I am trying to replicate the OFFPOV column in Table 1 titled “Difference between Official Poverty Rates and PUMS Poverty Rates” on the web page Official Poverty Rates Using the Cps Public-Use Microdata Series. I need to do this with the underlying income, threshold and weights. My current code is typically off by 0.4% that is 13% can come out as 13.4% or 12.6%.

Here is what I am doing. Year-by-year I fliter out data where OFFPOVUNIV is not equal to 1 (i.e. require OFFPOVUNIV==1). For income I am using OFFTOTVAL (I have also tried using FTOTVAL with similar sized errors). For the cutoff I am using OFFCUTOFF. For the weights I am using WTSUPP (I have also tried using HWTSUPP, again with similar sized errors).

Year-by-year for all observations that passed the OFFPOVUNIV==1 test, I sum the WTSUPP values. Call this Total_WTSUPP. Next I create a dummy equal to 1 if OFFTOTVAL<OFFCUTOFF and 0 otherwise. I then multiply this dummy by WTSUPP. Call this Pov_WTSUPP. I then calculate the poverty rate as Pov_WTSUPP/Total_WTSUPP.

As an example of the error, in 1988 the official rate is 13% and by my calculation it is 13.37%. Am I using the wrong income or threshold variable? Am I using the weights as intended? Thank you to whoever can help me resolve this problem.

There are a couple of different ways you could do this using the online analysis tool. The first is using the collapse syntax, where you could specify something like the following: “ftotinc(c: 10000, 1)”. The second is using the basic recode syntax, where you would manually recode values as: “ftotinc(r: (0-10000, 10001-20000, etc.)”.

Of course you could also easily perform this task by downloading IPUMS USA microdata and using Excel or another statistical software (such as STATA, SAS, or SPSS). In fact, using the microdata will allow you to analyze household income by family size without creating arbitrary income brackets.

It actually looks like you are running the calculation correctly, it just seems that the table layout is a bit tricky. It should be noted that the CPS ASEC refers to the previous calendar year. Therefore, the “survey year” provides statistical estimates for the previous year. In Table 1 on the page you’ve linked above, the “survey year” is in the second column and the “poverty year” (e.g. the actual year of reference) is in the first column. So, for example in 1988 when you calculate 13.37 you’ll notice that this corresponds to “poverty year” 1987 and “survey year” 1988. Which, in Table 1, has an OFFPOV value of 13.4.

I hope this makes sense. It is just an easy fix of interpretation of the table.

Thank you. That solved the problem.