1 The Causal Question

Can a colonial institution abolished two centuries ago still shape poverty today? This is the central question of Dell [2010], one of the most elegant applications of a geographic regression discontinuity design in the historical development literature. The paper asks: did the Spanish colonial forced labour system known as the mita cause the communities subject to it to be significantly poorer today than communities just outside its boundary?

The answer, documented with meticulous attention to identification, is a striking yes. Communities that fell within the mita boundary consume roughly 25% less and have 6 percentage points higher rates of childhood stunting than communities just outside, despite the mita's abolition in 1812. Understanding how Dell establishes this causal claim and what mechanisms it reveals is the subject of this case study.

2 Historical Background: The Mita System

The mita (from the Quechua word for "turn" or "season") was a system of draft labour established by Viceroy Francisco de Toledo in 1573. The Spanish Crown required communities within a defined catchment area roughly 200 miles around the mining centre of Potosí, in present-day Bolivia to send one-seventh of their adult male population to work in the silver mines each year.

The boundary of the mita catchment was drawn on the basis of travel time to Potosí: communities within a specified distance were included; those just beyond were exempt. This boundary, established in 1573, was largely unchanged until the mita's abolition in 1812. It cuts through the altiplano the high Andean plateau spanning southern Peru and Bolivia- creating a geographic discontinuity that Dell [2010] exploits more than four centuries later.

The mita was brutal. Mortality in the mines was high, and forced labour disrupted agricultural production and community organisation. But its effects did not end with the abolition. Dell documents that the mita led to the consolidation of large haciendas (estates) inside the boundary, while outside the boundary communities maintained indigenous community (comunidad indígena) structures with stronger collective land tenure. These differential institutional trajectories persisting through independence, land reform, and market integration are the mechanisms through which the mita continues to affect outcomes today.

3 Identification Strategy: Geographic Regression Discontinuity

The key identification insight is that the mita boundary was drawn for administrative convenience travel time to Potosí rather than to maximise or minimise any economic outcome. Communities just inside and just outside the boundary were otherwise similar: same altitude, same climate, same pre-mita ethnic composition, same access to markets. The boundary generates a sharp discontinuity in mita exposure that is plausibly unrelated to any confounding determinants of contemporary poverty.

3.1 The Estimator

Let mᵢ ∈ {0,1} indicate whether district i was inside the mita boundary, and let dᵢ denote distance from district i to the mita boundary (signed negative for inside, positive for outside).  The geographic RD estimator is:‍


where yᵢ is the outcome (log household consumption or childhood stunting rate), f(dᵢ) is a flexible polynomial in distance from the boundary that controls for smooth geographic gradients, and Xᵢ includes geographic controls (elevation, slope, proximity to Potosí). The coefficient β identifies the effect of mita exposure on contemporary outcomes for communities near the boundary.

The boundary is irregular it follows mountain contours and watershed boundaries rather than a straight line. Dell handles this by computing the minimum Euclidean distance from each district's centroid to the boundary, and by including the distance to Potosí as an additional control to absorb any direct gradient from the mining centre.

3.2 Boundary Fixed Effects

A further innovation is the inclusion of boundary segment fixed effects. Dell divides the mita boundary into segments and includes indicators for which segment each district is closest to. This absorbs any heterogeneity across parts of the boundary, so identification comes from within-segment, cross-boundary variation. This is equivalent to a matched comparison of adjacent communities on either side of the same boundary segment the cleanest possible implementation of the geographic discontinuity design.

Figure 1: Schematic of the geographic RD design around the mita boundary. The regressionestimates the outcome jump at di = 0.

4 Data and Setting

Dell links several data sources spanning four centuries:

• Colonial records: Spanish colonial tribute and census records from the late 1600s document the extent of the mita catchment area.

• Contemporary household surveys: The 2001 and 2006 Peruvian household surveys (ENAHO) provide measures of per-capita household consumption, used as the primary outcome.

• Childhood stunting: Height-for-age z-scores from the Demographic and Health Survey (DHS) capture long-run nutritional deprivation.

• Geographic covariates: Altitude, slope, soil quality, and distances to markets from GIS sources provide controls for geographic heterogeneity.

The analysis focuses on districts within 100 kilometres of the mita boundary, a bandwidth chosen to balance the precision of geographic RD (which increases with smaller windows) against statistical power. The sample comprises approximately 1,400 district-level observations.

5 Key Findings

5.1 Consumption and Stunting

The headline result is that districts inside the mita boundary have log consumption approximately 0.27 standard deviations lower than comparable districts just outside. This corresponds to roughly 25% lower consumption. The childhood stunting rate the share of children whose height-for-age is more than two standard deviations below the WHO reference is 6 percentage points higher inside the mita boundary.

Both results are robust to a range of polynomial specifications for f(dᵢ), alternative bandwidth choices, and the inclusion or exclusion of geographic controls. They persist when the sample is restricted to ever-smaller windows around the boundary, consistent with a genuine discontinuity rather than a smooth gradient.

5.2 Mechanisms: Haciendas, Road Infrastructure, and Public Goods

Dell documents three interacting mechanisms:

1. Hacienda consolidation. The mita facilitated the consolidation of large private estates inside the boundary by undermining indigenous community land tenure. Hacienda prevalence in the late colonial period is significantly higher inside the boundary. Colonial-era haciendas were associated with extractive labour relations and low investment in tenant welfare.
2. ‍Road infrastructure. Districts inside the mita boundary have significantly less road infrastructure today. Dell argues that hacienda owners historically lobbied against roads that would have allowed their tenants to seek work elsewhere, while comunidad-structured communities outside the boundary built roads to access markets. ‍
3. Public goods provision. School attendance rates and access to public services are lower inside the mita boundary, consistent with lower state capacity or community collective action in areas shaped by extractive colonial institutions.

Formally, Dell estimates the effect of mita exposure on each mechanism using the same geographic RD design, then asks whether the contemporary poverty gap is mediated by these mechanisms. The mediation analysis is suggestive rather than causal the mechanisms are themselves potentially endogenous but the pattern of results is internally coherent.

6 Validity Checks

Dell [2010] conducts an unusually thorough set of specification tests:

• Manipulation test. There is no discontinuity in the density of district centroids at the mita boundary (analogous to the McCrary density test), ruling out selection of districts into mita status.

• Covariate balance. Pre-treatment geographic characteristics altitude, slope, distance to rivers show no discontinuity at the boundary, consistent with the boundary being drawn for administrative rather than geographic reasons.

• Placebo boundaries. Using falsified boundaries shifted 50 kilometres inward or outward from the true boundary yields near-zero effects, confirming that the estimated effects are not an artefact of geographic gradients.

• Alternative outcomes. The pattern of results holds for multiple proxies of long-run development: school attendance, access to electricity, land titles.

7 Limitations and Lessons

The design is compelling but not without limitations. First, as with all geographic RD designs, the estimates apply to communities near the boundary and may not generalise to more heavily mita-exposed communities deep inside the catchment. Second, several potential confounders altitude, proximity to Potosí, ethnic composition are controlled through the polynomial f(dᵢ) and covariates, but unobserved geographic heterogeneity at the boundary cannot be ruled out definitively.

Third, the mechanisms are identified under additional assumptions beyond the boundary exogeneity maintained for the reduced-form estimate. The mediation analysis relies on sequential ignorability conditions that may not hold in a setting with two centuries of compounding institutional change.

Despite these caveats, Dell (2010) remains a methodological benchmark. It demonstrates the power of geographic discontinuities for historical identification, introduces boundary segment fixed effects as a tool for absorbing spatial heterogeneity, and provides a rich institutional account of the persistence mechanisms connecting colonial extraction to contemporary poverty.

8 Conclusion

Dell [2010] establishes that the Spanish colonial mita caused persistent poverty in the communities it burdened, with effects visible more than two centuries after its abolition. The identification strategy a geographic regression discontinuity at the mita boundary, combined with boundary segment fixed effects and rich specification testing sets a high bar for the historical institutions literature. The paper's finding that institutions matter for long-run development is not new; what is new is the rigour with which it is documented for a specific, well-defined institutional intervention.

References

1. Calonico, S., Cattaneo, M. D., and Titiunik, R. (2014). Robust nonparametric confidence intervals for regression-discontinuity designs. Econometrica, 82(6):2295-2326.
2. Dell, M. (2010). The persistent effects of Peru's mining mita. Econometrica, 78(6):1863-1903.
3. Imbens, G. W. and Lemieux, T. (2008). Regression discontinuity designs: A guide to practice. Journal of Econometrics, 142(2):615-635.
4. Acemoglu, D., Johnson, S., and Robinson, J. A. (2001). The colonial origins of comparative development: An empirical investigation. American Economic Review, 91(5):1369-1401.
5. Nunn, N. (2008). The long-term effects of Africa's slave trades. Quarterly Journal of Economics, 123(1):139-176.
6. Lee, D. S. (2008). Randomized experiments from non-random selection in U.S. House elections. Journal of Econometrics, 142(2):675-697.

‍