2.1.2. Piketty: the Accidental Pattern of Income Inequality

In a volume of over 800 pages, Piketty (2001) ⁹ uses individual income tax returns (1915–98), wage tax returns (1919–98) and inheritance tax returns (1902–94) to create a homogeneous data set of income inequality, wage inequality and wealth inequality. His findings question the arguments supporting the belief of a natural fading away of income inequality over time. “Mostly accidental”: this is how Piketty qualifies the decline in income inequality that took place during the first half of the twentieth century in France, overall as being related to the historical hazards of capital income and progressive taxation on very large fortunes.

Piketty’s publication of 2001 triggered a series of other empirical analyses applying the same methodology to different countries, as commented by Atkinson (August 2003):

‘There has recently been a revival of interest among economists in the distribution of top incomes. The pioneering study by Piketty (2001) produced estimates of the long-run distribution of top incomes for France. Following his lead, Atkinson (2002) made estimates for the United Kingdom, and Piketty and Saez (2003) made estimates for the United States. Estimates are now also available for Germany (Dell, 2002), Canada (Saez and Veall, 2003), Netherlands (Atkinson and Salverda, 2003), India (Banerjee and Piketty, 2001), and Australia and New Zealand (Atkinson and Leigh, 2003). In using data from the income tax records, these studies use similar sources to the earlier work of Bowley (1914) and Stamp (1916 and 1936) in the UK, and Kuznets (1953) in the US. (See also the survey by Kraus, 1981.) The findings of the recent papers are however of added interest, since the data provide estimates covering nearly all of the twentieth century – a length of time series unusual in economics. Moreover, the techniques are considerably more developed. ’

Thomas Piketty and Emmanuel Saez (2004) shed light on the empirical aspects of income inequality in the United States, using the Internal Revenue Service’s (IRS) annual publications on Statistics of Income to create a homogeneous time-series data set on the upper shares of income and wages from 1913 to 1998. The methodology used by Piketty and Saez follows very closely that used by Kuznets (1953) on Upper Income Groups in Income and Savings. There are several differences, though. The first difference lies on the definition of the upper income shares, based on individuals for Kuznets, and tax units for Piketty and Saez. In their paper (long version, 2004, p. 4), a “tax unit is defined as a married couple living together (with dependents) or a single adult (with dependents), as in the current tax law.” Because “Kuznets did not correct for the re-ranking,” he “misclassified in the top shares large families with high total income but moderate income per capita.” ¹⁰ Relative to Saez and Piketty’s, Kuznets’ shares are therefore underestimated. This discrepancy does not impact the pattern of upper income evolution over years though, because the decrease in the number of individuals per tax unit occurred across all income groups over the century.

The second difference deals with the treatment of capital gains and numerous other data adjustments. They received much less attention in Kuznets’ work of 1953 than in Piketty and Saez (2004). One should be aware, however, that the IRS micro-files released from 1960 to 1995, and improving the statistics of the initial IRS tables, significantly helped Saez and Piketty in the corrections of their estimates, were not at the disposal of Kuznets in his time.

The evidence for rising inequalities in the U.S. since the 1970s leads the authors to re-interpret the Kuznets curve as a sinusoidal graph. “A new industrial revolution has taken place, thereby leading to increasing inequality, and inequality will decline again at some point, as more and more workers benefit from the new innovations.” This argument matters even more once applied to the state-level analysis as the technological revolution occurred first on both the east and the west coasts.

The lack of empirical evidence supporting the theoretical reduction in inequalities over time also holds in the convergence literature.