Doing harm when you think you’re doing good – part 2
Evil?
If you become a fundraisers, you won’t raise extra money for charity. Rather, you’ll shift money from one charity to another.
Suppose Oxfam spends an extra £1 on fundraising. We’d expect this to raise them at least £4^{1}. So, it looks like they’ve generated an extra £3 for charity. Great!
But what would have happened if they hadn’t spent that £1? If this was an easy opportunity to raise money, then another charity would take it instead. Charities as a whole are raising about as much money as they can from the public. This is what we’d expect. Charities are aiming to do as much for their cause as possible, so if there were easy money to raise, they’d take it. And intuitively, people have a relatively stable ‘charity budget’ . They avoid future giving once they’ve ‘done their part.’
This means that Oxfam wasn’t actually generating new money for charity by doing more fundraising, rather, it was shifting it from somewhere else. So the extra fundraiser wasn’t doing any good^{2}.
But it gets worse.
If you graph how costeffective charities are, you’d expect to find a curve a bit like the one on my very basic graphic below^{3}:
The key point is that the median is significantly less than the mean. In other words, the effectiveness of the majority of charities is less than the effectiveness of the average charity. That sounds a bit odd, but it’s just because the average effectiveness is pulled up a long way by the small number of really good charities at the top. (Similarly, the average wage at a company is normally higher than what the majority of the staff are paid).
This means that most fundraisers at the margin are shifting money to charities that are less effective than the average charity. So, they are reducing the overall effectiveness of the charity sector. They are actually reducing how much good gets done!
It’s pretty likely that the effectiveness of the median charity^{4} is significantly less than the mean. it could easily be as low as one fifth^{5}. if it were one fifth, then even if every £1 of fundraising only caused 80p to move from the mean to the median, it would negate the good done by the £4 spent by the median. see below for the calculation^{6}. So, even if most of the money raised wouldn’t have been given by the public otherwise, no good is being done.
In fact, we’ve just seen that extra charity fundraising doesn’t cause much extra money to be given to charity. Most or perhaps almost all of the money is just shifted from other charities. Assuming the money comes from the mean charity, then the majority of fundraisers are shifting money from more effective causes to less, significantly reducing how much good is done by charity overall. The majority of fundraisers look like they’re doing good, but taking into account the indirect effects could mean they’re actually making the world worse off^{7}.
Before we get carried away, it’s important to note that that fundraising as a whole is still doing good^{8}. Moreover, if you’re a fundraiser, and you’ve got no idea how effective your charity is, then you should expect it to have the mean effectiveness (not median). That means that if you become a fundraiser you should expect to be doing good (or at worst be neutral). If you’re working for a charity that’s more effective than the mean, then you’re doing a lot of good.
If, on the other hand, you’ve got reason to think that your charity is less effective than the mean, then please stop!
The reverse, however, is true for the minority of fundraisers who work for charities that are more costeffective than the mean. They’re shifting money from less effective charities to more effective ones, potentially doing a huge amount of good.
If you’d like to find out more about effective giving, check out our page.
Thanks to Will MacAskill for the inspiration
You might also enjoy:
References and notes

According to the Institute of Fundraising’s Fundratios survey:
http://www.instituteoffundraising.org.uk/careerguide/fundratios/ ↩ 
More accurately, additional fundraising seems to cause additional giving of about £1.9, rather than £4, so the extra ‘profit’ to the charity sector is only 90p per £1. It’s not clear that this can go much lower, due to the risks of fundraising. So, fundraisers probably cause some extra money to be donated to charity, but it’s much less than it looks.
/2012/10/whydontcharitiesspendmoreonfundraising/
The nation’s charity budget is equal to about 1.1% of GDP. http://www.philanthropyuk.org/resources/usphilanthropy
Of course, there may still be ways to significantly increase how much is given to charity in total. It’s just that it would require a new approach. In fact, we’re very hopeful that 80,000 Hours and GWWC are one of these new approaches. ↩ 
We’d expect the distribution of interventions theoretically to be lognormal because most characteristics that make for an effective intervention will be distributed normally. The total effectiveness will arise from the product of all these characteristics. Multiplying normal distributions gives you a lognormal distribution.
This has direct empirical support. The DCP2 found that health interventions were lognormally distributed. The most effective intervention generated 300 DALY per $1000, while the least generated only 0.02. The median intervention generated about 5. The mean generated about 25.
There is a difference, however, between the distribution of interventions and the distribution of charity effectiveness. If we assume that charities choose interventions randomly (which might not be so bad an assumption), then the distributions would be the same. However, it seems like charities care at least a little about effectiveness. This will mean that the distribution of charities will move further to the right. We’d expect, however, for it to still have that kind of shape. ↩ 
Or more accurately, the average effectiveness of all charities that are less effective than the mean. By taking using the, we’re making an underestimate of the gap. ↩

For the health interventions in the DCP, the median effectiveness was about 5, while the mean was 25. We’d expect the mean in practice to be slightly less due to reversion to the mean. On the other hand, if we’re considering all types of intervention (rather than just health), the overall dispersion will be much larger, making the gap between the mean and the median significantly higher again. So, assuming the median effectiveness is about one fifth of mean effectiveness is likely to be an overestimate. ↩

 Let the median charity have an effectiveness of 1 unit/£, where ‘1 unit’ is just whatever is of value in the area being considered, for instance ‘healthy people.’
 The mean charity has an effectiveness of 5 unit/£
 The median spends £1 on fundraising, and raises £4
 £3.20 is money that wouldn’t have been given otherwise, 80p is moved from the mean charity to the median
 In reality, the median charity spends the (41) = £3 of ‘profit’, creating 3×1 = 3 units of value
 But if they hadn’t fundraised, then the mean charity would have had an extra 80p. It would have cost them 20p to raise this money, so they would have had an extra 60p to spend. This would have produced 0.6×5=3 units of value
 So, no good has been done overall

More realistic figures:
* The median spends £1 on fundraising, and raises £4
* £0.50 is money that wouldn’t have been given otherwise, £3.50 is moved from the mean charity to the median
* The £3 of ‘profit’ creates 3×1 = 3 units of value at the median charity
* But if they hadn’t fundraised, then the mean charity would have had an extra £3.50. Subtracting the fundraising cost of 90p, means it would have produced 2.6×5 = 13 units of value.
* So, the median charity is overall causing the loss of 13 – 3 = 10 units of value. Impressively, this loss is over three times as large as the good they apparently do directly. ↩ 
Assuming the mean pound that goes to charity does more good than a pound spent on consumption ↩