News & Updates

November 3, 2021 | Alison Pagan

Robin Hardwick | Washington Defender Association
July 13th, 2021



This post analyzes the 530 individuals recorded in the Department of Corrections database who were sentenced as an adult for a crime they committed as a juvenile. As the database only includes current inmates, all of these people remain incarcerated for crimes they convicted under the age of eighteen as children. In State v. Houston-Sconiers 188 Wn.2d 1, 391 P.3d 409 (2017) the Supreme Court ruled that judges have the right to exercise leeway when sentencing children in adult court, including discretion to depart from otherwise mandatory sentences. However, this ruling only applied to future cases, excluding the 530 individuals currently unfairly impacted and being punished for actions taken as juveniles.

September of 2020 saw a change to the legal precedence when the case Domingo-Cornelio (196 Wn.2d 255, 474 P.3d 524) determined that the ruling in Houston-Sconiers is retroactive. Thus, people sentenced prior to the Houston-Sconiers decision in 2017 are now eligible for re-sentencing as well. The case Ali 196 Wn.2d 220, 474 P.3d 507 (2020) determined that folks are eligible if they also “demonstrate actual and substantial prejudice by a preponderance of the evidence and there are no other adequate remedies available”. (Translation: individuals could be re-sentenced if they can prove the court was unfair toward them and they are not eligible for any other re-sentencing methods). These two landmark cases have the possibility to positively impact the lives of hundreds of currently incarcerated individuals, providing an opportunity at life that was previously deprived from them.

The goal of looking at these 530 individuals is to determine who Washington State has been commuting into long sentences while they were just children. A common argument against giving juveniles’ lengthy sentences as though they were adults is that their brains have not yet finished developing. Without a fully functioning prefrontal cortex, children are at higher risk of making impulsive decisions, not registering risks or consequences, and lacking social intelligence. While such conversation is critically important when understanding the intersection of age with the prison industrial complex, it is also crucial to determine what other roles may prejudice certain children over others to make riskier decisions. For instance, Black and Indigenous children not only face racism from their peers and community members, but also inherit generational trauma and the racial wealth gap. Faced with less options for the future than white kids with more affluent backgrounds, it is highly probable that they are driven to crime at a younger age due to these inequitable environmental and socioeconomic factors.

Regrettably, the data provided does not include race. Instead, to answer some of these crucial questions, I will interrogate which counties are convicting the highest proportion of juveniles for these crimes, and what the average age of conviction is per county for those convicted under the age of 18. Our goal is to identify which counties should be focused on when re-sentencing, as well as to provide evidence to these counties in order for them to take accountability.


Before we can begin to look at specific ages and crimes, it is important to analyze the number of individuals eligible for re-sentencing under Domingo Cornelio and Ali, alongside how the 530 of them are distributed across the Washington State counties. Here we ask: which counties have unfairly targeted juveniles and treated them as adults in the court system?

Unweighted by Population

Let’s first take a look at these 530 individuals, across the sentencing counties, without consideration for population.

This graph shows us that the top three counties that have sentenced juveniles as adults (in order) are: Pierce, King, while Yakima and Snohomish are tied for third. Consider, then, that King County is the most populated county in Washington (with over 1.3 million individuals more than Pierce), while Yakima County is the eighth most populated county despite tying for third place.

What is also important to note about this graph is the counties that are left out—or rather, have no recorded instance of sentencing a juvenile as though they were an adult. These counties are Columbia, Jefferson, Lincoln, Pend Orielle, San Juan, Wahkiakum, and Whitman.

Weighted by Population (Bar Chart)

As we mentioned with the previous graph, Yakima County was especially problematic due to being the eighth most populated county, yet the third in number of juveniles sentenced as adults. Other counties also stand out here, such as Pierce and Okanogan. Unfortunately, my past research on disparity by county (and other demographic variables) in regards to Blake and Robbery 2 cases also sees these counties emerging as problematic. While 530 individuals being sentenced with no consideration for their age is highly concerning, it also points toward similarly concerning systemic issues across the state for multiple types of convictions.

Additionally, while Garfield is shown as having the largest proportion of their population affected, I want to call attention to the number of individuals actually impacted. The numbers to the right of the bars correspond to the number of individuals out of the 530 sentenced in that county as an adult. Only one individual was sentenced in Garfield County. However, as the least populous county (with a total of 2,225 residents), convicting just juvenile as an adult results in a much larger impacted proportion than more populated counties. Thus, I put less stock in the one individual from Garfield, as it does not provide concrete proof of systemic issues in Garfield compared to Yakima or Pierce, both of which have a much larger sample of impacted individuals.

Weighted by Population (Map)

The above visualization takes the last bar plot and turns it spatial, emphasizing which locations in the state are targeting their younger populations the most.

Chi-Square Test

Lastly, to determine if the patterns we have been identifying via the graphs have any merit, I run a chi-square test of significance to determine if the number of individuals targeted by county is statistically significant.

Counties Individuals Convicted                                   Individuals Expected to be Convicted
Adams 2 1.4155576
Asotin 1 1.5996659
Benton 16 14.4785982
Chelan 5 5.4687009
Clallam 3 5.4779807
Clark 31 34.5860623
Cowlitz 5 7.8341974
Douglas 1 3.0764276
Ferry 1 0.5402821
Franklin 7 6.7453451
Garfield 1 0.1576148
Grant 11 6.9232195
Grays Harbor 8 5.3171782
Island 2 6.0312262
King 104 159.5827850
Kitsap 13 19.2306301
Kittitas 3 3.3956241
Klickitat 1 1.5885443
Lewis 9 5.7171301
Mason 7 4.7297179
Okanogan 6 2.9924136
Pacific 3 1.5918028
Pierce 129 64.1070591
Skagit 13 9.1526360
Skamania 1 0.8559367
Snohomish 46 58.2347935
Spokane 31 37.0340143
Stevens 3 3.2389302
Thurston 10 20.5810167
Walla Walla 5 4.3041226
Whatcom 6 16.2394207
Yakima 46 17.7713654

The above table shows the first step of a chi-square test, wherein we compare the number of individuals actually convicted to the number of individuals we would expect to be convicted, based on the population of each county. (The expected column was calculated by multiplying 530, the total number of individuals impacted, by the proportion of the Washington state population the county makes up). With just this table, we can already begin to identify discrepancies between what we expect to see and what we actually see.


Finally, we run our chi-square test and find that a 0.0001 probability that county does not impact the number of juveniles sentenced as adults. In State vs. Gregory, the Washington State Supreme Court ruled that findings with a p-value under 0.11 were enough for the court to consider them as statistically significant and indicative of a relationship. Thus, our p-value is small enough to affirm the patterns we have been seeing—specific counties in Washington State do indeed target juveniles by sentencing them as adults more than others.

The second value printed out above is the Cramer’s V value. In this instance, it is very high, and indicates that 25.38% of the variation in terms of the number of individuals. For just one variable to be able to explain over a fourth of the variation in the way the court systems work here in Washington indicates an extremely problematic and insular system. Additionally, it helps to verify the relationship between two variables—which was proven to us in the chi-square test—though it stands as a check on the sample size of our results.

In short, we conclude that there is a strong relationship between the county a person resides in and the likelihood that children will be tried as an adult for a crime they committed as a juvenile.


The following section takes a closer look at how the numerical data collected about each of our individuals corresponds with the counties. While we know that certain counties target juveniles, here we ask questions such as: do counties target folks who were a certain age when they offended? What about a certain age at the time of sentencing? Does the sentence length vary by county, with certain counties locking away children for longer than others?

Table Overview

Different Ages and Sentence Lengths for Individuals Sentenced as Adults for Offenses Committed as Juvenile
Data provided by the Washington State Department of Corrections
County Average Age at Time of Offense Average Age at Time of Sentencing Average Sentence Length (in Years)
Adams 15.50000 16.50000 11.665000
Asotin 15.00000 20.00000 7.410000
Benton 15.62500 20.31250 13.035625
Chelan 16.40000 18.80000 13.132000
Clallam 15.00000 21.33333 10.220000
Clark 16.29032 19.87097 16.473548
Cowlitz 17.00000 18.40000 15.582000
Douglas 15.00000 21.00000 12.410000
Ferry 14.00000 22.00000 10.000000
Franklin 16.14286 18.28571 11.931429
Garfield 16.00000 21.00000 11.500000
Grant 15.54545 17.09091 17.476364
Grays Harbor 16.62500 19.50000 20.310000
Island 17.00000 18.50000 10.290000
King 16.26923 19.27885 15.166250
Kitsap 15.76923 20.61538 19.831538
Kittitas 16.66667 17.00000 5.916667
Klickitat 15.00000 16.00000 20.000000
Lewis 15.88889 18.77778 11.042222
Mason 16.42857 17.28571 31.581429
Okanogan 15.00000 17.83333 18.318333
Pacific 16.00000 17.00000 23.106667
Pierce 15.84496 18.90698 15.459767
Skagit 16.15385 18.61538 7.394615
Skamania 16.00000 16.00000 0.000000
Snohomish 16.19565 18.02174 17.037609
Spokane 16.45161 18.32258 15.281290
Stevens 16.33333 18.66667 16.666667
Thurston 16.30000 19.10000 11.801000
Walla Walla 16.00000 16.80000 19.230000
Whatcom 16.00000 22.33333 6.750000
Yakima 16.10870 18.65217 15.146087

The above table is an overview and first attempt to answer the above three questions, colored to call attention to ages that are significantly higher or lower than others in the table. The first two columns do not have much variation in color, though the second—the average age of the individual at time of sentencing—has slightly higher ages, though they are internally consistent. This is congruent with the general wait time between offense and a sentencing, which tends to be around one to three years of a difference.

However, the third column, average sentence length (in years) for the juveniles sentenced as adults, varies significantly by county. While it is certainly possible there is a third variable at work that explains some of the variation (such as offense type), the relationship certainly appears statistically significant at first glance.

Average Age at Time of Offense

For all of the following boxplots, the aquamarine squares correspond to the average x-axis value for the county, while the boxes contain the interquartile range (every value from the 25th to 75th percentile; the line represents the median value) in that county. The whiskers stretch out to encompass the other values, while the points are outliers.

The above boxplot provides a closer look at the information in the first column of the table. Average age at the time of offense stays between 14 and 17 years old for all of the counties, though there are notably some outliers that were not emphasized in the table that deserve attention. For instance, there is a nine-year-old child and four eleven-year-olds who were sentenced as adults for crimes they committed at an extremely young age. While there does not seem to be a significant relationship between county and age at time of offense when we consider the average, there is a handful of individuals who are imprisoned for situations that occurred before they were twelve years old.

To verify our above suspicions about the significance of the relationship between county and age at time of offense, I run an ANOVA (Analysis of Variance) test. ANOVA tests are a useful tool to determine if the relationship between a quantitative and categorical variable is statistically significant. Essentially, it looks at both the variation between groups (counties, in this instance) and within groups, and determines how significant our results are by producing an F-statistic.

To understand the between and within group premise, the following graph helps visualize the importance of considering both for ANOVA tests.

In the above visualization, the red dot corresponds to the mean for each group (a, b, and c). In Scenario 1, the means are close together, though they vary between groups in Scenarios 2 and 3. However, we wouldn’t say that Scenarios 2 and 3 show the same relationship between the x and y variables, right? This is because in Scenario 3, there is variation within groups, with a wide range of values in each group category. However, Scenario 2 has minimal variation within its groups, indicating that the variation between the two variables is the most statistically significant in Scenario 2.

Though the graphs look quite different, the boxplot we saw earlier for offense age shows a similar relationship as Scenario 1. The average is close between groups, while the variation within groups is also quite spread out. Thus, the results we obtain from the ANOVA test—a p-value of 0.217—is not very surprising. Since 0.217 > 0.11, our p-value standard for declaring significance in the legal realm, we conclude that the relationship between age at time of offense and county is not significant.


Average Age at Time of Sentencing

While the range of average ages at time of sentencing is higher on this boxplot than the previous, the variation between counties is not very large overall. Additionally, the variation within each individual county is spread out and overlapping. Remember, the vast majority of the juveniles sentenced as adults are contained within the boxes and whiskers (lines stretching out from the boxes).

We do see some outliers, however. There are some individuals who waited decades to be sentenced and are still incarcerated today. While these individuals were adults at the time of the sentencing, this does not detract from the fact that they were juveniles when the offense was committed. Thus, while it seems again that the relationship is not very statistically significant, there are individuals who were sentenced for crimes they committed as juveniles well into their late twenties, individuals who absolutely deserve a re-sentencing hearing.

Df Sum Sq Mean Sq F value Pr(>F) COUNTY       31    413   13.32   0.796  0.778 Residuals   498   8341   16.75

We run another ANOVA test to verify our observations about the boxplot, wherein we find that the p-value is 0.778. Therefore, since 0.778 > 0.11, we do not find evidence for a statistically significant variation between counties in terms of the age they sentence juveniles as adults.


Average Length of Sentence

We see in the above boxplot that variation between counties is much higher here than in the previous two graphs. For instance, we see that the average length of a sentence is quite high in Mason and Pacific Counties, while extremely short in Skamania, Skagit, and Whatcom. However, it is important to call attention to the variation within counties, which seems to overlap quite a bit between the individual counties.

Df Sum Sq Mean Sq F value Pr(>F) COUNTY       31   5391   173.9   1.313  0.123 Residuals   498  65955   132.4

While we have a low p-value of 0.123, we cannot quite determine that there is a relationship between the length of sentences and counties. But, what about the specific counties we identified as standouts for higher or lower average sentence length? Are they individually statistically significant? And, what if we control for the offense type (as the length of a sentence is most often determined by the type of offense)? We turn to a multiple linear regression below to answer these very questions.



To view the regression table for this section, refer to Table 1A in the Appendix. Keep in mind when reading the table that it takes practice and some context to understand what the results mean. Don’t be intimidated by this—we’ll talk through the results and what they mean!

The most important thing to know is to look at the values with stars after them. Model 1: County shows us that a handful of the counties we identified above do indeed have statistically significant variations individually. While all of the counties together gave a p-value of 0.123, Mason and Skagit Counties individually have p-values lower than 0.05, indicating that there is indeed a statistically significant relationship between these two counties and the average sentence length Washington State counties assign to juveniles tried as adults. Mason County assigns longer sentence lengths, an average of 16.42 years longer than those assigned in Adams County. Skagit County assigns juveniles to lower sentence lengths, 7.77 years on average less than Adams County.

But this model doesn’t consider how different types of offenses influence the length of sentences for juveniles sentenced as adults (since certain offenses have mandatory sentence limits). However, when we control for all of the different types of offenses individuals could have committed, even more counties emerge as statistically significant.

Upon controlling for offense type, Grays Harbor, Kitsap, and Mason also emerge as convicting juveniles to longer sentence lengths, which is significant at the p < 0.05 level. On the flip side, Skamania sentences juveniles to shorter sentences of 18.18 years on average compared to Adams County when controlling for different types of offenses. The relationship between Skagit County and sentence length was eliminated once we controlled for offense type.

More discussion of the impacts of different types of offenses is to follow, but Aggravated Murder 1, Assault 1, Murder 1, Murder 2, and Rape 1 result in longer sentences for juveniles sentenced as adults than other offenses (significant at the p < 0.05 level) when controlling for the county. None of the types of offenses result in a statistically significant shorter sentence length when Robbery 1 is the omitted category.



This last section takes a closer look at the type of offenses the 530 individuals were convicted of and determines if there is a relationship between the type of offense and the age a person was when they convicted the crime. Such an investigation allows us to see what ages are targeted the most for specific crimes and if specific counties tend to sentence juveniles as adults for committing certain crimes over others. With this knowledge we can strengthen our argument for the prioritization of re-sentencing certain individuals, individuals who may have been targeted specifically at a very young age because they committed a crime deemed less forgivable than others.



Offense Name Age at Time of Offense Count Proportion
Aggravated Murder 1 17 10 0.4545455
Aggravated Murder 1 16 7 0.3181818
Aggravated Murder 1 15 3 0.1363636
Aggravated Murder 1 14 2 0.0909091
Arson 2 17 1 1.0000000
Assault 1 17 19 0.5135135
Assault 1 16 13 0.3513514
Assault 1 15 5 0.1351351
Assault 1 – Firearm Or Deadly Weapon 16 24 0.6000000
Assault 1 – Firearm Or Deadly Weapon 17 13 0.3250000
Assault 1 – Firearm Or Deadly Weapon 15 2 0.0500000
Assault 1 – Firearm Or Deadly Weapon 14 1 0.0250000
Assault 1 – Great Bodily Harm 17 3 1.0000000
Assault 2 17 3 0.7500000
Assault 2 15 1 0.2500000
Assault 2 – Intentional And Causes Substantial Bodily Harm 17 3 0.7500000
Assault 2 – Intentional And Causes Substantial Bodily Harm 16 1 0.2500000
Assault 2 – With Deadly Weapon 17 8 0.8000000
Assault 2 – With Deadly Weapon 16 2 0.2000000
Assault 2 With Sexual Motivation 17 2 1.0000000
Assault 3 (Weapon Or Other Thing To Cause Harm) 16 1 1.0000000
Assault Of A Child 2 13 1 1.0000000
Burglary 1 17 4 0.6666667
Burglary 1 16 2 0.3333333
Child Molestation 1 15 8 0.2424242
Child Molestation 1 17 8 0.2424242
Child Molestation 1 14 5 0.1515152
Child Molestation 1 13 4 0.1212121
Child Molestation 1 16 4 0.1212121
Child Molestation 1 12 3 0.0909091
Child Molestation 1 18 1 0.0303030
Child Molestation 2 17 4 0.4000000
Child Molestation 2 15 3 0.3000000
Child Molestation 2 16 2 0.2000000
Child Molestation 2 11 1 0.1000000
Communication With Minor For Immoral Purposes 12 1 1.0000000
Custodial Assault 16 1 1.0000000
Driveby Shooting 17 2 1.0000000
Escape 1 17 1 1.0000000
Forgery 17 1 1.0000000
Identity Theft 2 17 1 1.0000000
Incest 1 14 1 0.5000000
Incest 1 16 1 0.5000000
Indecent Liberties (With Forcible Compulsion) 17 1 1.0000000
Indecent Liberties Victim Is Incapable Of Consent 16 1 1.0000000
Kidnapping 1 16 3 0.5000000
Kidnapping 1 17 2 0.3333333
Kidnapping 1 15 1 0.1666667
Manslaughter 1 17 5 0.5555556
Manslaughter 1 16 4 0.4444444
Manslaughter 2 14 1 0.5000000
Manslaughter 2 17 1 0.5000000
Murder 1 17 50 0.4950495
Murder 1 16 32 0.3168317
Murder 1 15 11 0.1089109
Murder 1 14 7 0.0693069
Murder 1 13 1 0.0099010
Murder 2 17 38 0.4318182
Murder 2 16 34 0.3863636
Murder 2 15 15 0.1704545
Murder 2 14 1 0.0113636
Parole Violation 16 1 1.0000000
Rape 1 17 7 0.4666667
Rape 1 15 4 0.2666667
Rape 1 14 2 0.1333333
Rape 1 16 2 0.1333333
Rape 2 16 2 0.3333333
Rape 2 17 2 0.3333333
Rape 2 14 1 0.1666667
Rape 2 15 1 0.1666667
Rape 2 – Victim Is Incapable Of Consent 16 1 1.0000000
Rape 2 With Force 17 2 0.5000000
Rape 2 With Force 9 1 0.2500000
Rape 2 With Force 16 1 0.2500000
Rape 3 16 1 0.5000000
Rape 3 17 1 0.5000000
Rape Of A Child 1 16 12 0.2666667
Rape Of A Child 1 17 9 0.2000000
Rape Of A Child 1 14 6 0.1333333
Rape Of A Child 1 15 6 0.1333333
Rape Of A Child 1 13 5 0.1111111
Rape Of A Child 1 11 3 0.0666667
Rape Of A Child 1 12 3 0.0666667
Rape Of A Child 1 18 1 0.0222222
Rape Of A Child 2 17 8 0.6153846
Rape Of A Child 2 16 3 0.2307692
Rape Of A Child 2 18 2 0.1538462
Rape Of A Child 3 12 1 0.3333333
Rape Of A Child 3 15 1 0.3333333
Rape Of A Child 3 17 1 0.3333333
Residential Burglary 17 4 1.0000000
Robbery 1 17 27 0.6923077
Robbery 1 16 9 0.2307692
Robbery 1 15 3 0.0769231
Robbery 2 17 1 1.0000000
Sexual Exploitation of a Minor 16 1 1.0000000
Sexual Misconduct with a Minor 1 16 1 1.0000000
Theft of Motor Vehicle 16 1 1.0000000
Trafficking Persons 2 16 1 1.0000000
Unlawful Possession of a Firearm 1 17 1 1.0000000


The above table begins to show how age and offense type are related. It records both the count of individuals for each age and offense type, as well as the proportion of people the recorded age convicted of that offense. For instance, 69% of the amount of Robbery 1 convictions are attributed to folks who were 17 at the time. The general trend from this first glimpse is that most charges (Robbery, Murder, Assault, etc.) are caused by folks 15 and up, with a handful of 14-year-olds. Crimes of a sexual nature have a larger range of ages, however, even including 11-year-olds. This trend indicates that courts tend to treat juveniles who commit sexual crimes much harsher, treating children as adults for these crimes despite their age. But for other crimes—even those some would call more extreme, such as murder—many children under 14/15 are sentenced as juveniles, an exception not granted in cases of a sexual nature.


Bar Plot

This graph visualizes the same trends that we saw above, wherein the majority of crimes are made up 16- to 17-year-olds (with some 14- and 15-year-olds). However, for charges of a sexual nature (Incest, Rape/Rape of a Child, Assault of a Child, etc.) we see a more diverse age range. Thus, the state has upheld an unequal system against juveniles by targeting younger individuals more harshly when they commit crimes of a sexual nature.



To test the above findings, we must look at each offense type individually and see how significant it is in relation to age at time of the offense. We do so through a regression, results for which are reproduced in the Appendix (Table 2A). Once again, Robbery 1 is the omitted category (meaning all of the coefficients are comparing the difference between the average offense age for Robbery 1 convictions to the corresponding offense). All of the statistically significant offenses are negative (meaning the offense age is less than it is for individuals convicted of Robbery 1) and are offenses of a sexual nature (Assault of a Child 2, Child Molestation, etc.). It is concerning that one type of offense targets juveniles, and that children as young as even 11 years old could be considered the same as an adult in court because they committed an offense the judge finds more reprehensible than other crimes.

It is worth noting that even Aggravated Murder in the First Degree—perhaps the most extreme charge a person can receive in a court of law in terms of sentence length—does not sentence younger individuals at a statistically significant rate.



I also ran some tests to determine if the types of offenses was statistically significant on the county level—or, in other words, if certain counties target an offense type more than others. However, the results were not statistically significant, and the charts do not show any conclusive patterns. These results are printed below if you are interested.


Bar Plot

This bar plot resembles a rainbow, and as such is difficult to discern any conclusive patterns from. (With so many categories for each variable, it’s near impossible to visualize in any meaningful way).


Chi-Square Test

Pearson’s Chi-squared testX-squared = 1405.7, df = 1364, p-value = 0.2109

The above is a Chi-square test for the relationship between type of offense and the Washington State counties. We obtain a p-value of 0.2109, which is not low enough to provide statistical significance.



The following section offers an analysis of how juveniles being sentenced as adults has changed over time in Washington State. In this analysis, we cover how county, offense age, and sentence lengths are impacted by the passage of time. Such an interrogation allows us to understand how Washington has changed in their treatment of juveniles over time.

The above histogram indicates that the phenomenon of sentencing individuals who commit offenses as children as if they are adults escalated in the mid-nineties. Historical context supports this finding, as the super predator myth began in the mid-nineties by Princeton professor John DiIulio. This mentality contributed to the shift in the criminal “justice” framework to support prosecuting children in adult court for all types of crime due to a new societal perception of an increase in drugs and violence. Everyone, even children (especially Black children), were caught up in this myth, and the general public began to think that even children were or would quickly evolve into these “super predators.” Thus, the state began to believe that if someone has committed a crime in what they deem is a grown body, they must do the same jail time that an adult would.

We see this ideology emerge within the state specifically through two Washington State Supreme Court cases. The first, State v. Scott 72 Wn.App. 207, 866 P.2d 1258 (1993), argues that youthfulness or young age is not a reason to sentence children to jail sentences shorter than the “standard range.” The second court case, State v. Ha’mim 132 Wn.2d 834, 940 P.2d 633 (1997), upholds Scott unless “the defendant could not appreciate the wrongfulness of her conduct.” However, in this case, the individual who was prosecuted for Robbery One was eighteen years old, and the court did not find evidence of such “appreciation.” Thus, the court still does not consider age itself for shorter sentences and requires outside factors to provide children leniency.

What this myth fails to consider is that the brain does not finish maturing until the mid-twenties. Children are more likely to be driven toward impulsive behavior, which encompasses drive-by shootings, robberies, and drug involvement. Therefore, the state considers the actions of children who are still developing as evidence that they will become “super predators” instead of realizing that immaturity and an inability to resist impulsive decisions is a direct consequence of young age. While Domingo Cornelio and Ali are steps in the right direction, and the number of juveniles sentenced as individuals has dropped off since 2018, there are still hundreds of people across the country incarcerated for offenses they committed as children and were sentenced as if they were adults.


Counties and Offense Date

The above boxplot looks at the period of time that the counties have most frequently sentenced juveniles as though they are adults. As we observed in the previous histogram, the majority of counties sentence children as adults after the mid-nineties. Just about all of the average dates (blue squares) are after 1995, excluding Klickitat and Skamania, who only ever sentenced one juvenile as an adult.

If we think back to the three scenarios shown in the ANOVA boxplot example, our graph follows the most closely with Scenario 1, where variation between groups is minimal and variation within groups is wide.

Df    Sum Sq  Mean Sq F value Pr(>F) COUNTY       31 3.530e+08 11387803   1.091   0.34 Residuals   498 5.199e+09 10439436

To double check our above observation, I run an ANOVA test. With a p-value of 0.34, we do not have support that counties differ in the time period that they tend to sentence juveniles as adults. While overall they sentence these individuals more after 1995, between counties there is no significant difference.


Offense Age and Date

As the above graph confirms, the number of juveniles sentenced as adults increases dramatically after the mid-nineties. Additionally, children under the age of fourteen were not ever sentenced as adults until after the mid-nineties, meaning that the “super predator” myth also allowed people to conceptualize of very young children as deserving of being sentenced as if they are adults.

Pearson’s product-moment correlationt = 2.4148, df = 528, p-value = 0.01608alternative hypothesis: true correlation is not equal to 095 percent confidence interval:0.01952026 0.18801390sample estimates:correlation coefficient (ρ)0.104517

To verify the pattern shown in our graph, I run a Pearson’s correlation. A correlation shows the strength of a linear relationship with our data, wherein a positive linear relationship means that as the x variable increases, so does the y variable (which means for us that as time goes by, the age at time of offense also increases). Conversely, a negative linear relationship means the opposite, wherein the y variable decreases as the x variable does.

Our p-value (0.01608) is less than 0.11, meaning that the relationship between age at time of offense and date of the offense have a linear relationship. ρ, or the correlation coefficient, is 0.105. This indicates a weak positive relationship between our variables. Such a value means that over time, the average age of individuals convicted slightly increases. We can attribute this to the large uptick in 16- and 17-year-olds convicted after the mid-nineties, even while a few children under 14 were sentenced as adults then too.


Sentence Length and Offense Date

As time has progressed, the length of sentences received by juveniles treated as adults has decreased. While this may seem counter-intuitive to all of the results from above, it is important to consider that more juveniles are being sentenced in recent years than before even though juveniles are not necessarily committing more crimes. Therefore, previous years have less data than recent years. Plus, since juveniles were not often sentenced as adults until after the 1990s, the juveniles who were sentenced as adults before then were convicted of the most serious offenses. These more serious offenses have a longer minimum sentence length, meaning that because in more recent years children are being sentenced as adults at a much higher rate for crimes with a shorter minimum sentence length, the overall sentence length has also been decreasing over time.

To see this difference in context, these are the offenses of the first ten juveniles in Washington State ever sentenced as adults:

Offense Name Date of Offense
Parole Violation 1962-05-07
Murder 2 1974-09-04
Murder 2 1975-02-14
Aggravated Murder 1 1976-12-15
Kidnapping 1 1978-01-27
Murder 1 1978-04-28
Murder 1 1978-08-09
Murder 1 1979-05-05
Murder 2 1979-08-13
Manslaughter 2 1980-04-04

while these are the offenses of the ten most recent juveniles sentenced as adults:

Name of Offense Date of Offense
Driveby Shooting 2019-09-28
Robbery 1 2019-08-16
Assault 1 2019-05-20
Assault 2 2019-05-16
Assault 2 – With Deadly Weapon 2019-04-17
Assault 2 – With Deadly Weapon 2019-03-14
Assault 1 – Firearm Or Deadly Weapon 2019-03-02
Assault 2 – With Deadly Weapon 2019-02-25
Assault 1 – Firearm Or Deadly Weapon 2019-02-23
Murder 2 2018-12-10

Pearson’s product-moment correlationt = -11.215, df = 528, p-value < 2.2e-16alternative hypothesis: true correlation is not equal to 095 percent confidence interval:-0.5049191 -0.3671559sample estimates:correlation coefficient (ρ)-0.4386107

To prove the above results, I run a second correlation to test sentence length against offense year. Here we find an extremely statistically significant relationship with a p-value of approximately zero, meaning we can conclude that there is a relationship between these two variables. As the linear model line shows on the above graph, the relationship is a strong negative one, shown in the ρ of -0.439.



As this section has discussed relationships across variables—proving claims by showing qualitative evidence of offense type—I affirm my above claims by running a final multiple linear regression, shown in the Appendix’s Table 3A. In the regression, we first look at the relationship between age at time of offense and the year of offense. Just as above, this relationship is significant and positive.

However, the next model controls for the different types of offenses, and the effect of age on time is no longer significant. This confirms the above patterns I identified, because juveniles have started to be sentenced as adults for less severe crimes since the 1990s and the popularization of the “super predator” myth. Just take a look at the types of offenses and the sign on coefficients marked as significant. Aggravated Murder 1 charges happened the most, on average, 14.17 years before the average year that juveniles were sentenced as adults for Assault 1 charges. Other more extreme crimes (Kidnapping 1, Manslaughter 2, Murder 1) are also statistically significantly more likely to occur earlier. Parole violation as well—though this could also point at how children are locked up now and thus unable to violate their parole.

Less severe crimes—crimes which juveniles are only treated as adults for since the 1990s due to the “super predator” myth—are significantly more likely to impact juveniles sentenced as adults in later years on average. Such crimes include Robbery 1, Assault 1 – Firearm or Deadly Weapon, Assault 2 – Intentional and Causes Substantial Bodily Harm, Assault 2 – With Deadly Weapon, Burglary 1, and Manslaughter 1.

Thus, we can also point at time periods to explain the increased number of juveniles who are sentenced for specific offenses. That is, for less severe crimes (Burglary, Assault, etc.), juveniles have been sentenced as adults later (after the 1990s) on average. These results are statistically significant.



In short, this report provides statistical support for certain county’s targeting of juveniles by sentencing them as adults more than others (including Yakima, Pierce, and Okanogan). Additionally, while the age a juvenile committed the offense and the age at the time of sentencing does not differ significantly by county, the length of sentence does vary for certain counties, even when controlling for the different offenses committed. Grays Harbor, Kitsap, and Mason convict juveniles for longer sentences on average, while Skamania sentences juveniles to shorter sentences. We also find evidence that the state as a whole treats juveniles harsher based on the type of crime committed, with children being sentenced as adults at as young as nine if the crime is of a sexual nature. Even for more violent crimes such as Aggravated Murder and Assault, juveniles are sentenced as adults until they are about 16. This is proved significant in a regression.

Evidence also points to the importance of time period and the “super predator” myth to explain the uptick in juveniles sentenced as adults in the mid-nineties. We can also determine that less extreme crimes (assault, burglary) only began to treat juveniles as adults in more recent years, with a regression wiping out the impacts of age on time and indicating such offenses explained the variation we saw by age. Lastly, we check to see if counties target specific offenses more than others. The results did not provide support for this hypothesis.

The counties mentioned above, individuals convicted after the 1990s, as well as the individuals who were targeted at an extremely young age in comparison to the other 530 individuals due to their offenses being of a sexual nature, all deserve immediate, swift, and targeted re-sentencing. Since Domingo Cornelio and Ali determined that people sentenced as juveniles deserve a new hearing with reconsideration of their age, for folks who were disproportionally and unfairly impacted—be that because of their county of origin or because their type of offense—they deserve swift re-sentencing.

Please reach out to the author at with any questions you may have.

This study was supported in part by the Paul K. Richter & Evalyn Elizabeth Cook Richter Memorial Fund.




Table 1A. Multiple Linear Regression on Sentence Length, by County and Offense Type

 VARIABLES  Model 1: County Model 2: Offense Type
(Intercept) 11.67 9.42
(8.14) (6.36)
Asotin -4.26 0.04
(14.09) (10.36)
Benton 1.37 -4.67
(8.63) (6.58)
Chelan 1.47 -0.08
(9.63) (7.28)
Clallam -1.45 -0.86
(10.51) (7.87)
Clark 4.81 1.37
(8.40) (6.41)
Cowlitz 3.92 2.54
(9.63) (7.25)
Douglas 0.74 5.04
(14.09) (10.36)
Ferry -1.67 -2.93
(14.09) (10.35)
Franklin 0.27 -3.76
(9.23) (6.96)
Garfield -0.17 2.67
(14.09) (10.54)
Grant 5.81 -0.85
(8.85) (6.69)
Grays Harbor 8.64 8.16
(9.10) (6.77)
Island -1.38 -2.64
(11.51) (8.60)
King 3.50 -1.49
(8.22) (6.25)
Kitsap 8.17 4.36
(8.74) (6.66)
Kittitas -5.75 -5.80
(10.51) (7.95)
Klickitat 8.33 -11.05
(14.09) (10.29)
Lewis -0.62 1.86
(9.00) (7.05)
Mason 19.92 * 9.79
(9.23) (6.96)
Okanogan 6.65 -0.65
(9.40) (7.12)
Pacific 11.44 3.76
(10.51) (7.86)
Pierce 3.79 0.62
(8.20) (6.26)
Skagit -4.27 -3.75
(8.74) (6.52)
Skamania -11.67 -19.67
(14.09) (10.42)
Snohomish 5.37 -1.87
(8.31) (6.35)
Spokane 3.62 -3.79
(8.40) (6.39)
Stevens 5.00 -4.55
(10.51) (7.84)
Thurston 0.14 0.31
(8.91) (6.93)
Walla Walla 7.56 0.91
(9.63) (7.42)
Whatcom -4.92 -6.84
(9.40) (7.16)
Yakima 3.48 0.45
(8.31) (6.34)
Aggravated Murder 1 10.25 ***
Arson 2 -1.72
Assault 1 4.08 *
Assault 1 – Firearm or Deadly Weapon 2.93
Assault 1 – Great Bodily Harm -2.10
Assault 2 -4.29
Assault 2 – Intentional and Causes Substantial Bodily Harm -4.07
Assault 2 – With Deadly Weapon -5.05
Assault 2 With Sexual Motivation 5.58
Assault 3 (Weapon or Other Thing To Cause Harm) -8.62
Assault of a Child 2 -6.87
Burglary 1 -2.12
Child Molestation 1 -2.05
Child Molestation 2 -3.23
Communication With Minor for Immoral Purposes -5.29
Custodial Assault -10.28
Drive by Shooting -9.78
Escape 1 -9.47
Forgery -7.61
Identity Theft 2 -4.35
Incest 1 -4.79
Indecent Liberties (With Forcible Compulsion) 2.85
Indecent Liberties Victim Is Incapable of Consent -7.05
Kidnapping 1 6.90
Manslaughter 1 0.34
Manslaughter 2 -2.57
Murder 1 21.63 ***
Murder 2 8.56 ***
Parole Violation -2.43
Rape 1 7.16 **
Rape 2 4.75
Rape 2 – Victim Is Incapable of Consent -2.83
Rape 2 With Force -0.32
Rape 3 -4.89
Rape of a Child 1 3.51
Rape of a Child 2 -0.59
Rape of a Child 3 -5.47
Residential Burglary -6.76
Robbery 2 -3.97
Sexual Exploitation of a Minor 0.95
Sexual Misconduct with a Minor 1 -13.07
Theft Of Motor Vehicle -7.96
Trafficking Persons 2 -0.79
Unlawful Possession of a Firearm 1 -1.30
Observations (N) 530 530
R2 0.08 0.58

All continuous predictors are mean-centered and scaled by 1 standard deviation.

*** p < 0.001;  ** p < 0.01;  * p < 0.05.


Table 2A. Multiple Linear Regression on Offense Age, by Offense Type

VARIABLES Model 1: Offense
(Intercept) 16.62 ***
Aggravated Murder 1 -0.48
Arson 2 0.38
Assault 1 -0.24
Assault 1 – Firearm or Deadly Weapon -0.39
Assault 1 – Great Bodily Harm 0.38
Assault 2 -0.12
Assault 2 – Intentional and Causes Substantial Bodily Harm 0.13
Assault 2 – With Deadly Weapon 0.18
Assault 2 With Sexual Motivation 0.38
Assault 3 (Weapon or Other Thing To Cause Harm) -0.62
Assault of a Child 2 -3.62 **
Burglary 1 0.05
Child Molestation 1 -1.59 ***
Child Molestation 2 -1.02 *
Communication With Minor for Immoral Purposes -4.62 ***
Custodial Assault -0.62
Drive by Shooting 0.38
Escape 1 0.38
Forgery 0.38
Identity Theft 2 0.38
Incest 1 -1.62 *
Indecent Liberties (With Forcible Compulsion) 0.38
Indecent Liberties Victim Is Incapable of Consent -0.62
Kidnapping 1 -0.45
Manslaughter 1 -0.06
Manslaughter 2 -1.12
Murder 1 -0.40
Murder 2 -0.38
Parole Violation -0.62
Rape 1 -0.68 *
Rape 2 -0.78
Rape 2 – Victim Is Incapable of Consent -0.62
Rape 2 With Force -1.87 **
Rape 3 -0.12
Rape of a Child 1 -1.70 ***
Rape of a Child 2 0.31
Rape of a Child 3 -1.95 **
Residential Burglary 0.38
Robbery 2 0.38
Sexual Exploitation of A Minor -0.62
Sexual Misconduct with A Minor 1 -0.62
Theft Of Motor Vehicle -0.62
Trafficking Persons 2 -0.62
Unlawful Possession of a Firearm 1 0.38
N 530
R2 0.25

All continuous predictors are mean-centered and scaled by 1 standard deviation.

*** p < 0.001;  ** p < 0.01;  * p < 0.05.


Table 3A. Multiple Linear Regression on Time (Measured in Years) by Offense Age and Type

VARIABLES              Model 1:  Age Model 2:  Offense Type
(Intercept) 2008.49 *** 2007.79 ***
(0.38) (1.18)
Offense Age 0.93 * 0.43
(0.38) (0.36)
Aggravated Murder 1 -14.17 ***
Arson 2 8.89
Assault 1 – Firearm or Deadly Weapon 6.11 ***
Assault 1 – Great Bodily Harm 8.22
Assault 2 3.57
Assault 2 – Intentional and Causes Substantial Bodily Harm 8.98 *
Assault 2 – With Deadly Weapon 9.76 ***
Assault 2 With Sexual Motivation -5.61
Assault 3 (Weapon or Other Thing To Cause Harm) 8.24
Assault of a Child 2 6.29
Burglary 1 7.17 *
Child Molestation 1 1.61
Child Molestation 2 2.28
Communication With Minor for Immoral Purposes -5.36
Custodial Assault 9.24
Drive by Shooting 8.89
Escape 1 8.89
Forgery 7.89
Identity Theft 2 8.89
Incest 1 2.09
Indecent Liberties (With Forcible Compulsion) 0.89
Indecent Liberties Victim Is Incapable of Consent 8.24
Kidnapping 1 -10.15 **
Manslaughter 1 7.38 **
Manslaughter 2 -10.08
Murder 1 -3.93 **
Murder 2 2.20
Parole Violation -45.76 ***
Rape 1 -1.00
Rape 2 -1.53
Rape 2 – Victim Is Incapable of Consent 7.24
Rape 2 With Force 0.68
Rape 3 7.57
Rape of a Child 1 -0.33
Rape of a Child 2 1.53
Rape of a Child 3 6.71
Residential Burglary 4.89
Robbery 1 6.79 ***
Robbery 2 7.89
Sexual Exploitation of a Minor 4.24
Sexual Misconduct with a Minor 1 6.24
Theft Of Motor Vehicle 9.24
Trafficking Persons 2 8.24
Unlawful Possession of A Firearm 1 -8.11
N 530 530
R2 0.01 0.40

All continuous predictors are mean-centered and scaled by 1 standard deviation.

*** p < 0.001;  ** p < 0.01;  * p < 0.05.