The flat margin effect

Here’s a claim to supplement the replacement effect:

The flat margin effect (FME): If you take a job that seems to have a strong (positive or negative) impact on the economy, the actual difference it makes to social welfare will be minimal.

So who is this relevant to? And why should you believe it?

On the first question, it’s relevant to any effective altruist thinking of taking a job in something like finance, which they’re concerned will contribute to an industry that harms the wider economy. Perhaps you don’t agree with the replacement effect, or perhaps you think it doesn’t apply in your case for some reason (maybe you’re going to be such an effective banker that you’ll actually make a big difference to the industry’s success).

It might also apply to people considering startups whose product would contribute to the economy (for example business-to-business services), who are wondering how much the positive value from doing so weighs against a higher income than they might get from following another path.

Unlike the replacement effect, it doesn’t apply – or at least not as much – to jobs where you directly improve people’s welfare (such as becoming a doctor). It will still be relevant to any economic effects these jobs might have beyond the direct good (or bad) they do.

As for why you should believe it, let me explain with an example. The New Economics Foundation estimates that bankers remove £7 from the UK economy for every pound they earn.(1) But using money as a metric only works if you have a sense of where the money would have otherwise been spent.

To answer that question, we need to consider where this economic damage is inflicted. Two likely answers to this are on the public and on the government – on current (UK) tax rates, at about a 63:37 ratio on average. First, let’s consider the impact that a loss of income can have on individuals.

A case study of individual spending

Suppose Karl has a weekly income (after tax) of £500, which he typically uses something like this:

• £100 on rent
• £30 on bills
• £30 on transportation
• £90 on food for home preparation
• £40 on eating out with friends
• £100 into long-term savings
• £50 into saving for a forthcoming skiing holiday
• £20 on clothes
• £80 on books, music and other minor luxuries

Now I’ll disrupt this rosy picture by introducing a destructive banker. Via the magic of thought-experiment vagary, he damages the local economy to the extent that Karl’s company, while still solvent, has to cut his income by 10%. Accordingly, he has to subtract £50 from his weekly expenditures

Assuming Karl’s a faintly rational sort, he’s going to find the expenditure that brings him the least satisfaction – presumably his luxuries – and subtract it from there. So while losing the money might inconvenience him, rather than doing so to the tune of the average £50 he spends, it’s going to inconvenience him to the equivalent of the worst £50 he would otherwise have spent.

Marginal government spending

So how would a comparable loss of income affect the government? Much the same, for exactly the same reason. If we assume the government is fairly rational and benevolent (choke back your mirth; I’ll revisit this hypothetical in the objections below), any money that comes out of its budget is going to come from its least effective spending.

My assumptions so far are are quite damning toward banking – I’m supposing bankers are taking large sums of money from an organisation that would otherwise have spent them wisely. So just how bad would it be?

Let’s have a quick look at real numbers. We know the NHS spend up to £30,000 to buy a QALY – because they tell us.

So even if we also assume (with no particular reason to) that the NHS is the worst of the UK government’s expenditures, our antisocial banker only needs to donate 4p per £10 he earns to make up the difference,(2) so long as he does so to the best causes – and that’s ignoring the replaceability effect. Everything after that is of net benefit to the world.

Hang on a second…

This is a controversial claim, so I’m going to anticipate a few objections – and concede some ground to one of them:

1) Why not just get a job that doesn’t do any harm?

If you can find a job that pays the same as a finance role and does less harm (or more good, depending on your perspective), then great. But most jobs pay far less, and as an effective altruist, the value you can get from the extra money would be enormous.

Even a moderate-sounding difference in the income from two jobs can make a big difference to your positive impact. As your income rises, it becomes easier to give away a higher percentage of your salary, effectively volunteering yourself for progressive taxation.

A banker on £500K a year can easily give away 80% of his income, effectively gaining £394,000’s worth of £50-QALYs on the above estimation (and still not accounting for the replaceability effect, which would make it almost a round £400K’s worth of QALYs). A nearly-as-successful businessman whose job’s social impact was neutral would clearly struggle to do anywhere near as well on an annual salary of, say, £400K.

2) Are you kidding? The government isn’t rational or benevolent!

Everyone will have a different view on this objection, and I’m not going to argue the point. But if anything, it makes my case for me: the less benevolent or rational the government tends to be, the less problematic it is to take money from them in the first place. Under a truly irrational rule, had your job not destroyed that £7, it might have been used to plant teddy bears in the Thames.

The same idea holds for individuals – the worse Karl is at using his money to satisfy his life-goals, the less he’s harmed by losing some of it.

3) This is all very well in theory, but I know that my government is cutting spending in areas that aren’t the least efficient.

I’m very cautious of such claims, on the grounds that I think politics is the mind-killer. If that argument is right, all of us should have much lower confidence in our political views than we usually do.

That said, assuming you do have reliable information to this effect, a couple of factors might mitigate it.

Firstly, how much worse are the cuts than they might be? We needn’t assume every cut will be perfect (indeed, that would imply a perfectly rational government who would spend the money even better than effective altruists).

Assuming a human government, it would be incredible if the cuts were to come from precisely the most efficient areas. It would be like trying to peel an orange without removing a single bit of the pulp – requiring so much effort it would ultimately be counterproductive. But as long as we can expect the cuts to tend toward the edges, the overall effect is much the same.

Secondly, governments don’t allocate all spending centrally. They tend to cut from departments, local councils etc, who then have to reallocate the remaining budget internally. And departments, like governments, have cause to behave broadly rationally. For example, if the NHS were to lose a sizable proportion of its budget, we would expect it to come from the treatments that cost £30,000 per QALY, rather than those which cost only a few hundred pounds .

Thirdly, if your information is about a specific policy the government is planning to implement, odds are the decision will have been made long before you can make a career change.

So for the FME not to apply (given that you’re contemplating a career where it might), you need to be able to predict a long-term government tendency to make particularly inefficient cuts iff the annual GDP warrants it, and otherwise to leave the status quo untouched. If you can predictably their irrationality but not the form it will take then the comments in 2) apply.

4) It’s wrong to harm people, regardless of whether you’re doing it to benefit others.

A full response to this would be far too large for this blog post, since it would need to address the underlying assumptions about morality. In lieu of a full argument against it, I would argue that most people don’t really behave as though this is true:

If you think it’s wrong ever to harm (or collude in harming) people for others’ benefit, do you agree that war is ever justified?

If so, obviously in some cases it’s permissable to harm others.

If not, do you minimise your income in order to minimise the funding your national military receives from your tax?

If so, then I can only admire your dedication and respectfully disagree on this issue.

If not, then you are already colluding in harm for prudential reasons, but for far less humanitarian benefit.

5) You seem to be assuming any damage to the economy will be evenly spread. What if it’s the poorest people whose livelihood suffers the most?

I actually think this is a valid objection – if this happens, it could substantially weaken the FME. I’ll say two things against it, though:

Firstly, it’s unlikely that that high a proportion of the wealth will come from the poorest people’s pockets, simply because their pockets are relatively small to begin with. Trivially, someone with no income can’t suffer from a proportionate income reduction. This capping effect probably isn’t that large, though – the relative poor in the UK have enough wealth between them to account for fairly large economic losses. However, it does scale well enough for us to draw an a priori probability inference: any loss to the economy is most likely to come from the area of median salary (approx £22,800) since that’s where the biggest pool of money is (PDF).

I’ll go into more depth in a subsequent post, since it seems too much of a digression here, but for now I don’t think we should assume the poor are hit disproportionately to the middle classes – in fact we should think it quite unlikely, unless we have strong evidence to the contrary. (In the next post I’ll argue that even if they are, the negative effect is less than we might think, but again that’s too much of a digression here)

For now, I’ll concede that some further research on where economic damage tends to hit would help give a sense of how big a factor this is.

Secondly, my argument isn’t that the FME completely cancels economic harm, only that it shows it to be substantially less pronounced than one might think. Given that the government receives over a third of the value of UK GDP (and similar amounts in other developed countries), that alone makes the effect quite powerful. If the spread of damage does turn out to be relatively even, the effect will be huge.

(1) Plenty of people would argue that banking has a positive effect. This is broadly irrelevant to the argument for the FME – the margins look flat in both directions – though obviously it means you might apply it differently.

(2) This is a slightly simplistic calculation, only considering the governmental aspect:

£10 earned subtracts (on the above estimate) £70 of value of the overall economy. If 30K-per-QALY treatments on the NHS are the worst government expenditure, then I assume the £70 comes from there, in which case the banker’s job costs 0.0023 QALYs in economic harm. The current best cause as judged by Giving What We Can and Givewell buys one QALY for approximately US$25, or £15.56 at current exchange rates, so an effective altruist could buy back the 0.0023 of a QALY for slightly less than four pence.

I had a lot of useful feedback when putting together this argument – thanks to Ben Todd, Ryan Carey and Patrick Brinich-Langlois for some very useful comments, and to Rob Wiblin for prompting me to think about the subject in the first place.