🚨Replication alert🚨 Do vaccinated children have higher income as...

@michael_wiebe
Michael Wiebe@michael_wiebe
40 views Jan 31, 2024 ~3 min read
Advertisement
1
🚨Replication alert🚨

Do vaccinated children have higher income as adults?

Atwood (2022) reports that childhood access to the measles vaccine increased adult income.
I run an event study and find trends that contradict a causal effect of the vaccine.

#EconTwitter
1/
Media image
2
Since the vaccine was rolled out nationally in 1963, there's no control group. To make a treatment vs control comparison, Atwood uses states with high and low pre-vaccine measles incidence.

aeaweb.org/articles?id=10…

2/
Media image
3
But this is *reported* incidence.

Measles is so contagious that basically everyone got infected in the pre-vaccine era. So there shouldn't be any differences across states in *actual* long-run incidence.

3/
Media image
4
There could be short-run differences, due to timing of the epidemic cycle. But these would average out over 1952-63, the period Atwood uses to calculate pre-vaccine measles incidence.

So why is there variation in reported incidence, if actual incidence is ~100% everywhere?

4/
5
I see two possible explanations:
(1) States differ in reporting capacity. Better health care, report a larger fraction of cases.
(2) States with worse initial health have more severe measles cases, and more severe cases are more likely to be reported.

5/
6
In (2), states with higher reported incidence benefit more from the vaccine because they have worse health.
But in (1), it's not clear why states with better reporting capacity would benefit more from the vaccine.

6/
7
Atwood doesn't discuss this, and instead seems to think measles is like malaria (see link), which has geographic variation in mosquito prevalence. But there's no source of long-run geographic variation for measles.

aeaweb.org/articles?id=10…

7/
8
Given the ambiguity in the treatment variation, we should be worried that Atwood's finding that the vaccine increased adult income and employment is capturing unrelated trends.

The paper doesn't run an event study to test for pre-trends, so I do it.

8/
9
The event study tests how the effect of the vaccine changes over time, by estimating one coefficient for each birthyear.

The treatment dose is access to the vaccine, which is 0 for cohorts born before 1949, increases by 1 per year until 1964, and is 16 afterwards.

9/
10
The idea is that everyone gets measles by age 16, so if you're born in 1949, you have one year of vaccine access (as a 15 year-old) before you get measles anyway.
Born 1950: 2 years access
...
Born 1964: 16 out of 16 years of access.

10/
11
So in the event study, we should observe a flat line before 1948, an increase over 1948-64, and a flat line afterwards.

Eg. people born after 1964 have the same vaccine access (full), so the effect should be the same from 1964-80.

11/
12
Here are the results.

We do see a flat line before 1948 and an increase over 1948-64. But the trend continues after 1964, even when there's no change in vaccine access. The 1964 and 1980 cohorts have the same access to the vaccine, so they should benefit the same.

12/
Media image
13
Hence, Atwood's results seem to be picking up other trends, instead of a treatment effect of the vaccine.

This doesn't mean that the vaccine failed. On the contrary:

But it does mean that this research design works better for diseases like malaria.

13/
@michael_wiebe
Michael Wiebe@michael_wiebe
The measles vaccine was a tremendous achievement.
Media image
14
If we want to study the long-term effects of the measles vaccine, we need a different approach.

See here for the writeup:
michaelwiebe.com/assets/atwood/…

I'll be submitting this to AEJ:EP as a comment. Atwood didn't respond to emails.

14/
15
I also have an extended commentary with results on:
- testing for a trend break in the event study
- the event study for disease outcomes
- heterogeneous effects, and more

michaelwiebe.com/assets/atwood/…

15/
Media image
16
If you want to see more replications, support me on Patreon:
patreon.com/michael_wiebe

If you have research you want to see scrutinized, please get in touch:
michaelwiebe.com/hire-me/

16/16
18
@michaelmina_lab
19
@SeriousStockman And it would have to be happening differentially in high measles states.
Actions
What You Can Do
  • Export as PDF or Markdown
  • Batch Export to Notion
  • Bookmark & Highlight
  • Screenshot Tweet
Create Free Account

Includes 7-day Premium trial

Advertisement