Maternal Smoking & the 1998 Master Settlement Agreement
Douglas E. Levy, Ph.D. Ellen R. Meara, Ph.D.
Dept. Health Care Policy, Harvard Medical School
Support Provided by AHRQ T32-HS00055 & NIA T32-AG00186
Price Responsiveness of Smoking
Elasticity of smoking participation
General population -0.25 Pregnant women
-0.5 (Evans & Ringel; Natality Files 1989-1992) -0.7 (Ringel & Evans; Natality Files 1989-1995) -1.0 (Colman et al.; PRAMS data 1993-1999)
The Master Settlement Agreement led to an immediate 22% (45¢) increase in cigarette prices Price increase from MSA could reduce
Smoking among pregnant women 10-20% Low birth weight 2-4%
DID IT?
Study Design
Examined four outcomes
Maternal smoking participation and intensity (# of cigarettes/day) LBW and birth weight in grams
Used interrupted time-series regressions to determine whether there were important changes in outcomes following the MSA Subgroup analyses by mothers’ age, race, education
Study Data
National Vital Statistics Natality Detail Files from 1996 to 2000 We exclude
States not reporting maternal smoking (CA, IN, SD, upstate NY) Non-singleton birth Mother <15 or >44 years old Incomplete data on outcomes and demographics
4% had missing data given first three criteria
Control variables:
calendar month, state, excise taxes, maternal demographics, pregnancy characteristics, prenatal care, heavy alcohol consumption
N = 9,638,863
Study Period
Infants conceived from January 1996 to February 2000
Adequately capture pre- and post-MSA trends without picking up later non-price effects of MSA
Exclude infants conceived from March 1998 through November 1998
Avoid pregnancies only partially affected by MSA
Summary Statistics
Outcome Smoking #Cigs/day (sd)
LBW
All
12.1% 1.3 (4.4)
5.9%
Smokers
100% 11.0 (7.4)
10.1%
NonSmokers
0% 0 (0)
5.3%
Birth weight, g (sd) 3352 (572) 3161 (575) 3378 (567)
Maternal Smokers Are…
More likely to be white Less likely to have attended college Younger Less likely to be married
Likely to have fewer and later prenatal care visits More likely to have had a prior preterm birth More likely to drink heavily during pregnancy
Regression Analyses
Yits = 0 + timet + MSAt + timet*MSAt + Montht + Demographicsi + us + eist Time is monthly count; captures secular trend MSA captures immediate changes beginning Nov. 1998 Time*MSA captures changes in trend following MSA Month captures seasonal variation in outcome
Predicted Smoking Prevalence
MSA, Nov. 1998
15%
10%
Absent the MSA
5%
With the MSA
0%
y Ma
v No
Ju
Ju
r-9 Ma 6
t-9 Oc 6
8 c-9 De
Ja n-0
8 n-9
9 l-9
-97
-97
0
Predicted Smoking Prevalence 15mos Post-MSA
Population All
Hispanic
w/o MSA 11.7%
3.2
w/MSA 11.4%
3.0
Rel. % -2.2%
-6.3
Pr(MSA= MSA*time = 0) p<0.0197
p=0.1064
15-19yo
17.8
16.3
-8.2
p<0.0001
Predicted LBW Prevalence 15mos Post-MSA
Population All
Hispanic
w/o MSA 6.3%
5.7
w/MSA 6.2%
5.5
Rel. % -1.8%
-4.8
Pr(MSA= MSA*time = 0) p=0.0402
p=0.3726
15-19yo
9.0
8.7
-3.5
p=0.1077
Why So Little Effect?
Perhaps remaining smokers are less likely to change habits in response to price hikes Perhaps earlier estimates based on changes in state excise taxes were confounded
Simultaneous changes in attitude towards smoking Simultaneous tobacco control legislation
Limitations
No control group Assumes trend is linear Maternal smoking is self-reported Major states were excluded from the analysis
20% of births during this period
Implications
Pros of price increases for tobacco control
Young women and their infants may benefit Price changes affect most smokers Less effective than commonly believed Regressive taxation Results in other populations? Relate to new evidence re: long-run price-responsiveness
Cons of price increases for tobacco control
Next steps