These policies may be published as Health Education standards or Public Education codes . These policies can also provide information on how existing sex education laws may be interpreted by local school boards.
We analyzed the state profiles on sex education laws and policy data for all 50 states  following the criteria of the Editorial Projects in Education Research Center  to identify the level of abstinence education Table 2. North Dakota and Wyoming. Analyses of the two data sets gave essentially identical results. In this paper we present the analyses of the more extensive 48 states law and policy data set.
Teen pregnancy, abortion and birth data Data on teen pregnancy, birth and abortion rates were retrieved for the 48 states from the most recent national reports, which cover data through  , . The data are reported as number of teen pregnancies, teen births or teen abortions per one thousand female teens between 15 and 19 years of age.
In general, teen pregnancy rates are calculated based on reported teen birth and abortion rates, along with an estimated miscarriage rate . We used these data to determine whether there is a significant correlation between level of prescribed abstinence education and teen pregnancy and birth rates across states. The expectation is that higher levels of abstinence education will be correlated with higher levels of abstinence behavior and thus lower levels of teen pregnancy.
Other factors Data on four possibly confounding factors were included in our analyses. Socio-economics To account for cost-of-living differences across the US, we used the adjusted median household income for for each state from the Council for Community and Economic Research: These data are based on median household income from the Current Population Survey for from the U.
Ethnic composition We determined the proportion of the three major ethnic groups white, black, Hispanic in the teen population 15—19 years old for each state  , and assessed whether the teen pregnancy, abortion and birth rates across states were correlated with the ethnic composition of the teen population. To account for the ethnic diversity among the teen populations in the different states in a multivariate analysis of teen pregnancy and birth rates, we included only the proportion of white and black teens in the state populations as covariates, because the Hispanic teen population numbers were not normally distributed see below.
Medicaid waivers for family planning Medicaid-funded access to contraceptives and family planning services has been shown to decrease the incidence of unplanned pregnancies, especially among low-income women and teens . According to the Guttmacher Institute, the national family planning program prevents 1. Since the increasing role of Medicaid in funding family planning was mainly due to the efforts of 21 states to expand eligibility for family planning for low-income women who otherwise would not qualify for Medicaid, we analyzed whether these Medicaid waivers for family planning services available in some states but not in others could bias our results.
We determined which states had received permission as of from the Federal Medicaid program to extend Medicaid eligibility for family planning services to large numbers of individuals whose incomes are above the state-set levels for Medicaid enrollment .
We assessed whether the waivers access to family planning services had an effect on our analysis of teen pregnancy and birth rates across states, specifically whether they could bias our analysis with respect to the effects of the different levels of abstinence education. Except for teen abortion rates and Hispanic teen population data, all variables were normally distributed.
The distribution of the Hispanic teen population across states was not normal: Teen pregnancy and birth rate distributions included outliers, but these outliers did not cause the distributions within abstinence education levels to differ significantly from normal, thus all outliers were included in subsequent analyses.
For all further statistical analyses we used SPSS . Correlations We used non-parametric Spearman correlations to assess relationships between variables, and for normally distributed variables we also used parametric Pearson correlations, but these results showed the same trends and significance levels as the non-parametric correlations.
As a result, we only report the results for the non-parametric correlations here. We tested for homogeneity of error variances Levene's Test and for equality of covariance matrices Box test between groups. For pairwise comparison between abstinence levels, we used the Bonferroni adjustment for multiple comparisons. Results Among the 48 states in this analysis all U. To the contrary, teens in states that prescribe more abstinence education are actually more likely to become pregnant Figure 2.