SSI Estimates

My thoughts are not likely to be terribly reassuring...

People lie (sometimes unwittingly). We spend a great deal of time talking about sampling error, but we often forget to discuss all of the other types of error that affect surveys. There is considerable evidence that self-report is subject to recall bias, mis-interpretation of the question, and (sometimes) outright fabrication. See, for example: www.npr.org/.../mischievous-responders-confound-research-on-teens

So, it may be that the respondents mis-interpreted the question and recorded some income as "SSI" that isn't really SSI. It may be that the respondent fabricated an answer (income questions are notoriously touchy). It may be that those values were imputed.

It may also be that they're collecting more income than they "should," whether because of a mistake on the part of the government or because of fraud...

As for how to deal with it in your analysis... it depends on your purpose.
You could drop those cases, top-code those cases, or treat them differently in your sample.

My thoughts are not likely to be terribly reassuring...

People lie (sometimes unwittingly). We spend a great deal of time talking about sampling error, but we often forget to discuss all of the other types of error that affect surveys. There is considerable evidence that self-report is subject to recall bias, mis-interpretation of the question, and (sometimes) outright fabrication. See, for example: www.npr.org/.../mischievous-responders-confound-research-on-teens

So, it may be that the respondents mis-interpreted the question and recorded some income as "SSI" that isn't really SSI. It may be that the respondent fabricated an answer (income questions are notoriously touchy). It may be that those values were imputed.

It may also be that they're collecting more income than they "should," whether because of a mistake on the part of the government or because of fraud...

As for how to deal with it in your analysis... it depends on your purpose.
You could drop those cases, top-code those cases, or treat them differently in your sample.

I agree with Beth's assessment about misreported and misclassified income.

I just want to add that there are a few explanations for legitimate annual payments that are higher than the $13,200 threshold.

The ACS question doesn't ask who the beneficiary is, only who collects the income. The relative of a beneficiary who resides in an institution (nursing home, hospital, etc.) may collect SSI on that relative's behalf. If the respondent is also beneficiary, he/she could report up to twice the threshold. Unless the other person is in the household, there is no way to confirm a relationship. You also sometimes see this with people managing aging parents' income/earnings who may or may not live with their kids.

Also, SSI and SSDI will sometimes make lump-sump back payments to compensate for time waiting for the disability determination was being made.

Both these situations may not explain the $30k reports, but a hard threshold may not be appropriate either.