We conclude that in the Roman Empire as a whole, a ‘middling’ sector of somewhere around 6 to 12 per cent of the population, defined by a real income of between 2.4 and 10 times ‘bare bones’ subsistence or 1 to 4 times ‘respectable’ consumption levels, would have occupied a fairly narrow middle ground between an élite segment of perhaps 1.5 per cent of the population and a vast majority close to subsistence level of around 90 per cent. In this system, some 1.5 per cent of households controlled 15 to 25 per cent of total income, while close to 10 per cent took in another 15 to 25 per cent, leaving not much more than half of all income for all remaining households.
This estimate gives imperial Rome a Gini index roughly equivalent to the United States today. This is, however, an example of the limitations of the measure. As Milanovic, Lindert and Williamson noted, in pre-industrial societies the necessity of leaving the laboring poor enough food for them to subsist, work and replace themselves and the relatively low productivity of such societies limited the surplus production that elites could control.
Strikingly, they estimate that 10-20% of the population had incomes below subsistence: the unemployed or underemployed: not actually starving to death, but malnourished, unhealthy, stunted and with energy only for minimal activity. This underclass was disturbingly ubiquitous in pre-industrial societies.
Compare imperial Rome with England and Wales in 1688. This was before the industrial revolution, but England benefited from global trade, the printing press, and New World plants. According to Gregory King's estimate, it had a much larger middle class comfortably above subsistence and had a higher average per capita income. As a result, its Gini measure of inequality was significantly higher than that of imperial Rome.
And yet, in a real sense the England of 1688, becoming a nation of shopkeepers, had greater equality of welfare.
Or consider this thought experiment: imagine an imperial Rome with the incomes of the top and bottom 10% doubled, and other incomes unchanged. Measured by Gini coefficient, there's a significant increase in inequality. In terms of equality of welfare, there's a big improvement, since bottom 10% move from going to bed hungry to not, and the top 10% just get better stuff.