[fabbari]

Did you actually look at the numbers? You should give it a try. The best you can get from those numbers is that people are more likely to donate to the Democratic party than the Republican one.

For example: true, 99.4% of the donations from the USDA employees went to the Democratic part. 99.4% of how many donations? Less than 3,000. - Source [0] - and that's an overestimate, since that filter will include donations from 2015.

How many employees work for the USDA? About 100,000 [1].

That means that 97,000 people made no donations - to either party.

[0] https://www.fec.gov/data/receipts/individual-contributions/?...

[1]

[keeda]

Also, from the very same govexec link:

> The lopsided donations do not necessarily reflect how the federal workforce is voting. The former State Department secretary led the businessman by 5 percentage points among federal employees in a July poll by the Government Business Council, the research arm of Government Executive Media Group, with 42 percent of respondents saying they would vote for Clinton, compared to 37 percent who said the same for Trump.

[rayiner]

Why wouldn’t donations be a reasonable proxy? 3,000 samples is more than enough to have a statistically significant analysis of a population of 100,000.

[bigbadfeline]

> 3,000 samples is more than enough to have a statistically significant analysis of a population of 100,000.

It's not a valid sample because it's not a random selection - statistical sampling isn't only about sample size.

One reasonable conclusion that can be drawn from the above numbers is that the Democrats in government are less stingy about donations, or that the Republicans there don't really like the GOP and are there for the money alone.

[fabbari]

It is not a population sample - IE: we didn’t randomly pick 3000 people and observed who they donated to. Or in other words: we asked the whole population who they wanted to donate to and 97,000 said “none”.