This article is an open-access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).

The minimum legal drinking age (MLDA) was increased in the U.S. in the late 1980s in an effort to reduce intoxication-associated injuries, especially those related to motor vehicle accidents. This paper explores distal (secondary) effects of changing MLDA on indices of infant health, and whether changes in drinking behaviors or birth composition contributed to these effects.

State- and year-fixed-effects models are used to analyze the relationship between MLDA, drinking behaviors, and birth outcomes. We studied the effects of different MLDA (age 18, 19, 20, or 21 years) when potential mothers were 14 years old by merging two population-based datasets, the Natality Detailed Files and the Behavioral Risk Factor Surveillance System between 1985 and 2002.

A MLDA of 18 years old (when potential mothers were 14 years old) increased the prevalence of low birth weight, low Apgar scores, and premature births. Effects were stronger among children born to black women compared with white women. Moreover, a younger MLDA was associated with an increasing proportion of very young and high school dropouts for black women. Furthermore, older MLDA laws at age 14 years decreased the prevalence of binge drinking among black women.

Increasing the MLDA had longer term, distal impacts beyond the initially intended outcomes, specifically on birth outcomes (particularly among infants born to black women) as well as school drop-outs and binge drinking patterns among black young females. The older MLDA, intended initially to reduce problematic drinking behaviors, appeared to alter broader social contexts that influenced young women during their early childbearing years.

Broadly applied changes in social or public health policies may yield unintended consequences, in addition to initially targeted changes [

In the mid-1970s and early 1980s, organizations such as Mothers Against Drunk Driving argued to raise the MLDA after seeing increased MVA-related deaths associated with lowering the MLDA to 18 years old [

Of particular relevance to our report, a recent publication noted that a MLDA of 18 years at the time of mothers’ conception year was associated with more adverse outcomes among births [

Studies already have found that alcohol consumption during pregnancy is associated with adverse consequences for the fetus [

Researchers have shown that social settings will influence individual behaviors. In an environment where alcohol control is lenient, adolescents tend to begin drinking relatively earlier [

Overall, this study examined: (1) the relationship between alcohol-control environment during young women’s early childbearing years, indicated by the MLDA laws at their 14th birthday, and infant health indicators; (2) whether changes in birth composition, drinking behaviors, or formation of drinking habits contributed to change in infant health; and (3) whether racial differences relative to the above issues existed.

The study employs three data sources to understand the relationship between alcohol-controlled environment and infant health. The first one is the data extracted from the Vital Statistics Natality records from 1970 to 1992, census data that include all birth certificates, to estimate the relationship between MLDA laws at women’s age 14 and birth outcomes that occurred from ages 15–21. These data included infant information regarding birth weight, the 5-minute Apgar score and gestation length, and also mothers’ characteristics. The APGAR score is determined by evaluating the newborn baby on five criteria (Appearance, Pulse, Grimace, Activity, Respiration) on a scale from zero to two, then summing up the five values thus obtained. The resulting APGAR scores range from zero to 10. Low APGAR scores refer to scores below 7. We restricted the sample to young mothers who were 21 years old or younger at the time of delivery. In addition, since the MDA changed between 1970 and 1989, we further restricted the sample to women whose 14-years of age was between 1970 and 1989. Babies whose health indicators at birth were missing were dropped from the sample; if only one or two indicators were missing, the record was still kept in the sample. In the estimation, data were aggregated into cells according to residence, race/ethnicity, and year of mothers being aged 14 years. Panel A in

Despite of the Natality database’s rich information regarding infants’ health and their mothers’ characteristics, with its substantial underreporting, it is not a reliable source of data regarding drinking behavior.

To circumvent these issues, we used data from another nationally representative survey to measure alcohol drinking among women, in keeping with the methods of other reports and researchers [

Panel B of

The data on MLDA laws come from the Distilled Spirits Council of the U.S. (DISCUS). We use the minimum drinking age at the estimated year of age 14 years as the indicator for the relevant policy regime. In some years, laws have split among different type of alcohol (beer, wine or liquor): beer was legal at age 18, but liquor was legal at age 21. In those instances, we coded the MLDA as age 18 years.

The following equation estimates the long-run effects of MLDA on infant health and drinking behaviors:

where _{st}_{s,age–14}_{s}_{t}_{st}

^{2} = 1,160, df = 46, p < 0.001). This impact was largely associated with the difference between a MLDA of 18 and one of age 19 and age 21. The MLDA-19 decreased the LBW rate by 0.16 percentage points, while the MLDA-21 decreased the rate by 0.24 percentage points (LR ^{2} = 11,605, df = 48, p < 0.001). Results were split by race in Panels B and C. The MLDA polices, however, predicted no change in LBW of children born to white mothers, while they were significantly predictive for children born to black women. For example, when black women faced a MLDA-18 at the age of 14, the probability of their children weighing less than 2,500 g at birth increased by 1.8 percentage points (LR ^{2} = 7,605, df = 43, p < 0.001), and moving to an MLDA of 21 made the largest contribution to these effects by reducing the LBW 2.8 percentage points (LR ^{2} = 7,611, df = 45, p < 0.001).

The relationship between the MLDA at women’s age of 14 and low Apgar scores (below 7) is shown in Columns 3 and 4. On average, living in a state with a more lenient drinking age policy (^{2} = 15,605, df = 56, p < 0.001) when they became mothers at an older age when compared to other women. Exposure to any older age policy, ^{2} = 15,108, df = 58, p < 0.001). Racial differences were still observed with this birth health indicator. Children born to white mothers who at the age of 14 lived in a state that adopted a MLDA-18 were 0.36% points (LR ^{2} = 2,137, df = 46, p < 0.001) than those whose mothers lived in a state with older MLDA laws. Such effects were around 10 times larger for black women: children born to black mothers were 3.7% points (^{2} = 10,604, df = 48, p < 0.001) more likely to have low Apgar scores than their relative comparable groups. Age-21 laws were a significant driver to decrease the rate of low Apgar scores than other two age categories in either racial group. Along with the statistical significance, these results reveal a strong relationship between the drinking age laws at women’s young teenage years and probability of bearing a child with LBW or with low Apgar scores. Last, a MLDA of 18 predicted a 0.02 percentage points (LR ^{2} = 10,304, df = 46, p = < 0.001) increase in the prevalence of premature births of women at their later life. These age policies, however, had no observed effects among white mothers but among black mothers. Children born to black mothers who at the age of 14 experienced an MLDA-18 were 0.6 percentage points (LR ^{2} = 64,532, df = 43, p < 0.001) higher in being delivered pre-maturely than other children. States with a MLDA-21 predicted the largest decrease, roughly one percentage point (LR ^{2} = 6,532, df = 45, p < 0.001) in premature birth for black mothers. In summary, results in

How could the changes in MLDA laws affect birth outcomes? ^{2} = 64,695, df = 45, p < 0.001). Moreover, lenient alcohol availability, as reflected in the lower MDLA, increased the proportion of high school dropouts by 0.5 percent [F (72, 29927) = 1,286]. This effect was mainly associated with MLDA-21 laws, under which the proportion of mothers with less than a high school education dropped by 1.1 percentage point [F (75, 29914) = 1,365]. The age-specific legal access to alcohol had little impact on the composition of white mothers. While compositional shift is one possible explanation for the change in birth outcomes, results suggest that this shift only contributes to a very modest degree of improvement in birth outcomes for black mothers. No such effects are found among white mothers.

Using the individual-level data from BRFSS, ^{2} = 3,163, df = 68), and age-19 and age-21 partly explained these changes.

The stepwise change in MLDA laws had little effect on drinking among white women as the MLDA of age 21 years decreased the probability of alcohol drinking up to one percentage point, and had no significant effects on reducing binge drinking. In contrast, establishing the MLDA at age 21 years decreased the probability of alcohol drinking among black women by 2–8% points (LR ^{2} = 3,169, df = 70). Moreover, MLDA differences during early teenage years predicted differences in binge drinking between young white and black women, with the younger MDLA associated with more binging among blacks. Those who lived in a state where the MLDA was age-18 were 2.2 percentage points (LR ^{2} = 1,148, df = 63; p < 0.001) more likely to binge than those who lived in states with an older MLDA. Exposure to age-20 or -21 MLDA reduced the binge drinking by about 6.6 percentage points (LR ^{2} = 12,052, df = 65).

These results indicate that the MLDA changed drinking behaviors among white and black women in a different way. For black women, a movement away from age-18 as the drinking age decreased overall drinking probability to a relatively modest degree and decreased binge drinking substantially more. Apparently, an increase in the legal drinking age modified the formation of habitual drinking by effectively reducing the availability of alcoholic beverages. For white women, an older MLDA modestly reduced overall drinking, but not binge drinking. This suggested that it was moderate white drinkers rather than heavy ones who were sensitive to changes in MLDA.

Our results showed that an older MLDA when girls turned age 14, which we postulated was related to the “alcohol environment” in which they spent most of their teen years, was related to improved infant health, particularly among those born to black women. These were associated with a lower incidence of LBW, low Apgar scores and premature births and hence, indicated that a stricter MLDA was related to improved infant health on average. This result is consistent with other studies since a stricter alcohol access policy is associated with lower risk for risky sexual behaviors, hence lower risk for unplanned births. [

Future work is needed to clarify how and why MLDA laws had different impacts among black and white young women. In addition, this study reflects a lack of data that simultaneously include reliable information regarding an infant’s health and his/her mother’s drinking behaviors. If such ideal data had existed, we could find whether the MLDA has a causal impact on birth outcomes through drinking behaviors, using two-stage least square models. This current study is unable to draw out this causality claim. Moreover, not all policies, such as expansion of Medicaid for pregnant women, were accounted for in this study. In addition, some other factors like insurance status, utilization and frequency of pre-natal care may influence infant health, but were not included in the analysis. Some of these variables are not available in the Natality data. Hence, the omitted variable bias may still exist. Despite these limitations, this study suggests that measures to improve the social context (

The authors thank John Cawley, Xiangqiang Fang, Curtis Florence, George Jakubson, Donald Kenkel, Jordan Matsudaira, Kosali Simon and Douglas Young for helpful suggestions and discussion, and Thomas Dee for sharing his minimum legal drinking age laws data. Thanks also go to Ann Russ for her editorial assistance and proofreading.

The authors declare no conflict of interest.

Summary statistics.

All | White | Black | ||||
---|---|---|---|---|---|---|

| ||||||

Average Percentage | SD. Dev. | Average Percentage | SD. Dev. | Average Percentage | SD. Dev. | |

white moms | 51.05 | 49.01 | ||||

% born below 2500 g | 10.68 | 7.54 | 7.96 | 2.534 | 13.54 | 9.70 |

% born below 1500 g | 2.07 | 3.87 | 1.456 | 1.52 | 2.70 | 5.34 |

% Apgar score below 7 | 2.96 | 6.14 | 2.44 | 4.70 | 3.51 | 7.32 |

% pre-maturity birth | 14.87 | 9 | 11.19 | 3.76 | 18.77 | 11.06 |

% mothers married | 23.93 | 29.07 | 35.21 | 31.95 | 12.10 | 19.68 |

% mothers high school dropouts | 59.30 | 31.55 | 60.66 | 30.58 | 57.85 | 21.48 |

% mothers high school grads | 33.76 | 25.34 | 33.26 | 24.55 | 34.29 | 26.13 |

% mothers with some college | 6.72 | 9.47 | 5.85 | 7.41 | 7.59 | 11.17 |

% mothers smoke during pregnancy | 34.53 | 10.55 | 39.88 | 14.19 | 27.77 | 11.27 |

% mothers drank during pregnancy | 1.77 | 4.54 | 1.92 | 5.44 | 1.4564 | |

N | 11,051 | 7,448 | 4,603 | |||

% drink | 51.11 | 50.00 | 52.65 | 50.00 | 45.72 | 50.00 |

# drink per time | 2.71 | 2.31 | 2.73 | 2.33 | 2.67 | 2.21 |

% binge drink among drinkers | 50.97 | 50.00 | 21.12 | 50.01 | 25.63 | 45.68 |

% high school dropouts | 6.91 | 25.36 | 6.45 | 24.57 | 10.65 | 30.7 |

% high school graduate | 31.21 | 46.33 | 30.39 | 46 | 37.8 | 48.49 |

% some college | 30.15 | 45.89 | 30.06 | 45.84 | 30.91 | 46.21 |

% college or more | 31.74 | 46.54 | 33.14 | 47.12 | 20.72 | 40.52 |

N | 284,382 | 245,718 | 38,619 |

The effects of MLDA on birth outcomes.

Low Birth Weight | Low Apgar Scores | Premature Birth | ||||
---|---|---|---|---|---|---|

1 | 2 | 3 | 4 | 5 | 6 | |

MLDA18 | 0.0014 |
0.0112 |
0.0002 |
|||

p-value | <0.0001 | <0.0001 | 0.0514 | |||

MLDA19 | −0.0016 |
−0.0180 |
−0.0002 | |||

p-value | 0.0020 | <0.0001 | 0.0511 | |||

MLDA20 | −0.0005 | −0.0103 |
0.0001 | |||

p-value | 0.2174 | <0.0001 | 0.2145 | |||

MLDA21 | −0.0024 |
−0.0182 |
−0.0004 | |||

p-value | <0.0001 | <0.0001 | 0.0002 | |||

N | 11,051 | 11,051 | 5,792 | 5,792 | 9,900 | 9,900 |

MLDA18 | 0.0000 | 0.0036 |
0.0000 | |||

p-value | >0.1 | <0.0001 | >0.1 | |||

MLDA19 | 0.0000 | −0.0050 |
0.0000 | |||

p-value | >0.1 | 0.0003 | >0.1 | |||

MLDA20 | 0.0000 | −0.0040 |
0.0000 | |||

p-value | >0.1 | 0.0002 | >0.1 | |||

MLDA21 | 0.0000 | −0.0061 |
0.0000 | |||

p-value | >0.1 | <0.0001 | >0.1 | |||

N | 7,448 | 7,448 | 3,471 | 3,471 | 5,677 | 5,677 |

MLDA18 | 0.0181 |
0.0372 |
0.0061 |
|||

p-value | <0.0001 | <0.0001 | ||||

MLDA19 | −0.0202 | −0.0257 |
−0.0056 | |||

p-value | 0.0071 | <0.0001 | 0.2102 | |||

MLDA20 | −0.0073 | −0.0368 |
−0.0027 | |||

p-value | 0.2671 | <0.0001 | 0.5029 | |||

MLDA21 | −0.0283 |
−0.0498 |
−0.0095 | |||

p-value | <0.0001 | <0.0001 | 0.0075 | |||

N | 4,603 | 4,603 | 2,321 | 2,321 | 4,223 | 4,223 |

Notes: 1. Marginal effects are presented in the Table. 2. The dataset is the aggregated Natality files 1970–1992. Sample is restricted to mothers younger than 21 years old. Models include controls such as mothers’ education, age, marital status, smoking behaviors during pregnancy, real income per capita and real beer taxes (federal plus state level). 3. The numbers in the column headings represent the type of models we refer to. For instance, Column 1 refers to the model that includes MLDA-18 as well as other co-variates. All models include state and year fixed effects. Marginal effects are reported and standard errors in parentheses according to probit estimation. Sample is restricted to mothers younger than 21 years old. 4.

Statistically significant with a p-value <0.10,

Statistically significant with a p-value <0.05,

Statistically significant with a p-value <0.01.

The effects of MLDA on compositional change of births.

Mother Age 15–17 | % High School Dropouts | |||
---|---|---|---|---|

1 | 2 | 3 | 4 | |

MLDA18 | 0.0069 |
0.0004 | ||

p-value | 0.1682 | 0.8501 | ||

MLDA19 | −0.0028 | 0.0016 | ||

p-value | 0.6404 | 0.5252 | ||

MLDA20 | −0.0253 | 0.0005 | ||

p-value | 0.0046 | 0.8892 | ||

MLDA21 | −0.0053 | −0.0028 | ||

p-value | 0.9286 | 0.2477 | ||

N | 11,051 | |||

MLDA18 | −0.0015 | −0.0049 | ||

p-value | 0.8191 | 0.0471 | ||

MLDA19 | 0.0078 | 0.0048 | ||

p-value | 0.3128 | 0.0937 | ||

MLDA20 | −0.0102 | 0.0057 | ||

p-value | 0.3519 | 0.1606 | ||

MLDA21 | −0.0012 | 0.0048 | ||

p-value | 0.8736 | 0.0937 | ||

N | 7,448 | |||

MLDA18 | 0.0160 |
0.0056 |
||

p-value | 0.0336 | 0.0958 | ||

MLDA19 | −0.0139 | −0.0014 | ||

p-value | 0.1161 | 0.6405 | ||

MLDA20 | −0.0047 | −0.0042 | ||

p-value | 0.713 | 0.4629 | ||

MLDA21 | −0.0093 | −0.0106 | ||

p-value | 0.2856 | 0.0108 | ||

N | 4,603 | 4,603 |

Statistically significant with a p-value <0.10,

Statistically significant with a p-value <0.05,

Statistically significant with a p-value <0.01.

The MLDA and the drinking behaviors.

Drinking (0/1) | Binge Drinking (0/1) | |||
---|---|---|---|---|

1 | 2 | 3 | 4 | |

MLDA18 | 0.0015 | −0.0095 | ||

p-value | 0.6048 | 0.0102 | ||

MLDA19 | −0.0057 |
−0.0003 | ||

p-value | 0.0880 | 0.9364 | ||

MLDA20 | −0.0042 | 0.0404 | ||

p-value | 0.4206 | 0.6541 | ||

MLDA21 | −00056 |
0.0149 | ||

p-value | 0.1543 | 0.0039 | ||

N | 284,382 | 154,080 | ||

MLDA18 | 0 | −0.007 | ||

p-value | >0.1 | 0.3846 | ||

MLDA19 | −0.0046 | −0.0021 | ||

p-value | 0.1945 | 0.6486 | ||

MLDA20 | −0.0057 | 0.056 | ||

p-value | 0.2928 | 0.4037 | ||

MLDA21 | −0.0085 |
0.0257 | ||

p-value | 0.0421 | 0.0974 | ||

N | 245,718 | 140,328 | ||

MLDA18 | 0.0015 | 0.0219 |
||

p-value | 0.8421 | 0.0200 | ||

MLDA19 | −0.0282 | −0.0046 | ||

p-value | 0.0482 | 0.9603 | ||

MLDA20 | −0.0823 |
−0.0659 | ||

p-value | 0.0006 | <0.0001 | ||

MLDA21 | −0.0234 |
−0.0656 | ||

p-value | 0.1256 | <0.0001 | ||

N | 38,619 | 13,752 |

Statistically significant with a p-value <0.10,

Statistically significant with a p-value <0.05,

Statistically significant with a p-value <0.01.