Inequality and Risk

Paul Graham argues that economic inequality is good because people want to be rich. In order to achieve this, they take risks (both by investing capital and and founding companies); the resulting innovation and growth benefit society. He claims the only way to reduce inequality is to take money from the rich. This reduces the incentives for risk-taking, which in turn threatens growth and the ability of a country to resist poverty and subjugation to others. I recommend reading his original essay: he’s a good writer (though this is certainly not his best).

Graham argues convincingly that inequality encourages risk-taking, and that risk-taking leads to growth. But the argument lacks balance: he treats reducing inequality and eliminating inequality interchangeably. Worse, he doesn’t account for the benefits of reduced inequality and taxation. His reasoning reduces the success of a society to the relationship between inequality, risk, and innovation1.


Graham says that, “reducing economic inequality . . . is identical with taking money from the rich” – which by removing the incentive to take risks and innovate, threatens growth. But rich people are naturally risk-averse: they have the most to lose. So by Graham’s argument, it might make sense to take some money from the rich in order to encourage continued innovation2. As for the poor, they have little lose and everything to gain; it is they who are natural risk takers. So it is equally reasonable to criticize giving money to the poor.

This latter argument is quite common, first on the same grounds as Graham (that giving money reduces the incentive to work), and second because it eliminates spare capacity in the labor market (i.e. unemployment), which keeps wages and inflation down. Of course the second reason threatens the first, for limiting the gains the poor can make (by keeping wages low) also weakens the incentive.

The evidence is that the poor do take greater risks, as the popularity of lottery tickets demonstrates. Rich investors, on the other hand, tend to avoid risk. Only a few investors are venture capitalists3; many more pursue low-risk blue-chip investments or engage in speculation and short-term trading. Neither of these does much to encourage innovation.

Market Failure

There is much truth in Graham’s argument that inequality promotes risk-taking. But it does not follow that maximum inequality correlates with maximum growth. Quite the opposite: the world is full of destitute countries combining desperate poverty with extraordinary wealth. Their inequality hardly correlates to economic or social success. Nor is it clear that maximum growth is necessarily desirable: other western countries with less inequality than the United States (chiefly due to higher taxes on the wealthy) are hardly basket cases. Growth statistics are notorious for their weakness in measuring real-world wealth (e.g., reconstruction following a disaster counts as “growth”, but the creation of social assets like happiness and business relationships does not). America may have the most money, but it also has great poverty and the highest rate of incarceration of any western country. The result of reducing inequality need hardly result in “making your whole country poor”. By eliding reducing inequality with eliminating it, and citing growth as the sole measure of a society’s success, Graham sidesteps nearly all of the other factors that make a society work, and avoids any dispute about the trade-offs between one good (risk incentives, say) and another (such as stability).

Profit-seeking individuals and corporations are constitutionally ill-suited to taking into account “externalities”, i.e. the costs that their actions impose on others – even when collectively they too suffer the consequences. Environmental damage, social problems, and threats to democracy and all fall into this category. These kinds of market failures must be addressed by other mechanisms. For example, public institutions are more effective at delivering services when competition is undesirable or costs and risks are best pooled. These can include law, education, health care, policing, transportation, energy, sanitation, environmental regulation, standards, disaster relief, defense, diplomacy and so on4. The rule of law allows entrepreneurs to devote their risk-taking to innovation by increasing predictability and reducing risk through enforceable contracts. Other programs similarly provide benefits like security, education, and quality of life that allow businesses – especially small ones – to focus on what they do best. Graham recognizes this when he mentions that educating the poor can increase their wealth and inequality at the same time. Entrepreneurs may feel that they have succeeded alone, but that is an illusion: the money that governments take from the rich is the profit not just of individual enterprises but of the whole of society.

What is relevant is that these services are paid for disproportionately by progressive taxes. As economists say, increasing wealth suffers from “diminishing marginal utility”, i.e., the more money you have, the less significant any additional income becomes. Accordingly, $100 taxed from a wealthy person may be insignificant to them, where a homeless person would be devastated by such a loss; yet that $100, when put towards to the sorts of services listed above, has the same value regardless of its source. It may even provide more benefit to the rich taxed than when untaxed5. This is one case where taking from the rich benefits everyone.


Graham addresses one market failure in detail. He suggests that the main reason for the opposition to inequality is the translation of wealth into power, and that the solution is to attack corruption through greater transparency (presumably enforced by tax-funded regulation). It’s a good idea, but eliminating corruption (were that even possible) will not break the connection between wealth and power. To the extent that we meet our needs through private enterprise, business power is political power. This power lies disproportionately in the hands of those who influence business decisions, including company officers, shareholders, and consumers. In the case of shareholders and consumers, this influence corresponds directly to wealth. Economic inequalities lead to power inequalities regardless of corruption (corruption may in fact serve to help).

These inequalities of power are self-reinforcing: rich folks want their kids to also be rich. They tend to believe in the competence of people like themselves, who are also generally rich, and so the same class of people is also hired and promoted. This kind of nepotism need not manifest as corruption. People understand the tendency to hire friends and family as human; transparency may therefore be ineffective at reducing such behavior. Yet it reduces productivity because it reduces the impact of merit and effort (e.g. risk-taking) on success. Those poor people who are willing to take risks (because they want to be rich) may become discouraged because they know they lack the advantages possessed by the rich folk. The rich folk, meanwhile, are less likely to take risks because they have more to lose, and in fact have an incentive to use their power to prevent anyone else from innovating and threatening their situation.

Again, I agree that inequality encourages people to take risks and thereby aids productivity. However, it also creates power differentials and class differences which reduce productivity. It seems likely that a balance is to be preferred over extremes of inequality. Taking from the rich may be beneficial: it reduces the imbalance produced by power differentials, it encourages rich folks to continue working, and it signals to the poor that the competition is more fair.

Maximum Change

There is one other assumption in all of this: that innovation should be maximized. Not all innovations are good. Contentious technologies include nuclear weapons, the replacement of streetcars with buses, and designer drugs. But even if we accept that innovation is good, maximum innovation may not be. Societies and cultures often have difficulty adapting to rapid change. The result can be tragedy, from dire environmental damage to the barbarism of the early twentieth century. Such catastrophic risks, combined with the other costs and risks of great economic inequality, surely indicate that innovation cannot be pursued as an ideal isolated from other social concerns.

Paul Graham provides good illustrations of the benefits of risk-taking, and he explains the dangers of eliminating economic inequality. But he goes too far when he argues against reducing economic inequality in general. His innovative start-ups are just one factor in a complex environment: there are other important costs to inequality, including threats to the entire system. Reducing inequality can be an important tool both for encouraging innovation and for managing changes it brings about.


1 See also Tim Bray’s rebuttal of Graham’s argument.

2 This is one of the reasons why copyright terms should not be unlimited: past works turn into cash cows for large corporations, which remove the incentive to fund new material. Placing such works in the public domain promotes creativity without denying society the wealth that the old works represent.

3 The prevalence of venture capital in Graham’s industry is a special case. In many industries (and in Canada), VCs are few and far between. Furthermore, most entrepreneurial businesses – although risky – are not innovative (restaurants and corner groceries, for example).

4 One may quibble over the list (I certainly think there are roles for the market in many of these areas – for example, energy and water price regulation can encourage overconsumption), but the fact of market failures is beyond dispute.

5 Of course, it is only natural that anyone – rich folk included – would prefer that someone else pay the tax. It might even be to their benefit to bluff and oppose taxes which benefit them: if others with more to gain pay the tax, then they will still benefit. This applies whether the “bluffing” is conscious or not.