Sebastian Olascoaga

The puzzle

Mexico's Gini fell from 2008 to 2012. Top-fractile income shares climbed over the same years.

By 2013, Mexico's official Gini coefficient had fallen from 0.59 to roughly 0.49, the kind of trend that featured in international reports celebrating Latin America's "inequality improvements" of the early 21st century. The household-survey infrastructure those Ginis rest on (ENIGH and its peers) was doing what household surveys do. It was undercounting the top.

This thesis asked whether, with access to the full set of Mexican personal-income tax microdata, you could see the part of the distribution that surveys cannot. The methodological challenge is harder than it sounds. In a country with substantial tax noncompliance, even the tax data do not tell the truth at face value. You need to know how much, and where, taxpayers are inflating their deductions and understating their income.

How we did it

2.5 million tax returns a year, corrected with an "anonymous disclosure" benchmark from the Economic Census.

The data is the universe of Mexican personal-income tax microdata 2009 to 2012, about 2.0 to 2.5 million personal returns and 20 to 25 million employer-reported wage records per year, accessed through an OECD-Mexican government collaboration. The income denominator is built from INEGI's national accounts (Balance of primary income, adjusted, ≈59 to 60% of GDP). The adult-population control comes from 2005 and 2010 census interpolations.

The methodological contribution is a discrepancy-based evasion adjustment, adapted from Fiorio & D'Amuri (2006) and Matsaganis et al. (2010). The intuition is straightforward. When filers face a real-time tax incentive to overstate deductions, they will. When they don't (as in the anonymous, mandatory 2009 Mexican Economic Census, covering 5.14 million business units) they won't. Comparing average deductions across the two sources gives an evasion rate for each category, which deflates the tax-declared values to reconstruct true income. Three income definitions (Revenue, Net, and Adjusted) bracket the truth.

What it shows

Concentration at the top is static, and the same individuals hold those positions across years.

Top shares. After adjustment, the top 1% holds 13.2% (2009) rising to 13.6% (2012) of national income. Top 0.01%, 3.2% in 2012. Top 0.001%, 1.7% in 2012 (peak 1.85% in 2011). Definition matters dramatically. The 2012 top 1% share is 10.2% (Net) vs 13.6% (Adjusted) vs 19.5% (Revenue). The revenue measure puts Mexico above nearly every country in the WTID except Colombia.

Growth capture. Of all real income growth in Mexico 2009 to 2012, the top 1% captured 8%, the top 0.1% captured 5%, the top 0.01% captured 3%, and the top 0.001% captured 2%. Real income growth within the top 0.001% was +40.8% in three years.

International standing. Mexico's top 0.01% share is 3.50× Spain's, 2.70× France's, 1.73× Uruguay's, and 1.04× the United States'. The inverted Pareto coefficient at P99.99 reaches 3.89 in 2011. "At the same level of the 1920s of the USA and the UK," in the thesis's own words.

The locked-in part. Spearman rank correlation between 2009 and 2012 individual incomes is 0.81 (p = 0.001). Of those in the top 0.01%+ in 2009, 52.89% filed there again in 2012, well above the Colombian benchmark of 26.7 to 33.7%. Restricted to taxpayers who filed in both years, 60.67% stayed. Mexico's super-rich are dense, growing faster than everyone else, and staying put.

Taxation of the top. Effective average tax rates of top fractiles range from 3.6% to 13.36%. The top 0.01 to 0.001% has taxable income equal to only ~20% of revenue, while the 5 to 1% fractile has ~80%. A structure of legal exemption that the thesis documents in painstaking institutional detail.