Supplementary Electronic File
I.
Control variables and summary statistics
Table A1
Definition of Control Variables
Variable
Coding
Age
=Child’s age
Number of children
categories
2 children
=1 if the number of children below age of 18 in the household is 2.
3 or more children
=1 if the number of children below age of 18 in the household is 3 or more.
Mother’s age categories
Mother below 34
=1 if mother’s age is below 34
Mother below 44
=1 if mother’s age is below 44 and above 34
and above 34
Race/Ethnicity
Hispanic
=1 if the child is Hispanic
Black
=1 if the child is Black
Income
=1 if the household annual income is above 20,000$
Mother’s marital status
=1 if mother is currently married
Mother’s education
=1 if mother has 13 years of education or more
People in the household
=Number of people in the household
Health insurance status
Private
=1 if person holds employer or union provided health insurance.
Public
=1 if person Medicaid, S-CHIP, or American-Indian health insurance.
Military
=1 if person holds TRICARE, CHAMPUS, or CHAMPUS-VA.
=1 if there is any period of the time after age 11 that teen did not have any
No Insurance Since 11
health insurance coverage
No doctor visit last year
=1 if person has not visited a doctor in the past 12 months
=1 if teen has already lung condition rather than asthma, heart condition,
diabetes, a kidney condition, sick cell anemia or other anemia, weakened
Teen health status
immune system because of chronic illness or caused by medicine taken by
chronic illness
=1 if any other members of teen’s household have lung condition rather
than asthma, heart condition, diabetes, a kidney condition, sick cell anemia
Household health status
or other anemia, weakened immune system because of chronic illness or
caused by medicine taken by chronic illness
Tdap uptake status
=1 if teen has had at least one shot of Tdap since age 10 years
Table A2
Summary Statistics
Physician
recommendation
0.537
(0.499)
Education for parents
state
0.519
(0.500)
Age
14.972
(1.387)
14.997
(1.393)
15.004
(1.394)
0.498
(0.500)
0.069
(0.253)
0.510
(0.500)
0.096
(0.295)
0.502
(0.500)
0.095
(0.293)
0.074
(0.262)
0.354
(0.478)
0.0794
(0.270)
0.430
(0.495)
0.0759
(0.264)
0.429
(0.495)
0.066
(0.248)
0.378
(0.485)
0.226
(0.418)
0.135
(0.341)
0.108
(0.311)
0.096
(0.295)
0.857
(0.350)
0.846
(0.361)
0.883
(0.321)
0.545
(0.498)
0.718
(0.450)
0.628
(0.483)
0.686
(0.464)
0.620
(0.485)
0.705
(0.456)
4.057
(1.305)
4.285
(1.262)
4.251
(1.274)
0.299
(0.458)
0.073
(0.260)
0.045
(0.208)
0.115
(0.319)
0.315
(0.464)
0.0471
(0.212)
0.083
(0.275)
0.160
(0.367)
0.289
(0.453)
0.054
(0.227)
0.078
(0.268)
0.155
(0.362)
0.039
(0.195)
0.335
(0.472)
0.037
(0.190)
0.340
(0.474)
0.037
(0.189)
0.343
(0.475)
0.618
(0.486)
0.626
(0.484)
0.607
(0.488)
School mandate states
2 children
3 or more children
Mother below 34
Mother below 44 and
above 34
Hispanic
Black
Income
Mother’s marital status
Mother’s education
People in the household
Public health insurance
Military health insurance
No Insurance Since 11
No doctor visit last year
Teen health status
Household health status
Tdap uptake status
1
Other state
0.539
(0.498)
II. Robustness of policy effects
We want to be sure that null effects of existing policies is a fair conclusion, so we engage in
several robustness checks in Table A3. First, the strong impact of physician recommendations
might be attenuating the impact of the policies if those recommendations are more likely in states
with mandates and educational programs than those in our control group. In panel A, we remove
the recommendation variable as a control and find that there is still no discernable effect of either
policy. In fact, the effects weaken. In panel B, we combine school mandates and education for
parents, considering it as one policy to promote HPV vaccination. In the case of DC, the policies
were concurrent so this is sensible. School entry requirements for HPV vaccines in Virginia are
also not strict, and they offer a very liberal opt out. As a result, one might expect that school
mandates would only affect the vaccine decision through the educational content they provide
anyway. Regardless, there is no impact in considering both policies together.
As a next attempt to assess policies, we note that using the entire set of states that failed
to enact a policy as control states is problematic if those states that focus on HPV are simply
more health conscious to begin with and perhaps more likely to promote HPV both before and
after official policy enactment. We test for this by limiting our control group to only those states
that at least introduced a state mandate that ultimately failed. It is these states that are likely
more comparable prior to the enactment of the vaccine. In panel C, we see that for at least
initiation of the vaccine, the results are more promising. However, it is only education for
parents that is a significant predictor of vaccine take-up and even this effect does not hold for
vaccine completion. Moreover, the effect is essentially zero for the 13 and under targeted group.
We cautiously note here that this provides some support for educational material as a policy to at
2
least initiate the vaccine sequence but again cannot see any meaningful support for school
mandates.
Only two jurisdictions had a school mandate. We focus on Virginia and more
systematically assess the impact of that state’s school mandate through a synthetic control
approach. The synthetic control method is a data-driven procedure that provides a single control
unit as a weighted average of characteristics of several potential comparison units. The weights
determine the relative contribution of each control unit to the counterfactual of interest. Despite
the many benefits, this method is designed for aggregate level panel data while we are using an
individual level repeated cross section database. However, the Center for Disease Control and
Prevention utilizes the same individual level database in order to estimate the HPV vaccination
rate among girls between 13-17 years old. We use the data on vaccination rate, along with
aggregate state level data on different characteristics to build our synthetic group and obtain the
weights. Then, we use these weights in the basic specification to estimate the effect of school
mandates.1 Panel D presents these results and again show no influence of the Virginia school
mandates.
Finally, only Missouri and DC enacted their educational programs for parents after 2008.
For that reason, we engage in a final test where we repeat the synthetic control approach to
identify the effect of educational programs for parents. We drop all the jurisdictions that either
1
The synthetic control is created by matching on the unemployment rate, median income, the relative
percentage of urban population, the relative percentage of Hispanic population, the relative percentage of
Black population, population density, the percentage of the female population above age 25 with a
bachelor degree or higher, the percentage of the under age 18 population with public health insurance
plans, and the percentage of married households. The weights yielded by this approach indicate that
vaccination rate trends in Virginia prior to enactment of school mandate is best reproduced by a
combination of Alaska (1.7%), Maryland (61.1%), Minnesota (7.7%), North Dakota (7.2%), and Utah
(21.8%).
3
enacted school mandates, or enacted their educational programs for parents before 2008.2 Panel
E presents these results. Education for parents is a significant predictor of vaccine initiation and
this effect does not hold for vaccine completion.
Table A3
Robustness Checks on policy impacts- Marginal effects from probit models
Vaccine Initiation
VARIABLES
(1)
Panel A-Robustness Check for Multi-collinearity
School mandates
0.0252
(0.0410)
Education for parents
0.0282
(0.0203)
Vaccine Completion
(2)
Uptake Before Age 13
(3)
0.0369
(0.0386)
0.0175
(0.0327)
0.0053
(0.0184)
-0.0196
(0.0203)
Observations
35,624
35,624
Panel B- Consider School Mandates and Educational Programs Together as HPV promotion
HPV vaccine promotion
Observations
0.0278
(0.0194)
0.0096
(0.0177)
-0.0109
(0.0196)
35,624
35,624
25,799
Panel C- Using As Control Groups Only Those States that Introduced a Mandate
School mandate
0.0595
0.0516
(0.0426)
(0.0405)
Education for parents
0.0907***
(0.0312)
Observations
17,758
Panel D- Synthetic control for School Mandate in Virginia
School mandate
0.0296
(0.0672)
Observations
3,562
Panel E- Synthetic control for education for parents in Missouri
Education for parents
25,799
0.0736*
(0.0442)
0.0244
(0.0347)
0.0463
(0.0294)
-0.0082
(0.0297)
17,758
13,060
0.0269
(0.0621)
-0.0225
(0.0402)
3,562
2,578
0.0325
(0.0383)
0.0249
(0.0421)
Observations
5,350
5,350
3,774
Note: Regressions include state and year fixed effects as well as the interactions of age and year. The numbers in
parenthesis are robust standard errors. *** p<0.01, ** p<0.05, * p<0.1
2 Synthetic of Missouri is the combination of Alabama (22.4%), Alaska (12.6%), Hawaii (15.1%),
Kentucky (7.8%), Montana (14.6%), Ohio (12.6%), South Carolina (3.2%), and Tennessee (11.7%).
4