Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/196650 
Year of Publication: 
2019
Series/Report no.: 
IZA Discussion Papers No. 12152
Publisher: 
Institute of Labor Economics (IZA), Bonn
Abstract: 
Econometric analyses in the happiness literature typically use subjective well-being (SWB) data to compare the mean of observed or latent happiness across samples. Recent critiques show that com-paring the mean of ordinal data is only valid under strong assumptions that are usually rejected by SWB data. This leads to an open question whether much of the empirical studies in the economics of happiness literature have been futile. In order to salvage some of the prior results and avoid future issues, we suggest regression analysis of SWB (and other ordinal data) should focus on the median ra-ther than the mean. Median comparisons using parametric models such as the ordered probit and logit can be readily carried out using familiar statistical softwares like STATA. We also show a previously as-sumed impractical task of estimating a semiparametric median ordered-response model is also possi-ble by using a novel constrained mixed integer optimization technique. We use GSS data to show the famous Easterlin Paradox from the happiness literature holds for the US independent of any paramet-ric assumption.
Subjects: 
ordered-response model
mixed-integer optimization
median regression
subjective well-being
JEL: 
C24
C61
I31
Document Type: 
Working Paper

Files in This Item:
File
Size
539.26 kB





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.