Does Having a Minor Child Affect Criminal Charges and Sanctions Imposed on Female Defendants?

Abstract

This study examined effects of having a minor child(ren) on the probability of being prosecuted, convicted, and if convicted, the sanctions that were imposed. Data were state-wide court and birth records of criminally-charged women in North Carolina, a state with sentencing guidelines. We hypothesized that (a) prosecutors would be less likely to prosecute and more likely to lower an offense class and (b) judges (when they had discretion) would be more lenient for women in sentencing with minor children than without. Having a minor child(ren) reduced the probability of prosecution; given prosecution, conviction rates fell. When the judge had discretion, having minor children reduced the probability of an active sentence. Having a minor child had no effect on minimum sentence length for women with active sentences. Presence of a minor child affects prosecutorial and judicial decisions affecting women charged with a criminal offense.

Introduction

From 1980 to 2017, U.S. state and federal incarcerations for women grew nearly 10-fold from 13,206 in 1980 to 111,360 (The Sentencing Project, 2019). The growth in sanctioning of women is concerning, partly because 70–79 percent of women in jail are mothers, most with dependent children (Bloom, Owen, & Covington, 2004; McCampbell, 2005). A substantial body of research has shown that incarceration of a parent can have harmful effects on child(ren), such as leading to sleep disruption and unhealthy eating behaviors, homelessness, food insecurity, worse self-reported health, and worse mental health (Gifford et al., 2019; Jackson & Vaughn, 2017; Turney, 2014a; Wildeman, 2014; Wildeman, Goldman, & Turney, 2018). Yet, relatively few studies have examined how parental status affects decisions throughout the criminal justice process that ultimately leads toward sentencing. In contrast, the role of sex on sentencing decisions has been extensively studied, and results find that women tend to have shorter incarceration stays than men (Kruttschnitt & Green, 1984; Spohn & Beichner, 2001; Steffensmeier & Demuth, 2006).

While studies of sex-specific differences in sentencing assume that women are more likely to have parenting roles, few studies have examined whether parental status affects sentencing, or more generally, how being a parent changes how arrests are resolved by the criminal justice system. It is unknown how status as a mother affects procedural decisions--starting from the time the individual is charged with a criminal offense, when the decision is made to prosecute the charge(s), whether to convict conditional on being prosecuted, as well as the sentencing decision, conditional on being convicted on at least one charge. A substantial percentage of charges are never prosecuted (Bellin, 2019).

This study addressed these research questions: 1. Of cases involving women who were charged with a criminal offense, in what percentage of closed cases were the charged offenses prosecuted? Did prosecuted offenses result in a conviction? Among those convicted, what percentage were sentenced to incarceration in jail or prison? And, if sentenced to incarceration, what was the length of the minimum sentence? 2. How did the presence of minor children affect the probability of prosecution, conviction, sentence type, and sentence length? Since judicial discretion has been limited by sentencing guidelines, we expected that prosecutors would have more opportunities to exercise discretion than judges. We hypothesized that when the defendant had at least one minor child, relative to having no minor children, (1) prosecutors would be less likely to prosecute and would be more likely to lower an offense class (2) judges, when they had the discretion to do so, would show more leniency in sentencing. We found empirical support for both hypotheses.

This study used statewide criminal court data from North Carolina on all criminal charges filed against women with disposition dates between 2009–12. While several states have released criminal court data, what makes the North Carolina data unique is that we could link court data to birth records data from 1992–2012 for an entire state to identify whether or not women appearing in criminal court records had a minor child of a specific age on the date when charges were resolved. The ability to merge statewide birth records with statewide court records resulted in a much larger data set than has been available in previous studies.

Background

Legal Context

As of 2018, 25 states had sentencing guidelines and/or sentencing commissions; states vary in the prescribed length and type of incarceration (Mitchell, 2017; Robina Institute, 2018). In North Carolina, the punishment for individuals who are convicted of misdemeanor and felony offenses committed on or after October 1, 1994 has been subject to structured sentencing law. Under this law, the punishment an offender receives is based on the offense severity and the defendant’s prior criminal record. Punishment types for convicted offenses include: (a) active sentence (incarceration in jail or prison); (b) intermediate --supervised probation with at least one of following conditions: special probation, residential program, electronic house arrest, intensive supervision, day reporting center, or drug treatment court (Spainhour & Hall, 2014); and (c) community punishment which is any other non-active sentence (Spainhour & Hall, 2014). Some convicted offenses require active punishment while other offenses allow for community or intermediate punishment. For certain offense classes and prior record levels, a judge has discretion over active, intermediate, or community punishment. Despite strict sentencing guidelines, sentencing practices vary greatly within state, depending partly on attributes of the judge assigned to the case and the county court, including court size and caseload pressure (Johnson, 2006; Ulmer, Eisenstein, & Johnson, 2010; Ulmer & Johnson, 2004).

In North Carolina, judges can incorporate mitigating and aggravating factors into their decision. One mitigating factor is whether the defendant supports dependents (North Carolina Sentencing and Policy Advisory Commission, 2009).¹ Since North Carolina had sentencing guidelines throughout this study’s observational period, our results are likely to be more representative of states with than without such guidelines.

Before implementation of sentencing guidelines and sentencing commissions, judges had considerable latitude when weighing costs versus benefits of incarceration in relation to other non-incarceration sanctions (Kramer & Ulmer, 2009). As a result of strict sentencing guidelines, judicial discretion in sentencing has been curtailed. It has been hypothesized that the discretion that judges have lost has shifted to increased prosecutorial discretion (e.g., Shermer & Johnson, 2010). The prosecutorial office can exercise discretion at numerous steps in the process such as in deciding whether or not to pursue or decline criminal charges, factoring in the available evidence to obtain a conviction and resource constraints such as staffing (Miller & Wright, 2008; Wright, 2012) and whether or not to charge the arrestee with the original charge or reduce the charge to a lesser offense. Up until final disposition, the prosecutor’s office can negotiate with the defendant and her attorney for a conviction through plea bargaining (Shermer & Johnson, 2010). Plea bargains allow defendants to choose the certainty of a sanction—but at a reduced level—over possibly higher sanctions (Bibas, 2011; Landes, 1974).

Theoretical Perspectives

Several theories have been advanced to explain why women may receive different sentences than men including focal concerns theory (Steffensmeier, Ulmer, & Kramer, 1998; Steffensmeier, 1980), informal social control theory (Black, 1976; Sherman, Smith, Schmidt, & Rogan, 1992) and familial paternalism theory (Cho & Tasca, 2018; Daly, 1987b). These theories are interrelated, but even taken separately, they account for reasons which may lead to different sentencing practices applied to criminal cases involving women versus men. These theories acknowledge how family roles and responsibilities may result in the observed sentencing disparity between women and men and may be extended to suggest that mothers of minor children may receive more leniency than women not caring for minor children.

In focal concerns theory, judges and other court actors are guided by three concerns in reaching sentencing decisions: (1) blameworthiness—defendant culpability and having the punishment fit the crime, (2) protection of the community, and (3) practical constraints, e.g., jail capacity, and consequences including the potential effects of incarceration on the defendant’s children and other family members. By this rationale, court actors may be more lenient on mothers relative to non-mothers because of their parenting role. Similarly, when individuals have higher levels of informal social control mechanisms, such as being economically dependent on others, they are more likely to abide by the laws and thus, formal control mechanisms are less needed (Black, 1976; Sherman et al., 1992); this also suggests a reason for mothers receiving more lenient sentencing. Similarly, familial paternalism theory suggests that because the caregiving role of children and dependent relatives falls disproportionately on women and because of this informal control mechanism, they are subject to more lenient sentences than men.

To inform theoretical explanations regarding what motivates court actors sentencing decision, interviews and surveys with court actors have been conducted (Daly, 1987b, 1989b; Freiburger, 2010). Interviews with 35 prosecutors, defense attorneys, probation officers, and judges from Springfield Massachusetts demonstrated support for familial paternalism theory whereby court actors expressed concern for protecting family life, men and women’s employment to support their families, as well as those who depend on the defendant (Daly, 1987b). Judges in Pennsylvania were interviewed to explore why women tended to have more lenient sentences than men and revealed that objectives toward judicial protection were not directed at women, but rather at their children (Daly, 1989b). To elicit the judges’ views on the effect of defendants’ familial roles on sentencing outcomes, 360 Court of Common Plea judges were surveyed using vignettes (Freiburger, 2010). Results found that if the defendant performed a caretaker role, the probability of incarceration decreased. The effect size was larger for men than for women defendants.

Previous Research on Role of Family Responsibilities in Sentencing

Few studies have examined the effect of having a child on sentencing outcomes. The earliest studies examined the role of the defendant’s dependents—but were not specific to children (Bickle & Peterson, 1991; Daly, 1987a, 1989a). For example, how case outcomes vary as a function of defendants’ familial roles (i.e., marital status (single or married)) crossed with having dependents (yes or no) was studied in a sample of 1,621 defendants (11% women) from a single lower criminal court in New York City during 1974–75 (Daly, 1987a). Having dependents and whether the defendant was married or not was associated with increased leniency by the court, more so for women than for men. Subsequently, this sample was supplemented with data from a Seattle Court in 1973, and the earlier study’s finding of the importance of family ties in sentence severity were reinforced (Daly, 1989a).

Likewise, the effect of familial responsibilities on being sentenced to prison or receiving a less severe sentence was examined in a sample of 514 defendants (24% female) who were convicted of forgery in eight federal district courts during 1973–78 (Bickle & Peterson, 1991). For women, familial roles (marital status, having economic dependents, providing emotional support) did not predict whether or not the convicted individual received a prison sentence. For men, receiving a prison sentence was less likely for those who provided significant emotional support to dependents and more likely for those who were married, cet. par.

The introduction of judicial sentencing guidelines has been used to study how court officials considered the defendant’s sex and parenting role in the decision to incarcerate (Griffin & Wooldredge, 2006; Koons-Witt, 2002). Analysis of a Minnesota samples revealed that defendant’s sex alone did not affect the probability of incarceration, but women with dependent children had a significantly lower probability of being incarcerated than women without children both before and after sentencing guidelines were implemented (Koons-Witt, 2002). Koons-Witt (2002) inferred that the adoption of sentencing guidelines did not reduce judges’ ability to base sentencing decisions on defendants’ familial roles.

However, other studies did not confirm this finding (Griffin & Wooldredge, 2006; Stacey & Spohn, 2006). Results from a sample of Ohio felony convictions suggested that the length of prison sentence was less impacted by the defendant’s sex and number of dependent children after guidelines were implemented relative to before the guidelines were implemented (Griffin & Wooldredge, 2006). Stacey and Spohn (2006), using data from three U.S. district courts, found that while female defendants received shorter sentences than male defendants, there were no statistically significant differences in sentence length imposed on convicted offenders with and without dependent children.

More recent studies have tended to find that child care taking roles of women defendants are associated with increased leniency by courts (Cho & Tasca, 2018; Doerner & Demuth, 2014; Freiburger, 2011). Freiburger (2011) examined the effect of living with a child and the defendant’s family roles (e.g., whether parent visits child, pays child support) whether or not the defendant was sentenced to incarceration. Data came from 426 drug and property offense cases from Allegheny County, Pennsylvania. Results found that living with a child or paying child support was associated with a reduced odds of incarceration. A study using data from the U.S. Sentencing Commission found that defendants who had one or more dependents had a reduced probability of being incarcerated relative to defendants with no dependents; however, if incarcerated, the existence of dependents was not associated with a reduction in sentence length (Doerner, 2012). Further analysis revealed that the effect of dependents only applied to women (Doerner & Demuth, 2014). Using a sample of women who were incarcerated in Arizona in 2010 (n=419, 69% were mothers) Cho and Tasca (2018) found that maternal status was not associated with sentence length, but that mothers who lived with a child were more likely to have a shorter sentence length.

In sum, a variety of specifications have been employed to gauge the impact of parental involvement with minor children on sentencing outcomes. Most studies have been plagued by lack of statistical power, and often the jurisdictions from which the data were drawn have been small, much smaller than a single state. The studies have focused on sentencing decisions to the exclusion of decisions prior to sentencing. Our study seeks to fill these gaps.

Method

Data

Our data came from two sources. First, information on criminal offenses came from the Automated Criminal Infractions System (ACIS) maintained by the North Carolina Administrative Office of the Courts (AOC). ACIS contained separate records for each individual charged offense including charge date, information on the related state statute and offense description, and type and date of final disposition. The second data source was birth records from North Carolina Vital Statistics.

From the court records, we created a longitudinal file for each individual represented by one or more records. ACIS data contained identifying information, including the person’s name, address on the date on which criminal charges were filed, birth date, gender, and race/ethnicity. This information was used to create a unique identifier to permit tracking of criminal charges and associated outcomes by person over time. The data included information on how the defendant arrived in court (e.g., citation, warrant, criminal summons), the method of disposition of each offense (e.g., judge, jury, dismissed, deferred), verdicts (e.g., guilty, not guilty, prayer for judgement), sentences (e.g., jail and prison terms, fines, community service hours), and dates for each key event (e.g., charge date, disposition date, etc.). The first date was the date the charge was filed. This date often coincided with the date the offense was committed, such as for charges of Driving Under the Influence, but for other offenses, the delay could have been much longer depending on when the alleged offense was reported, the accuracy of the description of the alleged perpetrator, and the ability of the perpetrator to avoid police detection, among several factors (Beauregard & Martineau, 2014; Hipwell et al., 2018; Stefanska & Carter, 2019). The disposition date reflects the date the outcome of the charged offense was determined.

We used ACIS data on all felony and misdemeanor criminal offenses that were resolved by the judicial system from 2009–2012 involving women born in 1960 or later. We had access to ACIS data from 2005–2015. Limiting our sample to 2009–2012 allowed us to document who were mothers of minor children aged 0–17 years at the date of the case disposition. We obtained birth records from 1987 to 2012. We used ACIS data from 2006–2008 exclusively for conviction histories during the three years prior to case disposition.

The 10 most common offenses from the cases in our sample were: (1) speeding, (2) expired registration card/tag, (3) driving while license revoked, (4) no operator’s license, (5) operating a vehicle with no insurance, (6) misdemeanor larceny, (7) display canceled, revoked, suspended or altered registration certificate/tag, (8) simple assault, (9) Driving While Impaired, and (10) possession of drug paraphernalia.

We aggregated all charged offenses that were resolved by the judicial system on the same day, either the date when the prosecutor declined to prosecute, the conviction date, or the sentencing date, whichever occurred later, into a single observational unit, which we term a “case.” We used the charge date for the purpose of applying the guidelines applicable to the charge, such as charged offense class.

Second, we created a longitudinal file of children’s birth records from 1992 to 2012. As in other states, the birth record contained demographic information, such as birth date and sex of child, and as well as parents’ names and birthdates.

Third, we merged the two files as a means to indicate if a woman had a minor child at the time her criminal case was closed. Record linkage occurred in a staged matching process whereby in the first stage we used an exact matching algorithm based on mother’s name, including maiden name, and her birthdate from the birth records and defendant’s name, birthdate and gender from ACIS. Then, for records that did not match in the first stage, we relaxed the exact match criterion and considered a subset of the name (i.e. Elizabeth and Beth), spelling distance (using the SPEDIS procedure in SAS), and allowing for nicknames. Given that the merging of two longitudinal files resulted in a longitudinal file, we could determine whether a mother had minor children, the age of the child(ren), and the number of minor children by child birthdate at the date her criminal case was closed.

After merging court and birth records, 37 percent of cases involved a woman who was the mother of at least one minor child. For cases of women with minor children on the disposition date, nearly half (46%) had one minor child, 32% had two minor children, 15% had three minor children and 7% had four or more minor children. It is likely that women had minor children for whom we had no match. When there was no match between ACIS and a birth record, the woman was considered not to have been a mother of a minor child. This could have occurred if the child was born outside of North Carolina or if there was a discrepancy in the mother’s name or birthdate between the court and birth records. Since some women we classified as having no minor children probably had such children, our estimates of the effect of having minor children on criminal justice outcomes are likely to be underestimates.

This study was approved by the Duke University Institutional Review Board. Data use agreements were obtained from each public agency from which data were obtained

Empirical Specification

We defined separate dependent variables for whether: (1) the case resulted in a prosecution; (2) conditional on prosecution, a conviction; (3) conditional on a conviction, an active sentence; and (4) conditional on an active sentence, the length of the minimum active sentence, as specified by the guidelines on the effective charge date.

A binary variable was set to 1 if, in a given case, any charges were prosecuted. A charge was considered prosecuted if the method of disposition for that charge was: “jury trial,” “judge,” “magistrate,” “waiver-clerk,” and “waiver-magistrate.” A charge was considered not prosecuted if the method of disposition was: “capacity dismissed without leave,” “dismissed by court,” “deferred proceeding or deferred prosecution dismissal by DA,” “Habeas Corpus hearing (superior court only),” “capacity dismissed with leave,” “no true bill returned,” “no probable cause,” “dismissal without leave by DA,” and “dismissal with leave by DA.”

A binary variable for being convicted was set to 1 if conditional on prosecution there was a conviction on any charge--if the verdict was recorded as “responsible to lesser” or “responsible” if an infraction (misdemeanor traffic offenses were often reduced to infraction convictions), or “guilty to lesser” or “guilty” if misdemeanor or felony. Conviction outcomes were classified as “not convicted” if the verdict was: “not guilty,” “not responsible,” “prayer for judgment,” and “judgment arrested.”

Plea bargains often resulted in reduced severity of charges. To learn whether the presence of a minor child led to making a reduction in charge severity more likely, we set a binary variable to 1 if the most severe offense class on which a woman was convicted was less than the most severe offense class on which she was charged. For observations with only original misdemeanor offense charges, the dependent variable was 1 if the maximum charge class at conviction was less than the original maximum charge class. If the charge was not reduced during the judicial process, this binary variable was set to 0. If the maximum original charge was a felony, the dependent variable consisted of mutually exclusive categories of (1) no change in convicted class—the omitted reference group, (2) convicted on a lessor felony charge, and (3) convicted on a misdemeanor. When class information on charged and convicted offenses was missing from ACIS records for any charge in a case, the observation was dropped, since reduction of charge could not be determined.

A binary variable for active sentence indicated whether the woman received an active sentence for any convicted offense in a case. Convicted misdemeanants had to serve an active sentence of 90 days or less in a local jail; longer sentences could be served in either a local jail or a state prison (Spainhour & Hall, 2014; Spainhour & Katzenelson, 2014). A person convicted of a felony served an active sentence in state prison (Spainhour & Hall, 2014; Spainhour & Katzenelson, 2014). The binary variable for judge imposed non-mandatory active sentence was set to 1 if the judge chose an active sentence when sanctions other than incarceration were legally permissible.

We used the minimum active sentence length the judge or magistrate assigned to a conviction, a continuous dependent variable measured in days (in NC, Structured sentencing laws required that at minimum this amount of time must be served (Spainhour & Hall, 2014). For women with multiple convictions in different offense classes, the longest minimum sentence length of all convictions was considered to be the minimum active sentence. The ACIS data did not contain information on actual time served in jail or prison. Rather only the offense code and minimum incarceration times were specified.

The key explanatory variable was whether the defendant had a minor child (0–17) on the date the case was resolved. We also included covariates for race/ethnicity defined as mutually exclusive categories for black, white, and other (white, the omitted reference group); the woman’s age at disposition was categorized as ages 16–18, 19–24, 25–34, and 35 years or older (age 35+ was the omitted category). Legal explanatory variables included mutually exclusive binary variables for charge type of all charged offenses resolved on the same date: (1) only misdemeanor non-traffic, (2) both misdemeanor traffic and non-traffic offenses, and (3) any felony with only misdemeanor traffic as the omitted reference category; binary variables for judicial district; and count variables for the number of convictions, during the 3-years preceding the number of misdemeanor convictions and the number of felony convictions. In preliminary analysis, we included binary variables for most severe offense class charged. However, results on the presence of minor children did not differ from those we present.

We included judicial district fixed effects using 2009 definitions of judicial district. In analysis of minimum sentence length, for convictions subject to the structured sentencing law, we included binary variables for the most severe class on which the woman was convicted. In the analysis of having been convicted of a lesser offense, we included binary variables for the most severe class charged in the misdemeanor only sample, and a binary variable of charged with a class G or higher felony was included in the any felony charge model. Examples of class G felonies are second degree burglary, second degree arson, common law robbery, possession of firearms by a felon, and sale of schedule I or II of controlled substance. In all regressions, year fixed effects accounted for time trends in sanctioning. Convictions that received a life or a death sentence were not included in sentence length analysis.

Statistical Approach

When the dependent variable was a binary, we used logit analysis. We present marginal effects evaluated at the observational means. When the dependent variable was continuous, we used a generalized linear model (GLM). GLM is a flexible generalization of ordinary least squares (OLS) that allows for dependent variables with error distributions other than a normal. GLM generalizes OLS by allowing the linear model to be related to the dependent variable by means of a link function which allows the magnitude of the variance of each observation to be a function of its predicted value. Analysis were completed using STATA version 15 (StataCorp, 2018).

Results

Descriptive Statistics

There were 1.4 million criminal cases in North Carolina involving women which had disposition dates from 2009–2012 (Figure 1). Combining cases involving women with more than one charge, 836,384 women in North Carolina were involved in cases that were disposed by the courts between 2009–2012. In 2010 there were 3.25 million women aged 15–64 living in North Carolina (UNC Carolina Demography, 2018). Of these cases, fewer than half (45.3%) were prosecuted on any charge. Of those cases prosecuted, the vast majority led to a conviction on at least one charge (86.8%); very few convictions involved felonies (5.2%). More than half of the cases for which the maximum charge was a misdemeanor were reduced to a lower class of misdemeanor (53.2%); about two-thirds of felonies were reduced to a lesser felony or the initial felony charge was reduced to a misdemeanor at disposition (63.5%). The vast majority of those convicted did not receive an active sentence (95.1%), i.e., were not sentenced to jail or prison.

Figure 1.

Disposition of Cases

Compared to the population of North Carolina, black women were over-represented in the sample (37.3% of our sample versus 22% of the North Carolina population, Table 1) (United States Census Bureau, 2017). Women aged 16–18 were much less likely to have a minor child than older women in our sample. Women under 25 were less likely to have minor children than women at older ages.

Table 1.

Descriptive statistics: Women with and without minor children

Case Attributes (%)	All (N=1,401,003)	No minor child(ren) (n=883,234)	Had minor child(ren) (n=517,769)
Disposition of cases
Prosecuted (n=1,401,003)	45.2	47.3	41.8***
Convicted (n=633,939)	86.8	87.2	86.0***
Active sentence (n=550,285)	4.9	4.4	5.7***
Judge imposed non-mandatory active sentence (n=64,274)	23.6	24.7	22.2***
Convicted of a less serious offensea
All types of charges: Convicted of a less serious offense (n=532,455)	53.7	54.4	52.5***
Only misdemeanor charge: Convicted of a less serious misdemeanor (n=503,775)	53.2	53.9	51.7***
Any felony charge: Convicted of a misdemeanor (n=28,680)	53.1	53.5	52.4
Any felony charge: Convicted of a less serious felony (n=28,680)	10.4	10.4	10.4
Minimum active sentence length categorized by highest convicted offenseb	Mean in days (std. dev.)	Mean in days (std. dev.)	Mean in days (std dev.)
All convictions (n=26,192)	112 (424)	113 (433)	110 (410)
Misdemeanor only convictions (n=22,305)	27 (32)	27 (32)	28*** (32)
DWI convictions (n=789)	176 (219)	174 (219)	179 (221)
Any felony conviction (excluding drug traffic) (n=2,986)	664 (1035)	675 (1057)	647 (1002)
Drug traffic convictions (n=112)	1744 (850)	1741 (953)	1750 (670)
Race/ethnicity
White	51.8	53.4	49.1***
Black	37.3	34.1	42.8***
Other	10.9	12.5	8.1***
Mother’s age
16–18	5.8	8.5	1.2***
19–24	26.0	28.4	21.7***
25–34	34.7	27.9	46.1***
35–55	33.6	35.2	31.0***
Type of charged offenses on disposition date
Only misdemeanor traffic	72.7	71.6	74.6***
Only misdemeanor non-traffic	21.7	23.1	19.2***
Only misdemeanors, traffic and non-traffic	1.2	1.1	1.4***
Any felony	4.4	4.2	4.8***
Previous conviction(s) 1 day to 3 years before disposition
No previous convictions (n=1,401,003)	82.3	85.6	76.7***
Mean number of misdemeanors (std. dev.) (n=248,164)	2.15 (2.53)	2.04 (2.41)	2.27*** (2.64)
Mean number of felonies (std. dev.) (n=248,164)	0.23 (1.53)	0.21 (1.42)	0.26*** (1.65)

^*p<0.05;

^**p<0.01;

^***p<0.001 for t-test comparing to no minor child(ren) sample std=standard

^aDoes not include 17,830 cases missing either charge class or conviction class information

^bDoes not include 10 cases with life sentences or 550 cases missing conviction class information. 532 cases with missing conviction class were for contempt of court.

Included but not shown in independent variables: disposed year, 2009 judicial district, most severe offense class charged of all offenses with the same disposition date.

Women with minor children were 5.5 percentage points less likely to be prosecuted (Table 1)—41.8% versus 47.3%. However, conditional on being prosecuted, the probability of being convicted was only a slightly lower for women with minor children—86.0% versus 87.2%. Conditional on a conviction, a higher percentage of women with minor children were given an active sentence, 5.7% versus 4.4%. However, in the less common situation, when the sentencing judge had a choice between active sentence or other punishment type, women with minor children were 2.5 percentage points less likely to receive an active sentence—22.2% versus 24.7%.

Regression Results for Prosecution and Conviction

Controlling for the other factors, having minor children reduced the probability of prosecution by 0.065 (Table 2); the same as the difference unadjusted for other factors. Relative to the observational mean (45.2% prosecuted, Table 1), this represents a 14.4 percent reduction for having at least one minor child.

Table 2.

Effect of Having a Minor Child(ren) for Women’s Case outcomes

	Beta	Std. Error	Marginal Effect	95% CI for Marginal effects
Probability of being prosecuted and convicted
Prosecuted (n=1,401,003)	−0.262***	(0.004)	[−0.065]	[−0.067, −0.063]
Convicted conditional on being prosecuted (n=633,939)	−0.151***	(0.008)	[−0.014]	[−0.016,−0.013]
Probability of being convicted of a lesser charge
Convicted to a lesser misdemeanor1 (n=503,775)	−0.021**	(0.007)	[−0.005]	[−0.009,−0.002]
Any felony charge: Convicted of a lesser felony2 (n=28,680)	0.027	(0.049)	[0.002]	[−0.004,0.007]
Any felony charge: Convicted of a misdemeanor2 (n=28,680)	−0.004	(0.028)	[−0.002]	[−0.015,0.011]
Probability of active sentence
Active sentence3 (n=550,285)	0.002	(0.015)	[0.000]	[−0.000,0.001]
Active sentence given judge had a choice4 (n=64,274)	−0.141***	(0.020)	[−0.024]	[−0.031,−0.017]
Active Sentence Length Conditional on Receiving an Active Sentence
If woman had only misdemeanor convictions (n=22,305)	0.011	(0.017)	[0.249]	[−0.500,0.999]
If woman had any felony convictions (n=2,986)	0.013	(0.035)	[5.256]	[−22.635,33.148]

^*p<0.05;

^**p<0.01;

^***p<0.001 Std=standard

¹Sample limited to cases with only misdemeanor offenses;

²Sample limited to cases with any felony offenses;

³Sample limited to cases with women who had a conviction;

⁴Sample limited to cases with women who were convicted of an offense that the judge had choice to impose a non-mandatory active sentence or other punishment.

Notes: Models were estimated using logistic regression (probability of being prosecuted, probability of being convicted, probability of being convicted of a lesser charge for misdemeanors, probability of active sentence), multinomial logistic regression (probability of being convicted of a lesser charge for felonies, omitted not dropped to a lesser charge), and generalized linear model (active sentence length). For models of probability of being convicted of a lesser charge, 17,830 women-disposed-dates were excluded from the regression due to missing information on charge and/or conviction class. For models of active sentence length, ten women-disposed-dates with death or life sentences were excluded, 550 women-disposed-dates were excluded from the regression due to missing information on conviction class.

All models controlled for race/ethnicity (white (omitted), black, other race/ethnicity), woman’s age (16–18,19–24,25–34, 35–55 (omitted)), number of previous misdemeanor convictions, number of previous felony convictions, disposed year fixed effects, and 2009 judicial district fixed effects. Probability of being prosecuted and convicted and probability of active sentence models included type of offenses charged (only misdemeanor traffic (omitted), only misdemeanor non-traffic, only misdemeanors (traffic and non-traffic), any felony). The models for probability of being convicted of a lesser charge included most severe offense class charged (Class 3 (omitted), Class 2, Class1, Class A1) for misdemeanor model, and higher than an H felony for felony model. The models for active sentence length included most severe offense class convicted: for misdemeanor model (Class 3 (omitted), 2, 1, A1), for felony model (Class I (omitted), H, G, F, E, D, C, B2, B1). Class A felonies are death or life sentences and were excluded.

Much of the accounting for the presence of a minor child was done by prosecutors. Conditional on being prosecuted, having a minor child reduced the probability of conviction by 1.4 percentage points, slightly more than the corresponding unadjusted result in Table 1. This result implies that having a minor child had only a very small effect on the probability of conviction—relative to the observational mean, a 1.6 percent reduction.

In separate analysis (see Appendix), we disaggregated the minor children variable into binary variables for whether the woman had any minor children aged 0–3, 4–6, 7–10, and 11–17 at the disposition date. All parameter estimates were statistically significant relative to no minor children, the omitted reference category. Some parameter estimates for the child age group covariates were statistically significantly different from one another but the magnitude of the implied differences was small.

Regression Results for Being Convicted of a Less Serious Offense than the Original Charge

When only charged with misdemeanors, the probability that there would be a reduction in the most severe misdemeanor at conviction was about the same for women with and without minor children (Table 2). There were no statistically significant differences in charge reduction when the original charges included a felony. Both the number of previous misdemeanors and the number of felony convictions were related to a reduced probability of being convicted of a lesser misdemeanor and a reduced probability of having a felony charge reduced to a misdemeanor (results not shown).

Regression Results for Minimum Active Sentence

For women who were convicted of an offense, the presence of minor children was not associated with a reduction in the probability of receiving an active sentence (Table 2). However, when judges had the discretion to impose of an active sentence given the nature of the offense, women with minor children were 0.024 less likely to receive an active sentence. There were no statistically significant results for having minor children on active sentence length (Table 2).

Discussion

Our study yielded four major findings. First, having minor children reduced the probability that criminal charges were prosecuted and, conditional on being prosecuted, the probability that the defendant was convicted. However, if convicted, the probability that charges were reduced at conviction was not lower for women with minor children. Second, when the judge had discretion in imposing an active sentence, having minor children resulted in a reduction in the probability of an active sentence. Third, conditional on having an active sentence, women with minor children had the same active sentence length as women without minor children. Fourth, the age of the minor children had at most a small effect on the likelihood of prosecution or conviction.

Most of the reduction in criminal sanctions attributable to motherhood stemmed from prosecutors’ decisions. Judges accounted for the presence of children in deciding on whether to convict and whether the woman was to serve an active sentence when the sentencing guidelines permitted a choice of whether an active sentence was to be served. However, judges only had such discretion in a minority of cases.

Prosecutorial discretion has a large effect on case processing outcomes following a charge. Thus, it is important to analyze the process from filing of initial criminal charges through conviction, rather than focus on conviction alone which has been the focus of many previous studies. From a societal perspective, this study shows that even with sentencing guidelines, which are designed to reduce the importance of extra-legal factors in sentencing, the judicial system had some flexibility in taking account of maternal status in determining criminal sanctions.

Our finding regarding the importance of the role of prosecutorial discretion in court outcomes, given widespread implementation of sentencing guidelines by states, is not in itself new (see e.g., Kramer & Ulmer, 2009, and studies cited therein). A rationale for such guidelines was to reduce the importance of extra-legal factors in sentencing. Included in such factors is presumably the presence of minor children although the effects of guidelines per se on sentencing of mothers has been a neglected research area.

With stricter sentencing guidelines in place, the importance of prosecutorial discretion has grown as judges are limited in their sentencing options to state mandated guidelines (Bibas, 2004; Stuntz, 2004). Prosecutorial discretion can be seen both in the number of cases that are prosecuted and the percent convicted once prosecuted. For one, prosecutors are likely not to invest as much in cases in which conviction is unlikely. In our sample, more than half of the cases involving women were dropped in that no charges were pursued by the prosecutor, but for those that were pursued, the conviction rate was extremely high. One mechanism for securing convictions is for prosecutors to offer defendants a less serious offense (e.g., dropping the charge from a felony to misdemeanor) in exchange for a guilty plea. This proves to be true in our sample as well with over 50 percent having a conviction for a less serious offense.

With the growing importance of prosecutorial discretion, choosing to decline prosecution of a case because the woman has a minor child has implications for child well-being as well as for the criminal justice system more generally. The lower rate of prosecution of these women may reflect effective arguments made by defense counsel that the harm done to the child(ren) by prosecuting a parental crime exceeds the harm of not prosecuting the crime at all. Of course, there may be other factors at play such as lack of sufficient evidence to convict. The fact that women with children were more likely to have a conviction on a lesser charge also suggests that prosecutorial discretion is used to offer lower charges to women who are caretakers of children.

We also documented an effect of a minor child on sentencing when judges had a choice based on flexibility in sentencing guidelines. Being the caretaker of a minor child, especially the sole caretaker of a minor child, serves as a mitigating factor when determining a sentence. Although grandparents and other family members are often caretakers following incarceration of a female parent (Turney, 2014b), our administrative data did not include such information.

Accounting for the presence of a child in the home or the fact that woman may be the sole caretaker child has health implications for both the child and the family. Discretion showing acknowledgment of the child at prosecution, conviction, and sentencing can directly affect the wellbeing of the child. Previous studies have reported that children with an incarcerated parent are at risk for physical and mental health declines, early substance use initiation, criminal activity, being exposed to maltreatment or harsh parenting, and poor school outcomes (Haskins, Amorim, & Mingo, 2018; Turney & Goodsell, 2018).

Most of the research that has examined the intergenerational consequences of parental criminal involvement has focused on parental incarceration (Kirk & Wakefield, 2018). However, incarceration represents a small share of the experiences that criminally involved parents face; A much larger number of women are arrested and face prosecution—the act of being arrested alone has been linked to negative child outcomes (Jackson & Vaughn, 2017; Turney, 2014b; Wildeman, 2012; Wildeman et al., 2018). If one is convicted of a severe penalty such as a felony—certain rights are lost, such as voting rights and eligibility for pubic transfers. Even someone convicted of a misdemeanor, may be barred from employment in certain occupations (Jain, 2015). We do not know what the costs of lesser sanctions are.

To our knowledge, ours is the first study to gauge the effect of minor children on the disposition of criminal cases from the date criminal charges were initially filed through sentencing. The reason for a lack of studies is there are many ethical and privacy barriers to merging administrative datasets, in particular when minor children are involved. Because we were able to overcome these barriers and link information from the court and birth records, the North Carolina data provided a unique opportunity for this empirical analysis. Many previous studies of effects of minor children on sentencing outcomes were based on much smaller samples than ours and applied to jurisdictions below the state level (e.g., (Cho & Tasca, 2018; Freiburger, 2011; Stacey & Spohn, 2006)). Studies based on the U.S. Sentencing Commission data, which had larger samples, had to rely on whether or not the person listed dependent(s) but did not have information as to whether or not the dependent was a child or whether the person had a child that she was unable to claim as a dependent (Doerner, 2012; Doerner & Demuth, 2014).

Ideally, public policies would be set at levels that minimize the total cost of crime, including costs to victims, law enforcement agencies, other public agencies that deal with crime victims and offenders, as well as relatives of offenders. Establishing the optimal levels and mix of public policies would require substantial additional empirical analysis. This task is further complicated by consideration of equity issues. For example, how much weight should be placed on benefit to family members of offenders versus the value of a safe neighborhood to potential crime victims? Thus, we cannot know whether the current reductions in criminal sanctions documented in this study are excessive or inadequate. If the reductions were deemed to be suboptimal in either direction, it might be appropriate to adjust sentencing guidelines when the convicted offender is a parent. However, these adjustments would be made by judges, not prosecutors. If for example, a woman was given a more lenient sentence because she is a mother, this would be accounted for by judges. In response, prosecutors might treat cases involving mothers of minor children less leniently.

We acknowledge several study limitations. First, we could link mothers to their biological and adopted children, but we could not measure whether the child lived with the biological/adoptive mother, whether the child was otherwise dependent of the mother for financial support, and whether the child had other caregivers, such as a father, grandparent, other relative, or foster parent.

Second, our analysis was based on data from a single state. We know from past research that there are regional differences in sentencing outcomes (Holland & Prohaska, 2018) and such outcomes are dependent on the political context (Kim, Wang, & Cheon, 2018).

Third, our analysis was limited to a state with sentencing guidelines in place throughout the observational period. Conceptually, since guidelines are designed to limit judicial discretion, if anything, effect sizes of being a mother of minor children should have been larger effect in states without guidelines. Other differences, such as those attributable to effects of sentencing guidelines on racial disparities in criminal penalties, that have been documented have been small (Bushway & Forst, 2013), but, to our knowledge, there is no empirical evidence on the effects of having minor children post versus pre-implementation of guidelines.

Fourth, the court records only allowed for a limited accounting of the facts of each case, and since court records were only available from 2005 and forward, we could not account for defendants’ entire criminal history. Even if we had data to allow us to calculate an individual-specific criminal history, the lookback would only include crimes that occurred in North Carolina. Some individuals would undoubtedly have been convicted of crimes in other states. Such data may be available from the courts in print form but such records are not available electronically. The direction of omitted variables bias of the parameter estimates for minor children depends on the correlation of the minor children covariate with the omitted variable.

Fifth, with court data, we could determine minimum and maximum active sentence length, but we were unable to determine the actual sentence served. Thus, we were unable to determine whether mothers were less likely to receive more lenient sentences, holding other factors constant. We could only measure whether motherhood led to lower minimum sentences, which would have been fully determined by the charge(s) on which she was convicted.

Information on actual time served is not available from the court records, but can be found in jail and prison records. The vast majority of incarcerations occur in jails, and such data in North Carolina are maintained by individual counties, and there is considerable heterogeneity in the way that such records are maintained. Some women served more than their minimum sentence or received special sentence conditions such as serving on weekends only. Merging court and jail or prison records is a tedious process; as a result, studies that have accomplished this merge have included relatively few observations from small geographic areas, such as single county. Fifth, we were able to identify women with minor children, but unable to exclude the possibility that some women whose court records did not match with birth records were in fact mothers of minor children. To the extent this is so, this implies the differences we have documented are plausibly underestimates of the true effects of having a minor child.

In sum, while previous studies have compared sentencing between males and females convicted of a criminal offense and attributed the difference, at least in part, to recognition by the judicial system that child-rearing primarily remains the responsibility of mothers, this study, using administrative data, found empirical support for the hypothesis that women with minor children who are charged with a criminal offense do receive somewhat more lenient criminal penalties.

Appendix

Table A1.

Women with at least one minor child on date of disposition, by child age-group (%)

At least 1 child aged (n=517,769):	%
0–3 years	46.7
4–6 years	34.9
7–10 years	34.8
11–17 years	34.3

Note: Categories are not mutually exclusive so totals may sum to more than 100%

A2.

Probability of being prosecuted and convicted: Effect of minor child by age group

	Prosecuted	Convicted, conditional on being prosecuted
Any child aged 0–3 on disposition date	−0.174***	−0.072***
	(0.005)	(0.011)
	[−0.043]	[−0.007]
	<−0.045, −0.041>	<−0.009, −0.005>
Any child aged 4–6 on disposition date	−0.133***	−0.108***
	(0.006)	(0.012)
	[−0.033]	[−0.010]
	<−0.036, −0.030>	<−0.013, −0.008>
Any child aged 7–10 on disposition date	−0.115***	−0.094***
	(0.006)	(0.013)
	[−0.029]	[−0.009]
	<−0.031, −0.026>	<−0.011, −0.007>
Any child aged 11–17 on disposition date	−0.160***	−0.092***
	(0.006)	(0.013)
	[−0.040]	[−0.009]
	<−0.042, −0.037>	<−0.011, −0.006>
n	1,401,003	633,939

^*p<0.05;

^**p<0.01;

^***p<0.001

Notes: Models were estimated using logistic regression. Models controlled for race/ethnicity (white (omitted), black, other race/ethnicity), woman’s age (16–18,19–24,25–34, 35–55 (omitted)), number of previous misdemeanor convictions, number of previous felony convictions, type of offenses charged (only misdemeanor traffic (omitted), only misdemeanor non-traffic, only misdemeanors (traffic and non-traffic), any felony), disposed year fixed effects, and 2009 judicial district fixed effects.

Table A3.

Probability of being convicted of a lesser charge: Effect of minor child by age group

	Convicted to a lesser misdemeanor Beta, (std. error), [marginal effect, <95% CI>]	Any felony charge: Convicted of a lesser felony Beta, (std. error), [marginal effect, <95% CI>]	Any felony charge: Convicted of a misdemeanor Beta, (std. error), [marginal effect, <95% CI>]
Any child aged 0–3 on disposition date	−0.122***	−0.057	0.005
	(0.009)	(0.065)	(0.037)
	[−0.031]	[−0.004]	[0.003]
	<−0.035, −0.026>	<−0.011, 0.004>	<−0.014, 0.02>
Any child aged 4–6 on disposition date	−0.003	0.161*	0.067
	(0.011)	(0.069)	(0.041)
	[−0.001]	[0.007]	[0.011]
	<−0.006, 0.005>	<0.00, 0.015>	<−0.008, 0.030>
Any child aged 7–10 on disposition date	0.019	0.076	−0.020
	(0.011)	(0.070)	(0.041)
	[0.005]	[0.005]	[−0.008]
	<−0.001, 0.010>	<−0.003, 0.013>	<−0.027, 0.011>
Any child aged 11–17 on disposition date	0.036**	−0.084	−0.093*
	(0.011)	(0.068)	(0.039)
	[0.009]	[−0.002]	[−0.020]
	<0.004, 0.014>	<−0.009, 0.006>	<−0.038, −0.002>
n	503,775	28,680	28,680

^*p<0.05;

^**p<0.01;

^***p<0.001

Notes: Models were estimated using logistic regression (convicted to a lesser misdemeanor) and multinomial logistic regression (probability of being convicted of a lesser charge for felonies, omitted not dropped to a lesser charge). Models controlled for race/ethnicity (white (omitted), black, other race/ethnicity), woman’s age (16–18,19–24,25–34, 35–55 (omitted)), number of previous misdemeanor convictions, number of previous felony convictions, most severe offense class charged (Class 3 (omitted), Class 2, Class1, Class A1) for misdemeanor model, and higher than an H felony for felony model, disposed year fixed effects, and 2009 judicial district fixed effects. For models of probability of being convicted of a lesser charge, 17,830 women-disposed-dates were excluded from the regression due to missing information on charge and/or conviction class.

Table A4.

Probability of active sentence: Effect of minor child by age group

Covariates	Active sentence Beta, (std. error), [marginal effect, <95% CI>]	Active sentence given judge had choice Beta, (std. error), [marginal effect, <95% CI>]
Any child aged 0–3 on disposition date	0.015	−0.087**
	(0.019)	(0.027)
	[0.000]	[−0.015]
	<−0.000, 0.001>	<−0.024, −0.006>
Any child aged 4–6 on disposition date	0.103***	0.020
	(0.021)	(0.028)
	[0.002]	[0.004]
	<0.001, 0.002>	<−0.006, 0.013>
Any child aged 7–10 on disposition date	−0.033	−0.084**
	(0.022)	(0.028)
	[−0.001]	[−0.015]
	<−0.001, 0.000>	<−0.024, −0.005>
Any child aged 11–17 on disposition date	−0.028	−0.098***
	(0.021)	(0.027)
	[−0.000]	[−0.017]
	<−0.001, 0.000>	<−0.026, −0.008>
n	550,285	64,274

^*p<0.05;

^**p<0.01;

^***p<0.001

Notes: Models were estimated using logistic regression. Models controlled for race/ethnicity (white (omitted), black, other race/ethnicity), woman’s age (16–18,19–24,25–34, 35–55 (omitted)), number of previous misdemeanor convictions, number of previous felony convictions, type of offenses charged (only misdemeanor traffic (omitted), only misdemeanor non-traffic, only misdemeanors (traffic and non-traffic), any felony), disposed year fixed effects, and 2009 judicial district fixed effects.

Table A5.

Active Sentence Length Conditional on Receiving an Active Sentence: Effect of minor child by age group

	Only misdemeanor convictions Beta, (std. error), [marginal effect, <95% CI>]	Any felony conviction Beta, (std. error), [marginal effect, <95% CI>]	Any DWI conviction Beta, (std. error), [marginal effect, <95% CI>]
Any child aged 0–3 at disposed date	0.003	−0.036	−0.103
	(0.022)	(0.041)	(0.126)
	[0.078]	[−14.746]	[−10.468]
	<−0.896, 1.051>	<−47.448, 17.957>	<−35.484, 14.547>
Any child aged 4–6 at disposed date	−0.032	−0.065	−0.061
	(0.023)	(0.043)	(0.144)
	[−0.717]	[−26.433]	[−6.168]
	<−1.732, 0.298>	<−61.013, 8.147>	<−34.673, 22.337>
Any child aged 7–10 at disposed date	0.024	−0.019	−0.063
	(0.025)	(0.053)	(0.117)
	[0.552]	[−7.693]	[−6.425]
	<−0.538,1.643>	<−49.849, 34.464>	<−29.620, 16.771>
Any child aged 11–17 at disposed date	0.037	0.051	0.094
	(0.022)	(0.058)	(0.097)
	[0.834]	[21.026]	[9.572]
	<−0.142, 1.811>	<−25.074, 67.126>	<−9.801, 28.946>
n	22,305	2,986	789

^*p<0.05;

^**p<0.01;

^***p<0.001

Notes: Models were estimated using generalized linear model (active sentence length). Models controlled for race/ethnicity (white (omitted), black, other race/ethnicity), woman’s age (16–18,19–24,25–34, 35–55 (omitted)), number of previous misdemeanor convictions, number of previous felony convictions, most severe offense class convicted: for misdemeanor model (Class 3 (omitted), 2, 1, A1), for felony model (Class I (omitted), H, G, F, E, D, C, B2, B1), disposed year fixed effects, and 2009 judicial district fixed effects. Class A felonies are death or life sentences and were excluded. Ten women-disposed-dates with death or life sentences were excluded.