shortcoming by directly estimating the parameters corresponding to the disaggregated model using aggregated data for the dependent variable. Our contribution is directly related to the literature on nonlinearly aggregated models, which have been mostly applied in econometrics to obtain inference in time series models for data at higher frequencies than that at which observations are available. We provide empirical examples based on international trade and spatial econometric models.
with Jesus Crespo Cuaresma, WU, IIASA, WIC, WIFO
In many econometric applications, inference about bilateral relationships across economic units is desired, but only aggregated data exist for the variable of interest. In particular, often bilateral explanatory variables are available but the dependent variable is collected at an aggregated level. Such settings are common in econometric applications related to international trade (where certain variables may be observed at the country level but inference on bilateral country relationships is of interest) and regional economics (where variables of interest may be observed at a lower level of geographical disaggregation). This paper proposes a method to overcome this shortcoming by directly estimating the parameters corresponding to the disaggregated model using aggregated data for the dependent variable. Our contribution is directly related to the literature on nonlinearly aggregated models, which have been mostly applied in econometrics to obtain inference in time series models for data at higher frequencies than that at which observations are available. We provide empirical examples based on international trade and spatial econometric models.