The age-crime curve

View the most recent version.

The amount of recorded crime committed by members of the study population at different ages follows the familiar "age-crime curve". Figure 1 shows the crime rates per 100,000 population at each age from 5 to 17. The difference between the age-crime curves for the two birth cohorts over the age range where they overlap (8 to 14 years) is very small, and is statistically non-significant for most age groups.¹ The crime rates shown for 8 to 14 year olds are the averages for the two birth cohorts. The recorded crime rate is extremely low for 5 year old children (37 offences per 100,000 population) and rises slowly to 890 offences per 100,000 at the age of 11. From 11 years of age, it rises with increasing speed to a peak of 10,111 per 100,000 at 16 years, after which it begins to fall. This relationship between age and crime is so consistent with findings reported for different countries, periods of time, and different ways of measuring criminal activity that it has been claimed that it is invariant across social and cultural conditions (Hirschi & Gottfredson, 1983). The age-crime curves for males and females are broadly similar, but the level of crime by females is of course much lower, and peaks a year earlier (Figure 2).

The crime rates and percentage distributions reported above were calculated by counting each involvement in a criminal incident of each youth in the study population. Therefore, if a youth was involved in several incidents, he or she would be counted more than once in the statistics. This method of counting gives an accurate count of the total number and distribution of involvements in recorded crime, but it does not provide information on how many different youths are involved in recorded crime, nor how many offences of which types each youth is involved in. Theoretically, all 91,491 offences recorded over the 10 year period which involved persons born in 1987 could have been committed by the same person; or they could have been committed by 91,491 different people, or by any combination of the same and different individuals. In order to determine how many different individuals were involved in recorded delinquency during the period of observation, all the incidents involving the same youth were linked together. This procedure resulted in a new unit of analysis: the chronological series of recorded incidents in which a youth is involved, also known as the youth's delinquent career.

---

Note

1. Two-tailed t-values for the differences in the crime rates at ages 8, 9, ..., 14 are 1.15, -1.25, -0.05, -0.22, 2.36, 5.30, and 15.59. The t-statistics were calculated separately for each year-of-age group. For each age group, a file was created containing as many records as the sum of the total populations of the two birth cohorts at that age, including members of the cohort who were not alleged offenders at any age, and therefore not part of the study population. The records for persons accused at that age were coded with the number of recorded incidents for that person at that age. The remaining records for each age group, representing persons not accused at that age, were coded with zero incidents. Thus, the t-statistic for each age group compared the mean number of incidents allegedly committed by the entire population of each cohort, including non-offenders, at that age. These statistics should be interpreted with caution, since these are not normally distributed random samples. Satterthwaite's (1946) approximation for the degrees of freedom was used, since the variances of each pair of vectors of observations were unequal (SAS Institute, 2004). Only the t-value for 12, 13, and 14 year olds are statistically significant at p<.05 (p₁₂=.018, p₁₃<.0001, p₁₄<.0001). The relatively large difference in the crime rates for 14 year olds in the two cohorts is probably due to a slight under-counting of offences in the data for 2005, which would have resulted in a slight under-count of offences committed by 14 year olds born in 1990 (and of 17 year olds born in 1987). The reason for under-counting in the 2005 data is that a small number of incidents occurring each year are not reported to the UCR2 Survey in time to be included in the data for that year. When the UCR2 data are updated the following year, data for the immediately preceding year are restated to include incidents reported in arrears. Thus, the data for all years except the most recent year (2005) had been updated when the data were provided for this study. In the following analyses, unless otherwise specified, rates, proportions, etc. for the age range where the two cohorts overlap (8 to 14 years) are given as the mean of the values for the two cohorts.