1 The Causal Question

Disease burden is widely recognised as a constraint on economic development. But does eradicating a specific disease cause measurable, long-run improvements in human capital and income? This question is difficult to answer causally: health interventions are rarely randomly assigned, and regions that receive health investments may differ systematically from those that do not.

Bleakley [2007] provides one of the most celebrated attempts to answer this question, exploiting the Rockefeller Sanitary Commission's campaign to eradicate hookworm in the American South between 1910 and 1920. The paper is a landmark in both the economics of health and development and the application of difference-in-differences methods to historical data.

2 The Setting: Hookworm in the American South

Hookworm (Necator americanus) is a soil-transmitted parasite that enters the body through bare feet, migrates to the intestine, and feeds on blood. In children, it causes anaemia, stunting, and cognitive impairment; in adults, chronic infection reduces work capacity and productivity. By the early twentieth century, hookworm was endemic in the American South, with infection rates in some counties exceeding 40 percent of the population, particularly among poor, rural, and shoeless populations.

The Rockefeller Sanitary Commission for the Eradication of Hookworm Disease (RSC) was founded in 1909 and operated from 1910 to 1914 in 11 Southern states. The RSC provided treatment (thymol and Epsom salts, effective against the parasite) through county health dispensaries, and conducted educational campaigns about sanitation and shoe-wearing. By the end of the campaign, hookworm prevalence had fallen dramatically in affected areas.

The key empirical feature of the setting is that hookworm was not uniformly distributed across the South: it thrived in sandy, warm, moist soils. Counties with these soil characteristics had much higher pre-treatment infection rates. This differential pre-treatment exposure creates the identifying variation.

3 Identification Strategy

Bleakley [2007] uses a difference-in-differences design that exploits variation in two dimensions:

1. Pre-existing hookworm burden. Measured by county-level hookworm infection rates from RSC surveys conducted before the eradication campaign. High-burden counties experienced a larger "treatment" from the campaign than low-burden counties.
2. ‍Cohort of birth. Children born after the campaign began (post-1910) were exposed to less hookworm during childhood than children born before it. Children born in high-burden counties experienced a larger reduction in exposure than those in low-burden counties.

The identifying variation is the interaction of county-level hookworm burden with birth cohort: did children born in high-burden counties after the RSC campaign improve their outcomes more than children born in the same counties before the campaign, relative to similar cohort comparisons in low-burden counties?

The estimating equation is:

where yᵢ꜀ₛ is the outcome (school enrollment, literacy, occupational income score) for individual i in county c in survey year s;  Hookworm is the pre-campaign hookworm infection rate in county c; 1[born post] is an indicator for birth after the campaign;  γ꜀ are county fixed effects; δₛ are survey-year effects; and Xᵢ꜀ₛ are individual-level controls. The coefficient β identifies the causal effect of hookworm eradication under the parallel trends assumption.

3.1 The Identifying Assumption

The key assumption is that, absent the RSC campaign, the gap in outcomes between high-burden and low-burden counties would not have been changing differentially for children born after 1910 relative to those born before. This is a conditional parallel trends assumption: controlling for county and year effects, any pre-existing trend differences between high- and low-burden counties should be orthogonal to hookworm exposure.

Bleakley [2007] validates this in several ways. First, he examines pre-trends: did high-burden and low-burden counties diverge in cohort outcomes before 1910? He finds no evidence of pre-existing differential trends in school enrollment or literacy, supporting the parallel trends assumption. Second, he conducts a placebo test using northern states, where hookworm was rare: if the campaign explains the results, the same birth-cohort interaction should not appear in states that were not targeted. He finds no effect in the North, consistent with the causal interpretation.

4 Data

The study uses two main data sources. The 1920 Decennial Census of Population provides individual-level data on school enrollment (for children aged 10-14) and literacy (for adults aged 18-40), merged with county-level hookworm burden from RSC pre-campaign surveys. The 1940 Census provides occupational income scores and other long-run adult outcomes for cohorts exposed to the campaign in childhood.

A key challenge is that the RSC surveys measured infection rates for a subset of counties (those that received dispensary visits). Bleakley [2007] uses soil characteristics (sandy loam texture, temperature, humidity) as an additional proxy for hookworm suitability, and shows results are similar using either measure.

5 Key Results

5.1 Short-Run Effects on Schooling

Comparing children aged 10-14 in the 1920 Census, Bleakley [2007] finds that in high-hookworm-burden counties, school enrollment increased significantly more for cohorts born after 1910 than for those born before, relative to low-burden counties. The estimated effect implies that a 10-percentage-point reduction in county hookworm infection rate increased the school enrollment rate by approximately 5-8 percentage points—a large effect given that enrollment rates in the sample averaged around 60 percent.

5.2 Long-Run Effects on Literacy and Income

Using the 1940 Census, Bleakley [2007] finds persistent long-run effects on adult outcomes. Individuals who were children in high-burden counties during the RSC campaign have higher literacy rates and higher occupational income scores by 1940, relative to similar cohorts in low-burden counties or in the same counties born before the campaign. The occupational income score effect implies roughly a 20-25 percent earnings premium from being raised in a county that experienced a large reduction in hookworm burden.

Table 1: Summary of main findings in Bleakley [2007]

5.3 Effect Sizes in Context

These are large effects. To put them in perspective, a 20-25 percent earnings premium from a single public health campaign compares favorably with estimates of returns to a year of schooling (roughly 8-10 percent), suggesting that the main mechanism is not just additional years of schooling but improved cognitive capacity, physical health, and learning productivity conditional on enrollment. Bleakley [2007] shows that most of the earnings effect operates through improved literacy and schooling quality, not just quantity.

6 Mechanisms

What explains these effects? Bleakley [2007] proposes several channels:

• Cognitive capacity. Chronic hookworm infection causes iron-deficiency anaemia, which is associated with impaired cognitive development in children. Eradication would directly improve the cognitive function of children in school, making them better learners.

• School attendance. Hookworm infection causes fatigue and weakness. Healthier children attend school more regularly and are more productive when they attend.

• Returns to schooling. If hookworm lowered the productivity of schooling, then eradicating it could raise the returns to a given year of education, providing an amplification mechanism.

• Parental income. Improved adult health increases parental productivity and income, providing more resources for children's education—a second-generation effect beyond the direct impact on infected children.

Bleakley [2007] presents evidence consistent with all four mechanisms but cannot fully decompose them. Identifying the precise mechanism remains a limitation of the design.

7 Comparison and Extensions

The Bleakley study has spawned a rich follow-up literature on disease eradication and development. Bhalotra and Venkataramani [2015] and others apply similar DiD-in-exposure designs to malaria eradication. Baird et al. [2016] revisit the Kenyan deworming experiment and find long-run positive effects on earnings using a randomised design, broadly corroborating Bleakley [2007]. Miguel and Kremer [2004] also find that school-based deworming in Kenya significantly increased school participation.

A notable challenge in this literature has been the so-called "worm wars": Taylor et al. [2015] and the International Initiative for Impact Evaluation (3ie) disputed the findings of Miguel and Kremer [2004], sparking a methodological debate. Bleakley [2007] uses a different design (historical DiD rather than an RCT) and draws on a stronger institutional context, making it somewhat insulated from the specific critiques of the Kenyan trial.

8 Limitations

Several limitations deserve acknowledgment:

• Pre-campaign infection measurement. RSC surveys were not conducted in every county, and the sample of counties surveyed may not be representative. Measurement error in the hookworm variable would attenuate the estimates.

• Spillover effects. If eradication in high-burden counties also reduced transmission to neighbouring counties, the control group is also partially treated, and the DiD estimate understates the true effect.

• External validity. The results apply specifically to the US South in the early twentieth century—a low-income, largely rural setting. Extrapolation to modern developing-country settings requires caution.

• Long-horizon data quality. Matching individuals across 1910, 1920, and 1940 Censuses relies on name-matching algorithms with imperfect coverage, potentially introducing selection bias. if the matched sample differs systematically from the full population.

9 What We Learn

Bleakley [2007] is a template for using historical natural experiments to identify long-run causal effects of health interventions. The key design elements—a large-scale exogenous health shock with geographic heterogeneity in exposure, combined with cohort variation—allow credible DiD identification even without randomisation.The study provides compelling evidence that early-life disease burden has lasting effects on human capital and income. with implications for both historical interpretations of the North-South income gapin the US and for modern development policy.

Methodologically, the paper shows the power of combining institutional knowledge (soil suitability determines hookworm burden; the RSC campaign was an exogenous health shock)with careful pre-trend testing and placebo analysis to validate a difference-in-differences design in a challenging historical setting.

10 Conclusion

The Rockefeller Sanitary Commission's hookworm eradication campaign, conducted over a century ago, created a natural experiment that Bleakley [2007] transforms into rigorous causal evidence on disease and development. The evidence shows that hookworm eradication had large and lasting effects on school enrollment, literacy, and adult earnings—suggesting that disease burden is not just a correlate of underdevelopment but a genuine cause. The paper stands as a model of historical causal inference. and continues to inform both theeconomics of health and the design of public health interventions in developing countries.

References

1. Baird, S., Hicks, J. H., Kremer, M., and Miguel, E. (2016). Worms at work: Long-run impacts of a child health investment. Quarterly Journal of Economics, 131(4):1637-1680.
2. Bhalotra, S. and Venkataramani, A. (2015). Shadows of the captain of the men of death: Early life health interventions, human capital investments, and institutions. Journal of Human Resources, 50(2):445-497.
3. Bleakley, H. (2007). Disease and development: Evidence from hookworm eradication in the American South. Quarterly Journal of Economics, 122(1):73-117.
4. Callaway, B. and Sant'Anna, P. H. C. (2021). Difference-in-differences with multiple time periods. Journal of Econometrics, 225(2):200-230.
5. Imbens, G. W. and Rubin, D. B. (2015). Causal Inference for Statistics, Social, and Biomedical Sciences. Cambridge University Press, Cambridge.
6. Miguel, E. and Kremer, M. (2004). Worms: Identifying impacts on education and health in the presence of treatment externalities. Econometrica, 72(1):159-217.
7. Roth, J., Sant'Anna, P. H. C., Bilinski, A., and Poe, J. (2023). What's trending in difference-in-differences? A synthesis of the recent econometrics literature. Journal of Econometrics, 235(2):2218-2244.
8. Taylor-Robinson, D. C., Maayan, N., Donegan, S., Chaplin, M., and Garner, P. (2015). Public health deworming programmes for soil-transmitted helminths in children living in endemic areas. Cochrane Database of Systematic Reviews, 11.