Arrested development: relative school entry age and arrests during the teenage and young adult years

ABSTRACT

A large literature documents that there are significant academic and non-academic differences between the youngest and oldest students in a school cohort. This paper investigates if being the youngest in a cohort has any impact on an individual's propensity to commit crime by utilizing a data set that contains over 4 million arrest records spanning a 20-year period in California. While I find no persistent effect on the probability of arrest, the youngest students in a cohort have a higher risk of arrest for certain offenses at age 14, corresponding to the age when they would transition to high school.

KEYWORDS:

• School
• juvenile
• delinquency
• crime
• policy

JEL CLASSIFICATION:

• H75
• I21
• J13
• K42

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data Availability Statement

I would like to request an exemption from the data availability requirement, as this paper uses confidential data from the California Department of Justice and the California Department of Public Health.

Notes

1 Status offenses are offenses for which only juveniles can be charged, which include truancy, loitering/breaking curfew, running away, and incorrigibility.

2 Each record contains only one offense; therefore, if someone is arrested for more than one crime only the most serious offense is recorded.

3 Based on U.S. Census Data from the 2005 American Community Survey, the states that send the most people to California based on migration trends tend to have an earlier school entry cutoff date–with a majority of those states having a school entry cutoff date of September 1st (Education Commission of the States 2011)). This could possibly create biased estimates due to some degree of measurement error stemming from those who may have been exposed to a different school entry cutoff date than that of California. The most likely outcome is that those not born in California would have been exposed to earlier school entry cutoff date–effectively moving them from the ‘before cutoff’ group to the ‘after cutoff’ group for those born in the fall. However, given the large number of observations in the data set, this is not likely to create significantly biased estimates. Further, the primary bandwidth of the RD analysis used throughout the paper utilizes a 60-day range on either side of the school entry cutoff that would ignore those who are too far away from the school entry cutoff.

4 During this time frame, on average 74% of the population of all other states aged 24 or younger lived in their state of birth (U.S. Census Bureau 2004).

5 The MACR database shows where a person is sent after arrest, but not what happens to them after that step. For juveniles, this indicates that they are sent to the probation department for further processing and for adults this means that a formal complaint was sought.

6 Figure A.1 in the Online Appendix show birth counts for each year of available data.

7 Further details on this calculation can be found in the Online Appendix.

8 The denominator of the arrest rate for each subgroup is calculated using only the data relevant to that subgroup and excludes all others.

9 The California Education Code regarding school entry age prior to 1987 read that individuals must be 4 years and 9 months old by September 1, which is equivalent to turning 5 years old by December 1.

10 Shigeoka (2015) does find evidence of manipulation of births around the cutoff in Japan, where certain parents delay the birth of their child to just after the cutoff. However, in Japan the school entry cutoff is binding and the option of delaying entry is not available as it is in the U.S.

11 During the last part of the sample period, California did allow for certain students to begin kindergarten midway through the school year as long as they turned 5 by May 1st in the Spring of that school year. These mid-year entry students would effectively be moved from the group born after the cutoff to the group born before the cutoff; however, this is not a serious concern here considering the number of observations in the data set. Furthermore, for those born outside of the 60-day bandwidth, this will not be an issue as they will not be included in the estimates. That is, those who would be eligible for this mid-year entry are not part of the analysis since their birth dates are outside of the 60-day bandwidth that the main analyses of this paper uses.

12 For example, the results by gender were done by running separate regressions for the females in the data set and for the males in the data set.

13 I choose this particular age range for adults because the peak years of criminal activity are during the late teens and early 20s (Gottfredson and Hirschi 1983).

14 A felony arrest can result in the arrestee being sentenced to state prison if convicted and a misdemeanor arrest can result in the arrestee being sentenced to county jail, paying a fine, restitution, or probation.

15 When re-estimating Equation (1(1) $\begin{array}{rl}{Y}_i& ={\text{β}}_0+{\text{β}}_1\mathrm{BornAfte}{r}_i+{\text{β}}_2\mathrm{Distanc}{e}_i+{\text{β}}_3\mathrm{BornAfte}{r}_i\ast \mathrm{Distanc}{e}_i\\ & +{\text{β}}_4\mathrm{Distanc}{e}_i^2+{\text{β}}_5\mathrm{BornAfte}{r}_i\ast \mathrm{Distanc}{e}_i^2+{ϵ}_i\end{array}$(1) ) by the type or severity of offense, the corresponding arrest rate only includes those arrested for that particular offense, as shown in Equation (A.3) in the Online Appendix.

16 The arrest rates represent arrests rates per 100,000 people of the relevant population. The arrest rates were constructed using the method that was outlined in Section 2.2.

17 The procedure implements optimal bandwidth calculations proposed by Calonico, Cattaneo, and Titiunik (2014b), Imbens and Kalyanaraman (2012), and Ludwig and Miller (2007).

18 Table A.3 in the Online Appendix shows the various optimal bandwidth calculations.

19 The California Department of Education allows districts to decide early admissions, with the admission criteria being generally based on test results, maturity of the child, or preschool records.

20 Figure A.2 in the Online Appendix shows individual RD plots for status offenses and misdemeanor offenses for 14-year-olds.

21 The percent estimate given here is the percent difference in arrest rates at the cutoff, which was calculated by dividing the RD estimate by the constant–where the constant term also represents the mean arrest rate for those born before the cutoff since the running variable, the distance in days of an individual's birthday from the school entry cutoff, is re-centered around zero.

22 Beatton, Kidd, and Sandi (2019) use data from Australia to find that those who begin school early relative to their peers are more likely to receive school sanctions for certain offenses in their teenage years, but commit less criminal offenses of the same type in their early adult years. The results presented here finds a similar pattern of behavior where those who are early school starters do not commit more serious offenses relative to their older counterparts as they grow older.

23 Figures A.3 and A.4 in the Online Appendix show estimates for status offenses and misdemeanor offenses by age across gender and race/ethnicity.

24 I checked various school district websites and contacted district offices in California to get an approximate school year based on historical school calendars, though the years do not exactly match the time frame of my arrest data. I also use truancy arrests as a guide to the school year since it is an offense that typically only occurs during the school year as seen in Figure A.5 in the Online Appendix–where the arrest trend closely matches what I limit the school year to in my estimates. Thanksgiving break was approximated to be between November 23 and November 28, Christmas break was approximated to be between December 20 and January 4, and summer break was approximated to be between June 15 and September 1

25 Arrest frequency trends for the other types and levels of crime show a similar trend for juveniles.

26 Arrest frequency trends across other types and levels of crime show a similar trend for adults.

27 Figure A.6 in the Online Appendix shows individual RD plots for 14-year-olds for status and misdemeanor offenses during the school year and summer.

28 Tables A.1 and A.2 in the Online Appendix show estimates for status offenses and misdemeanor offenses by gender and race/ethnicity by time of year.