Regression Results Gasoline lead coefficients are highly significant (p-value < .0000) in all of the crime and unwed pregnancy regressions, as is the Roe v. Wade variable in the unwed pregnancy regressions for ages 15-19. The overall unemployment rate and percentage of the population in high-crime age brackets (ages 15-24) was not significant in any of the crime regressions. The teen unemployment rate was significant in the regressions for rape, robbery, and for all violent crime. In the case of pregnancies under age 15 and unwed pregnancies for other age brackets, the best-fit lag (highest R2and lowest p-value for lead) is equal to the median and mode age of unwed pregnancies for teens ages 15-17, and just above the median and mode ages for other age brackets. The median and mode ages of arrests for violent crime vary over time, and the broad age distribution results in medians that are several years higher than modes, but all of the crime regressions show a best-fit lag within the range of median and mode ages for arrests. Although other social trends and government policies are often cited to explain the rise and fall in unwed pregnancy and crime rates over recent decades, the role of childhood lead exposure seems to be especially apparent in the best-fit lag structures for gasoline lead regressions. In the case of the unwed pregnancy regressions, the best-fit lag for each age bracket is consistent with changes in lead exposure in the first years of life. In the largest categories of violent crime, where the median and mode ages for aggravated assault tend to be several years older than
Page 42 the median and mode ages for robbery, the best-fit lag for assault is four years greater than the best-fit lag for robbery. The best-fit lags in Figures 3 through 11 clearly match the rise, peak, and decline in gasoline lead exposure, and most also reflect the temporary peak and plateau from the mid-1950s through the mid-1960s. The fit between these temporal patterns, with lags consistent with the known risks of lead exposure in the first years of life, provide striking visual support for the association between lead exposure and undesirable social behaviors. The regression analysis of the temporal relationship between murder rates back to 1900 and gasoline and white lead exposure rates back to 1876 suggests that lead exposure may have influenced crime rates throughout this century. Although the R2for this regression is not as high as for the gasoline lead regressions, the results still show that lead exposure rates are highly significant (p-value < .0000) in explaining murder rates from 1900 to 1998. The best-fit lag of 21 years, versus 18 years for the best-fit lag between murder rates and gasoline lead, is also consistent with the different lags between gasoline and paint lead and their greatest impact on blood lead levels. That is, gasoline lead consumption has been strongly associated with blood lead levels with a lag of just one to two months. Paint lead, by contrast, is likely to pose the more pervasive hazard several years after it is consumed in paint, when the older paint begins to flake and peel and create paint chip and lead dust hazards.