A system is not a person

Public problems are conveniently explained by defective people: the poor are lazy, officials are corrupt, the boss is bad, voters are stupid. Personal qualities matter, but the system determines which behaviour is rewarded and which is possible at all.

Thomas Schelling showed how moderate individual preferences are able to produce an extreme collective outcome. In his simple model people did not demand a completely uniform neighbourhood; it was enough for them not to end up in too obvious a minority. A series of local moves still led to sharply divided districts.

The system here has no secret design. Each participant reacts to their immediate surroundings, and the outcome emerges from the reactions combined. So it is sometimes pointless to look for a single villain, though the rules and incentives will still have to be changed by particular people.

Poverty exists because the poor are lazy. Wealth, because the rich are talented. Corruption, because officials are bad. War, because one leader is insane.

Personal qualities matter. But the system determines which behaviour is rewarded, which is punished and which is possible at all.

An honest person in an opaque system can lose to someone who knows how to use connections. A kind manager in an organisation with impossible deadlines will still pass the pressure downwards. A politician with good intentions depends on the party, the funding, the bureaucracy and coalitions.

This does not remove personal responsibility. It shows why replacing one person sometimes changes almost nothing.

To understand a system it helps to look at:

• incentives;
• constraints;
• flows of money;
• access to information;
• how people are appointed;
• control mechanisms;
• the distribution of risk;
• the possibility of exit;
• feedback.

If a system produces the same result as the participants change, the cause probably does not lie in their characters alone.

Changing a rule is often more effective than endlessly searching for exceptionally good people. Good people also get tired, make mistakes and respond to incentives.

At the same time a good rule does not enforce itself. Any construction ultimately depends on people able to notice that it has stopped performing its original function.

Society is not a machine. It is more like a machine assembled from parts, each of which has its own opinion about the blueprint.

The test for a system is done by subtracting people. If the third sales manager in a row ends the year with falsified figures, three characters are unlikely to be the issue. More likely the target is set above what is achievable and the bonus is paid only for hitting it: under those conditions the honest employee simply leaves earlier than the rest, and the ones who stay are the ones who know how to draw the numbers. The fourth manager will be more honest than the previous ones right up to the end of the first quarter.

Hence an unpleasant corollary for anyone who likes looking for the guilty party: a system that gives the same result with different people in it is fixed by changing the rule, not by firing someone. The firing usually happens anyway — it is quicker and looks more like action.