Fighting Crime in the Cradle

Notes: Each cell represents a separate OLS regression with standard errors clustered at the birth county level in parentheses. Observations are at the birth county by birth month level and are weighted by the number of births in each county in 1964. The dependent variable is the fraction of individuals in a given birth county–birth month cohort who are later convicted of a crime or particular crime type in North Carolina by age 24. Columns indicate crime types (any, violent, property), and rows indicate severity (any conviction or felony). Mirroring FBI Part I definitions, violent crimes are defined only as offenses containing the words “murder,” “assault,” or “robbery” (rape is not included). Property crimes are defined only as offenses containing the words “burglary” or “larceny.” All specifications include birth county and birth month fixed effects, as well as baseline county characteristics interacted with a time trend in birth cohort. Baseline (1960) county characteristics include: percent of land in farming, percent of people living in families with less than $3,000, percent of population in urban area, percent Black, percent less than age five, percent greater than age 65, and percent of employment in agriculture. The sample is restricted to cohorts who were born between 1964 and 1974. Significance levels:

↵* p < 0.10;

↵** p < 0.05,

*** p < 0.01.

Notes: Each cell represents a separate OLS regression with each row denoting a different dependent variable and each column denoting a different specification. The dependent variable is the fraction of individuals in a given birth county–birth month cohort who are later convicted of a crime or particular crime type in North Carolina by age 24. All specifications include birth county and birth month fixed effects. Baseline (1960) birth county characteristics include: percent of land in farming, percent of people living in families with less than $3,000, percent of population in urban area, percent Black, percent less than age five, percent greater than age 65, and percent of employment in agriculture. “Additional birth county chars.” (also interacted with a trend in birth cohort) include population density, median income, median education, percent of adults with less than five years education, unemployment rate, per capita government expenditure, and Democratic vote margin in 1960 presidential campaign. Observations are at the birth county by birth month level and are weighted by the number of births in each county in the initial year of the sample period. The sample is restricted to cohorts who were bom 1964–1974 or 1968–1974 as noted. The latter sample enables the inclusion of time-varying county characteristic controls (birth county by birth month level) constructed from natality files. These “County natality chars.” include mean mother’s age, fraction of births to married parents, fraction white births, and fraction of births with an attending physician in a hospital. War on Poverty (WOP) controls include access to WIC (at birth) and Head Start (at age four), as well as per capita expenditures on Public Assistance Transfers, Medicaid, Community Health Centers, and Community Action Agencies. Standard errors clustered at the birth county level are in parentheses. Significance levels:

↵* p < 0.10;

↵** p < 0.05,

↵*** p < 0.01.

Notes: Each cell represents a separate OLS regression with standard errors clustered at the birth county level in parentheses. Observations are at the birth county by birth month level and are weighted by the number of births in each county in 1964. The dependent variable is the fraction of white or nonwhite individuals in a given birth county–birth month cohort who are later convicted of a crime or particular crime type in North Carolina by age 24. All specifications include birth county and birth month fixed effects, as well as baseline county characteristics (1960) interacted with a trend in birth month. Baseline (1960) birth county characteristics include: percent of land in farming, percent of people living in families with less than $3,000, percent of population in urban area, percent Black, percent less than age five, percent greater than age 65, and percent of employment in agriculture. The sample is restricted to cohorts who were born between 1964 and 1974 in the 75 counties where the number of births is available by race prior to 1968. Significance levels:

↵* p < 0.10;

↵** p < 0.05,

*** p < 0.01.

Notes: Each column presents coefficients from a separate OLS regression with standard errors clustered at the county level in parentheses. Observations are at the county by birth cohort by age level and are weighted by the number of births in each county in 1964. The dependent variable is the number of individuals per 100 within a given county cohort who are arrested at a particular age. All specifications include birth year, age, and county fixed effects, as well as baseline county characteristics (1960) interacted with a trend in birth year. Baseline county characteristics include: percent of land in farming, percent of people living in families with less than $3,000, percent of population in urban area, percent Black, percent less than age five, percent greater than age 65, and percent of employment in agriculture. The sample is restricted to individuals age 18–24 unless otherwise noted. Sample restricted to agencies accounting for at least 20 percent of a county’s population. Sample sizes vary due to differences in reporting across offenses. Significance levels:

↵* p < 0.10;

↵** p < 0.05,

↵*** p < 0.01.

Notes: Table shows back-of-the-envelope calculations of the discounted social benefits of later crime reduction from the 1964–1974 implementation of the FSP. Social cost estimates for each crime type (Column 1) are adopted from the preferred estimates of McCollister, French, and Fang (2010) and the lowest estimates from their literature review, both of which may be underestimates. The former estimates include victimization costs, criminal justice system costs, and the lost value of criminals’ time, but do not include private expenditures on crime prevention. The latter estimates include only victimization costs. The estimates of the change in arrests due to FSP implementation (Column 2) are based on the coefficient estimates from Equation 1 for each offense (contained in Table 5). The change in arrests is converted to a change in offenses (Column 3) using the ratio of offenses to arrests for each offense type. This ratio is operationalized conservatively as the minimum of the annual ratio of the UCR national estimate of offenses known to the UCR national estimate of arrests for 1980–2000 for the given offense. For murder/manslaughter this ratio is less than one due to either the UCR imputation process or a high rate of offenders per murder/manslaughter offense. This results in our estimated changes in murder/manslaughter arrests exceeding our estimated changes in murder/manslaughter offenses, potentially leading us to underestimate the social benefit from reductions in murder/manslaughter. Estimates of the discounted social benefit are produced by multiplying the dollar value of each offense’s social cost by the change in offenses implied by our estimates, discounted using various social discount rates. See Online Appendix C for details.