Some context
So for everyone too lazy to read the report (which you should because it is extremely interesting! this post leaves out a lot of details!), here is the TL;DR:
(method stuff is taken from page 11)
Amazon's response is (as expected) out-of-context bs. Yes, they reached the workers through social media campaigns, but targeted it at people who had listed amazon as employer on facebook, and they used geofencing for location too. The method has been established in prior research as being a viable way to reach a wide sample of a specific company. This is not just sending around a survey link on facebook...
In this survey, they used a few ways to weed out false responses:
1. captchas
2. asking for an alphanumeric warehouse code, and including a false value in the question prompt.
3. using advanced data analytics, to identify patterns of fraud (this is a bit vague but it's a tl;dr, think for example geographic location not matching a warehouse, or a cluster of similar responses that could indicate fraud).
4. They weighted the survey results by weighing the diversity characteristics (gender, ethnicity, etc.) of their survey with the published metrics amazon provided.
The survey was approved by the UIC Institutional review board.
And it is true this research was funded by an organisation with a specific agenda, but research can never be discarded because of that simple fact. They are essentially saying this is a fraudulent, non-scientific report. That's a very big claim to make towards university researchers, and they really don't have sufficient evidence to back that up.
Now, there are some caveats:
1. The percentage is inflated, because 90% of filtered respondents were "Warehouse Associate (Tier 1)", which i believe is lowest in the pecking order. They are going to have a higher injury rate than manglement.
2. You're more likely to respond to a survey like this if you have been injured.
3. Respondents who have been injured are probably more likely to make their employer seem a bit worse, especially after unpaid time off
4. This is a measure the worker's own measure of injury, i.e. not returning home in the same state they got to work at. This will not match OSHA numbers either. (this is mentioned in the report)
The report lists a lot of caveats too, most are mentioned in this article. (p.17-18)
I think it perfectly shows how amazon works as a company. Most other companies would at least respond with "We are very alarmed, and we will review this report and conduct our own research", which they will of course never do and hope people forget about it. No, Amazon immediately discredits the research and continues to push it's "OSHA and Unions are evil" propaganda. Disgusting.
Yes, this report inflates some numbers due to how participants are selected, and these injury rates are not apples to apples, but they acknowledge this personally, and they simply ignore disturbing facts like people being scared of reporting their injuries, and their Amcare facilities discouraging people from seeking medical attention (to be fair, people seem to be generally happy about the care they receive there).
They use some small issues with the research, blow them out of proportion, and use it as a strawman to disregard this research.
Biting the hand that feeds IT
