[Wikimedia-l] compromise?

[James Salsman]

Again, I am not suggesting canceling anyone's health insurance or
replacing it with increased salary. I am only trying to say that in
the case of when a parent or sibling faces catastrophic medical
expenses in the U.S., just over two years of the difference between
typical junior software engineer pay at the Wikimedia and Mozilla
foundations is the same amount that the average American who enters
bankruptcy because of medical expenses has in debt.

> On 5 January 2013 11:11, Thomas Morton <morton.thomas at googlemail.com> wrote:
>
>> So the foundation should NOT throw money at staff without showing that
>> paying extra would bring the charity significant increases in value.

If the nine reviews added to
http://www.glassdoor.com/Reviews/Wikimedia-Foundation-Reviews-E38331.htm
over the past two weeks does not establish that, then I can't imagine
anything will.

>>> A representative sample of 384 donors is sufficient to establish the
>>> answer with 95% confidence. I am not suggesting asking all however
>>> many million there have been.
>>
>> I call this number the magic 384, it's a common rookie mistake when
>> designing surveys for a million people.
>>
>> With a sample size of 384 you do get 95% confidence, with a confidence
>> interval of 5%. So the data is fairly meaningless (if 49% of your
>> respondents say X then that could represent anything from 44 to 54 percent
>> of the population).

If my preliminary informal survey of a much smaller number of donors
is representative, then the results will be much closer to 100%
agreeing that the Foundation should meet or exceed market pay than
50%.

>> You need around 12000 for any solid degree of confidence. And I believe we
>> have a lot more than a million donors across a wide variety of cultures.

That is absurdly excessive. There has never been a Foundation donor
survey of more than 3,760 donors, and that number was only chosen
because of a requirement to measure fine grained demographics in
categories for which few respondents were expected. 384 is plenty to
resolve a yes/no or below/meet/exceed question at the 95% confidence
level unless anyone has any actual evidence that the result is likely
to be close.

I am convinced that if asked, donors would think it is irresponsible
to pay so little that Oracle employees are more satisfied.