search

PSLmodels / ui_calculator

Calculator for unemployment insurance benefits
22 stars 12 forks source link

Adding DC? #3

Open raheem03 opened 4 years ago

raheem03 commented 4 years ago

Is there a reason DC is not included in the state_eligibility worksheet?

If I'm interpreting this this correctly, it looks like hqw = 1.5*, absolute_hqw = 1300, and abs_2hqw = 1950.

FWIW, I implemented this on my local copy and it gives reasonable looking results. Is there a reason we shouldn't do this?

(*This isn't quite right since it can be within $70 of 1.5 * hqw, but I think implementing that would require writing an exception or an additional parameter into the function.)

ganong123 commented 4 years ago

@raheem03

thanks for flagging this! no, I don't there was a good reason to exclude. in this issue, can you report average UI benefit and average weekly wage using the model applied to the CPS alongside the BAM benchmark? Assuming these are close, please make a PR adding DC to state_eligibility.csv. Thanks!

raheem03 commented 4 years ago

@ganong123

Re-ran using longer_example.R file (but did not filter out DC/WV) and found:

BAM CPS
aww 1111.17 1452.77
wba 360.63 393.17
rr .398 .353

Weekly earnings in CPS are about 30% higher. Seems high (only NJ and MD come close with 23% and 24% increases respectively), but given small sample size in DC high variability seems plausible. Perhaps as a result, benefits are about 10% higher.

ganong123 commented 4 years ago

@raheem03 thanks for flagging this issue and giving this a shot. one goal we have is to incorporate multiple years of data in #5. doing that will help sort out whether the problem is (a) sampling variability or (b) that the benefit model is misspecified in some way. let's hold off on adding DC until #5 is done.

raheem03 commented 4 years ago

Sounds good. Just to clarify, I felt the average weekly wages seemed a bit high (which is independent of the calculator); the benefits seem to be in a reasonable range. FWIW, I slightly adapted the procedure you all provided in the example code and ran the average weekly earnings for the March CPS in 2016-2019. DC wages do tend to be quite a bit higher than even other high-income states. They also don't seem to be moving around much. But, as you say, it might be good to validate against other years more rigorously.

Anyway, thanks for following up and looking into this!