Tuesday, October 3, 2017

The long-term effect of slavery on inequality today

According to a new working paper, 1800s slavery explains approximately 20% of income inequality in Brazil today. While the direction of the impact is not surprising, I'm impressed by its magnitude. I wonder how much investment in cash transfer programs would be necessary to achieve an effect of similar magnitude. Thanks John B. Holbein‏ for pointing to this study on Twitter.


Fujiwara, T., Laudares, H., & Caicedo, F. V. (2017). Tordesillas, Slavery and the Origins of Brazilian Inequality.

From the abstract:
"...To deal with the endogeneity of slavery placing, we use a spatial Regression Discontinuity framework, exploiting the colonial boundaries between the Portuguese and Spanish empires in current day Brazil. We find that the number of slaves in 1872 is discontinuously higher in the Portuguese side of the border, consistent with this power’s comparative advantage in this trade. We then show how this differential slave rate has led to higher income inequality of 0.103 points (Gini coefficient), approximately 20% of average income inequality in Brazil. To further investigate the role of slavery on economic development, we use the division of the Portuguese colony into Donatary Captancies. We find that a 1% increase in slavery in 1872 leads to an increase in inequality of 0.112. Aside from the general effect on inequality, we find that more slave intensive areas have higher income and educational racial imbalances and worse public institutions today"

ps. This paper also reminds me of this post on how presidential elections are impacted by a 100 million year old coastline in the USA. Hint: geology determined the distribution of productive land, which influenced the spatial distribution of African slaves which in turn influenced the electoral distribution. I'm not saying I'm convinced by this argument but I have to recognize it uses a quite inventive identification strategy.
image credit: Fujiwara et al (2017)

5 comments:

Raphael Gouvea said...

I have not read the full paper yet (I only glanced over it). Having said that, the large effect seems to be the result of generalizing the local treatment effect to overall inequality. If this is indeed the case, I would take the implications for overall inequality with caution...

Rafael H M Pereira said...
This comment has been removed by the author.
Rafael H M Pereira said...

Thanks for the heads up, Rapha! It also crossed my mind that perhaps the authors should have considered spatial auto-correlation in the analysis but it is hard to tell beforehand if this would significantly change the results.

Unknown said...
This comment has been removed by a blog administrator.
Eduardo Luís Baptista said...

Interesting, but what about other phenomena that could change this equation during the last 150 years such as migration? Explaining inequality geographically seems to me that it's a much harder task, which involves other significant criteria as well.
Valeu!