How can I approximate precise number of weeks worked (as from WKSWORK1) from WKSWORK2?


I’m trying to continue some analysis that used the values of WKSWORK1 to create an approximation of labor supply. However WKSWORK2 is the available variable for the years in question, and it groups these values into 6 categoricals of #weeks worked. What is the best way to approximate the actual number of weeks worked from these 6 categories?

I have considered just taking the average of each category:

1 = avg(1, 13) = 7 weeks

2 = avg(14, 26) = 20 weeks etc.

But this seems innaccurate.

Is there a better way to approximate the actual values of the WKSWORK2 categories for a given sample?

Unfortunately, there is not a universally-accepted best method to estimating the precise number of weeks. I have seen papers choose your method of simply using the midpoint of each interval. Others have restricted their analysis to full-year workers (i.e. WKSWORK2=6), such as this Brookings paper; although, that might not be feasible for your analysis.

You could also use data from prior ACS years for WKSWORK1 to guide how you impute weeks worked values.For example, in 2001-2007, you see peaks at 4-week intervals. Also, nearly half of those in the 40-47 week interval reported working exactly 40 weeks. Thus, choosing the midpoint might not be the best estimate.

Hope this helps.