Abstract

The following abstract was generated by the evaluation manager, with the help of a chatbot. Positives: (1) Novel epidemiological method for imputing historical infection rates. (2) Interesting comparison of epidemiological vs economic approaches. Limitations/suggestions: (3) Conclusion favoring the economic approach needs stronger support. (4) Concerned about identifying assumptions for ‘sharp cohort design’ — would like to see tests for ‘no cohort pre-trends’. (5) Needs more discussion of conceptual differences between mortality and infection rates. (6) Could elaborate more on applicability to other diseases.

Summary Measures

Managers’ note: The first evaluator’s metrics are omitted as the ratings given did not seem to reflect the content of the report.

Written report1

Using insights from epidemiological modeling, this paper develops a novel method to impute infection rates by birth cohort, which is subsequently used to test for long-run socioeconomic effects of being infected by measles as a child. This solves the problem that age-specific infection rates are generally not available, in particular not in the past. The authors exploit the fact that in a world with no immunization (and where measles is endemic) almost everyone will eventually be infected by measles during childhood. The invention of an effective immunization method (vaccination) then generates plausible exogenous variation in the share of a birth cohort infected by measles. The authors apply their approach to study the long-run effects of measles vaccination in the US and find insignificant effects on different socioeconomic outcomes when controlling for cohort effects. In other words, they cannot separate the effect of being infected with measles as a child from other cohort-specific trends.

On the other hand, if the authors instead use pre-vaccination mortality rates (age 0-5), share ever infected, or a combination, they find positive socioeconomic effects of immunization against measles, even when controlling for cohort fixed effects. The latter approach is often used in “the economics literature” studying the long-run implications of various disease eradication campaigns. One particularly interesting benchmark here is the interesting paper by Atwood (2022)[1], who studies the exact same campaign, also using US data. She concludes that measles vaccination had positive effects on income (among other variables) and demonstrates that this conclusion is robust to controlling for cohort fixed effects, along with a battery of other robustness checks. This is consistent with the current study when the campaign is evaluated using the “econ-approach”, but not when applying the “epi-approach”. The authors of the current study conclude that their evidence suggests that the econ-approach is better suited for isolating long-run effects of a disease intervention campaign compared to the epi-approach.

While I personally like this conclusion, the main question is whether it is warranted. I believe that the authors could do a bit more to support their interpretation. The main concern with the econ-approach is whether this captures other cohort trends unrelated to the disease campaign program (here vaccination against measles), since this strategy uses a cross-sectional variable to measure how much cohorts across the country could benefit from the campaign (i.e., intensity of treatment). In particular, this treatment measure captures the experience of multiple cohorts prior to the campaign. Thus, one could question whether this measure is well-suited for a sharp cohort design. As far as I can tell (and I might be wrong), Atwood (2022)[1] does not report any event studies for the long-run outcomes, where ‘event time’ is birth cohort. (She does provide a number of zero-stage-like event studies where contemporary mortality rates are the outcomes). Essentially, I would like to see that cohorts too old to benefit from the vaccination campaign did not experience improved long-run outcomes (i.e., no cohort pre-trends) and treatment dynamics as one would expect. If the authors would provide such evidence, I believe that their conclusion would be more robust, and very important for all studies applying the econ-approach (and this literature seems to be growing over time).

Below I provide a few additional (mostly minor) comments:

1) Beyond the issue of measurement (in particular cohort vs. period measures), the authors could elaborate a bit more on the conceptual difference between mortality rates and infection rates. These two rates are connected via the fatality rate. Is there any a priori reason as to why the fatality rate should vary across locations for a disease such as measles?

2) A study such as Acemoglu and Johnson (2007)[2], which is cited for using the reduced-form econapproach, does not apply this method in a sharp cross-cohort design. They compare countries over time, which means by construction they want to measure how disease reductions affected multiple cohorts (i.e., as such the entire population). Jayachandran et al. (2010)[3] does not apply this design at all (if I remember correctly). They instead compare diseases that could be treated with sulfa drugs to disease that could not treated with sulfa drugs to tell us something about how many life years this medical innovation saved. The authors should perhaps try to group (the relevant) studies into whether they apply the method in cohort designs and into studies that look at the effects on populations.

3) Although I think amounts to using the same sample variation, it is perhaps more intuitive controlling for survey-year fixed effects instead of age fixed effects.

4) I am not completely sure why the authors refer to Eq. (8) as a Bartik-type regression. Please elaborate. 5) The authors could discuss more whether their epidemiological approach can applied to other infectious diseases.

References

[1]Atwood, A. (2022). The Long-Term Effects of Measles Vaccination on Earnings and Employment. American Economic Journal: Economic Policy, 14(2), 34–60.

[2]Acemoglu, D., & Johnson, S. (2007). Disease and Development: The Effect of Life Expectancy on Economic Growth. Journal of Political Economy, 115(6), 925–985.

[3]Jayachandran, S., Lleras-Muney, A., & Smith, K. V. (2010). Modern Medicine and the Twentieth Century Decline in Mortality: Evidence on the Impact of Sulfa Drugs. American Economic Journal: Applied Economics, 2(2), 118–146.