To calculate a worldwide measure of income inequality, you need to work with data on the distribution of income for the population in every country--and for many countries, this data is mismatched and helter-skelter. You need to convert the income data for all countries into a common currency, like U.S. dollars. You then add up all the people in the world who fall into each income category. To do comparisons over time, you need to find data for different countries over time, and then also adjust for inflation. Christoph Lakner and Branko Milanovic of the World Bank take on this task in "Global Income Distribution From the Fall of the Berlin Wall to the Great Recession," published this month as Policy Research Working Paper 6719.
Here's how the global distribution of income has shifted over time. It used to be said back in the 1960s that the global distribution of income was bi-modal--that is, it had one hump representing the large number of people who were very low-income and then a smaller hump representing those in the high-income countries. In the blue line for the global distribution of income 1988, the remnants of that bimodal distribution are still visible. But over time, the highest point in the income distribution is shifting to the right, and by 2008, the world has moved fairly close to having a unimodal or one-hump distribution of income.
An obvious question here is to what extent these changes are about what has happened in China and in India, which after all combine to include about one-third of the world's population. Here's a figure showing the shifts over time, with the population of India and China shaded separately. You can see that the shift in the shape of the global income distribution an largely be traced to India and China. In particular, you can see that the hump to China is pretty much centered over the hump for India in 1988 and 1993, but by 2008, the hump for China is now shifting to the right more than the hump for India, reflecting China's faster rate of economic growth.
One final thought: In interpreting these charts, it's important to remember that the horizontal axis measures income on a logarithmic graph. That is, instead of each horizontal distance representing the same absolute gain in income, it represents the same proportional gain income. Starting from the left, the horizontal distance from $50 to $100 is the same as the distance from $100 to $200, which is the same as the distance from $200 to $400, or if you look off to the right, the sqame as the distance from $10K to $20K. In other words, relatively small movements to the right on this graph represent large changes in absolute value of incomes, especially as you get to the center and far-right of the graph.
Note: Hat tip to Howard Schneider at the Washington Post Wonkblog, where I saw the Lakner-Milanovic working paper mentioned.