# The Persistence of Poverty

Normally we think of goods in terms of declining marginal utility. You may pay \$5 for the first slice of cake but only \$2 for the second. A tenth slice might only be worth \$0.10 to you. But not everything works in terms of declining marginal utility. There are situations where marginal utility increases.

Consider the ‘screaming room’ from Mike Konczal’s summary of the main ideas in Charles Karelis’ book The Persistence of Poverty: Why the Economics of the Well-Off Can’t Help the Poor.

Now picture you are in a room with 10 people screaming. You hate it when people scream, and you can pay a person to get them to stop screaming. Would you pay in a similar way to the cake example? Would you pay a \$1 to get the first person to stop screaming, and a penny for the 10th person to stop screaming?

No. Getting one person to stop screaming would make very little difference in how much you dislike being in the room. Modern psychology tells us you might not even notice it. You’d probably only pay a penny to get that first guy to stop screaming. However getting the second guy to stop screaming might be worth 10 cents. And the last guy, the difference between some screaming and no screaming, might be worth the full dollar to you. The more quiet it got, the more a marginal difference in how quiet it is would be worth to you. There’s increasing returns to this good; the 10th guy not screaming is worth more than the first guy not screaming, which is the exact opposite dynamic of the 10th cake being less delicious than the first.

For those not involved with economic theory this might just elicit a shrug, but this mechanism turns everything on its head. Let’s say that instead of money, you are given 20 tokens to be used over 4 days, and each token gets you one slice of cake in room #1, and one person to stop screaming in room #2. In the cake room, the optimal decision is to consumption smooth – eat five slices of cake each day, so you use the tokens {5,5,5,5}. In the screaming room, all the enjoyment is not in getting a room with half screaming but in getting a quiet room, and instead of consumption smoothing the optimal choice is to binge – pay 10 people to stop screaming the first two days, and deal with a loud room the last two days – {10,10,0,0}. This will hold even with ‘nudges’, say offering two extra tokens if you have people consumption smooth, since the marginal utility isn’t increasing that much. The utility of {10,10,0,0} is greater than that of {5,5,5,7}.