JOHN G from MI posted over 5 years ago:

This seemed interesting so I went and read the 29 page paper this was drawn from. While the article above correctly summarized the article I believe the study had flaws. (If someone could correct me if I'm wrong I would appreciate.) The article had several hypothesis, they all were only focused on income. This is like a plant study where some got no fertilizer, some fertilizer A, etc. Then you could determine what was likely outcomes. What is missing is that humans are very difficult to analyze. I couldn't determine if some were working, not for the pay so much as what they got out of it in non-monetary ways. A doctor often doesn't retire but keeps working because they get tremendous satisfaction from helping. Same with other professions. What about teachers, some keep working because of the growth they see in their students. Some people may take a job as a clerk in a store just to be around people even though it may be minimum wage. I didn't see where any of this was factored in. these are my thoughts and questions

THOMAS M from IL posted over 5 years ago:

You make some great points, John. One of the unstated assumptions - at least, I didn't see it stated - is that the researchers attempted to account for different levels of non-wage incentive to work by segmenting the sample, for example, by job level. Although the authors assert a number of times that their effect size is large, in spite of the relatively large number of variables they included in their analysis, there remains a huge chunk of variability in people's return-to-work behavior that compensation did not predict - presumably including the variability due to factors that you mentioned. The way I think about your hypothesis is that the doctor's helping behavior or the teacher's student development or the clerk's social contact are actually leisure behaviors that happen to be statistically confounded with the variable of interest. The researchers interpret the behavior as work but some of the participants might describe it as leisure. I believe that because the highlighted results are part of a multivariate analysis, at best they predict "on the margin" and shouldn't be interpreted in isolation from the many other variables in their model. The criticism I'll offer is that I didn't see a cross-validation of their model - although I may have missed it as I only skimmed their research article. In a cross-validation, they would have taken their prediction equation and applied it to a small "hold-out" sample to see if it correctly predicted the elasticity they believed they found in their research sample. One of my former colleagues noted frequently that "people are complex and the world is messy." So my compliments to the researchers for attempting to predict real-world behavior - and to you for digging in deeper.