Does the criminal past of parents affect educational attainment of their children?

Iryna Rud

Chris Van Klaveren

Wim Groot

Henriette Maassen van den Brink

All authors are affiliated with The Top Institute for Evidence Based Education Research, TIER, Maastricht University, P.O. BOX 616, 6200 MD Maastricht, The Netherlands.

E-mail address of the corresponding author:

Abstract

There is a large body of empirical evidence that establishes the relationship between education and crime, and the intergenerational transmission of both criminal behavior and educational attainment. However, evidence on how criminal involvement of parents affects the educational outcomes of their children is limited to studies on relatively small samples of imprisoned populations. In our study, we use an extensive dataset which allows us to control for a wide range of individual and family characteristics, including educational attainment of parents. For this purpose, we apply a multinomial probit model on a matched sample.

We find that the probability of higher educational attainment is lower for children whose parents were ever criminally involved. In particular, the probability of obtaining a higher education degree decreases by eight percentage points if parents have a criminal past. This underlines that growing up in an unstable family environment can have a negative impact on the educational attainment of children, indicating that the effects of crime is extensive and long-lasting.

JEL Codes: I21, J13, J24, K14

Keywords: educational attainment; criminal involvement; intergenerational effects

1  Introduction

It is widely recognized that crime involves different types of social costs, related to security measures, property loss, victim services and so forth. In addition, it can entail intangible costs such as anxiety, inconvenience or suffering of close family members of criminals. For example, children of criminal parents are influenced by an unstable social environment in their family, and as a result they can become involved in criminal activity themselves or get school-related problems. While the intergenerational transmission of criminal behavior from parents to their children has been strongly established (e.g. Van de Rakt et al., 2010), only a few empirical studies have examined the effects of criminal involvement of parents on educational outcomes of children (see review by Murray et. al, 2012). Those studies do not fully control for important characteristics that could be related to both criminal involvement of parents and the educational attainment of the children, such as parental education levels, ethnicity, single parent household, household income and socio-economic status of parents (Thornberry and Farnworth, 1982; Chevalier et al., 2005).

In this study, we aim to examine the impact of criminal involvement of parents on educational attainment of their children. The main advantage of our study over previous research is that we take into account all those different variables that can correlate with the criminal status of parents and affect education of children. To do so, we first examine the effect of parents’ criminal involvement on educational attainment of children, applying a multinomial probit model which includes a number of individual characteristics as controls. Secondly, as there are systematic differences between children whose parents have engaged in criminal activities and those who have not, we apply matching methods which allow us to eliminate these differences and reduce reliance on the model-based adjustments.

We first group observations on educational attainment level of their parents and gender of parents to provide an exact match on these characteristics, since parental education is considered as a strong predictor of educational attainment of children (De Haan and Plug, 2009; Holmlund et al., 2011). Then, inside of each group we use the Mahalanobis distance method (Rubin, 1980) to match on a set the demographic characteristics of children and on household characteristics. As a result of matching each child whose parents without criminal involvement serves as a control for a child whose parents were criminally involved. We estimate the variable of our interest, educational attainment of children with the same multinomial probit model for the matched sample, and then compare the estimation results to those for the whole sample. Finally, we provide a series of robustness checks including controls for unfinished educational attainment spells and family circumstances, and redefining the outcome variable into years of education.

We believe that our study contributes to the literature in several ways. First, this study contributes to the research on the intergenerational effects of educational attainment and criminal behavior from parents to children. Most previous studies consider the intergenerational effects of either criminal behavior or educational attainment from parents to children (e.g. Holmlund et al., 2011; Hjalmarsson and Lindquist, 2011), or the relationship between education and crime on the individual level (e.g. Lochner and Moretti 2004; Machin et al., 2011). Second, in order to analyze the effects of criminal past of parents on educational attainment of their children, we make use of a large sample of the population from a nationally representative survey conducted in the Netherlands, the ‘Netherlands Kinship Panel Study’, whereas previous research is primarily done using an imprisoned population sample from the United States (e.g. Trice and Brewster, 2004; Parke and Clarke-Stewart, 2002). The data we use contain detailed information about individuals and their households, and therefore, allow us to apply matching techniques to separate the effect of criminal involvement of parents from other decisive effects that can shape the educational attainment of children.

The paper proceeds as follows. In section 2 we review the empirical literature relevant to our study. Section 3 explains the process of data preparation and reports descriptive statistics. Section 4 elaborates on the estimation strategy and reports the results of the multinomial probit analyses and the matching models and provides robustness checks. We conclude with a discussion of the implications of our findings and suggestions for future research.

2  Literature

The underlying assumptions in our study are based on the literature analyzing the relationship between education and crime, and on studies discussing the intergenerational transmission of educational attainment or criminal behavior from parents to their children. In particular, the association between education and crime has been well-documented in the economic literature (e.g. Lochner and Moretti 2004; Machin et al., 2011; Groot and Maassen van den Brink, 2010; Meghir et al., 2011). A growing number of empirical studies find that there is a negative relationship between educational attainment and criminal behavior. Only white collar crime, which usually is related to a certain level of income or a particular social status, is likely to be positively associated with educational attainment (Lochner, 2004; Groot and Maassen van den Brink, 2010). However, this type of crime is usually underrepresented in criminal statistics (Lynch et al., 2004). The above mentioned studies highlight that children of criminally involved parents are often those who have lower educated parents. Parental education is, perhaps, the most explicit characteristic that can be related to both criminal status of parents and education of children at the same time, but there might be many other factors (social environment, socioeconomic status of parents, size of the family, etc.), which are important to take into account while analyzing the effect of criminal involvement of parents on education of their children.

A considerable literature has focused on the intergenerational effects from parental background on their children outcomes. In particular, some studies have established a causal link between educational attainment of parents and their children (De Haan and Plug, 2009; Holmlund et al., 2011; Ermisch and Pronzato, 2010) and others have provided evidence on the transmission of criminal behavior from parents to their children (Farrington 2002; Van de Rakt et al., 2010; Hjalmarsson and Lindquist, 2011). However, little is known about the impact of criminal behavior of parents on individual outcomes of children. Research shows that parental involvement in crime has a negative influence on family stability and financial circumstances, and, as a result, the general well-being of children (Roguski and Chauvel, 2009; Travis and Waul, 2003). Children of criminal parents are more predisposed to delinquent behavior and emotional difficulties (Geller et al., 2009; Murray and Farrington, 2005; Wilbur et al., 2007) and they are often stigmatized by their peers at school (Gadsden and Jacobs, 2007; Nesmith and Ruhland, 2008). This already implies that an unstable social environment in the family can negatively influence children, also in terms of their study.

The effects of criminally involved parents on educational attainment of children have been primarily examined using imprisoned population data, and the results suggest negative effects with regards to children’s academic performance, school related problems, and failing classes (Gordon, 2009; Murray and Farrington, 2008; Neal, 2009; Sarri et al., 2011). Separation from the mother due to incarceration might not be the same as separation from the father. A number of studies analyze particularly the relationship between mothers’ imprisonment and educational outcomes of their children, such as academic performance, failing classes and dropping out of school, and find a negative association (Stanton, 1980; Henriques, 1982; Trice and Brewster, 2004). Moreover, the study of Trice and Brewster (2004) suggests that high school dropout is related not only to the imprisonment status of mothers but also to their mothers’ educational attainment. To our knowledge, a single study done on imprisoned fathers shows a negative effect on academic performance of children as well (Hagan and Foster, 2009).

We can conclude these studies that parents’ criminal involvement can have a negative impact on individual outcomes of children which can also lead to school-related problems. Studies on the imprisoned parents’ populations show direct negative effects on educational attainment of children. However, the existing research is mainly done using relatively small samples of imprisoned populations and applying a narrow set of covariates. To our knowledge, no empirical study analyses the effects of criminal involvement of parents on educational attainment of children from a general population sample. We believe that our study can bridge this gap by providing an analysis on a large sample of general population data and by accounting for those important differences in covariates between children whose parents were criminally involved and those whose parents were not.

3  Data and descriptive statistics

In this study, we use the Netherlands’ Kinship Panel Study (NKPS), a nationally representative two-wave survey held from2002 to 2004 and from 2006 to 2007 among 8,155 residents of the Netherlands aged 18 to 79 (Dykstra et al., 2004). The NKPS data were collected through self-reported questionnaires. In our study, we can only exploit the first wave of the NKPS only because this wave contains information on the criminal involvement status of parents and on the educational attainment of parents and their children. It also contains demographicand socio-economiccharacteristics of respondents and their closest family members, in particular their spouses and children. Respondents were asked to report about their children aged 15 and older, with a maximum of two children of their choice if they have more than two. First, we exclude observations which do not provide necessary information for the analysis, such as respondents without children or missing values of the educational or criminal status variables. Then, we recode the dataset in a new form, where each observation is a child of the respondent with information available on individual characteristics and parental characteristics (i.e. respondents and respondents’ spouses). As a result, we have for our usage 5,822 observations in total.

In the NKPS study, respondents were asked if they or their family members have ever been convicted in a criminal court or have ever been in contact with police for delinquent behavior, other than for traffic offences (Dykstra et al., 2004; Dykstra and Valk, 2007). Based on this information we create a dummy variable that takes a value of 1 if at least one parent in the household has ever been arrested and/or convicted and the value of the dummy is 0 if parents do not have criminal involvement. There are 242 children with either one or two criminally involved parents in our data, 97 of whom were convicted.

Educational attainment is recoded into five categories, reflecting the current system of education in the Netherlands. In the NKPS study, educational attainment refers to the highest level of education that an individual completed with a diploma (Dykstra et al., 2004). In Table 1, we report the descriptive statistics of several demographic characteristics of parents and their children, and characteristics of their households. The first column shows summary statistics for the group of persons whose parents have no criminal past (N=5,580), and the second column reports descriptive information for the group of persons whose parents have had criminal involvement in the past (N=242). Besides information on educational attainment, Table 1 shows variables on the ethnic origin of the parents, gender of parents by criminal involvement status, age of parents, marital status of the parents, income, a household size, number of children in the family, gender and age of the child, and urbanization level of the residence area (via a dummy variable).

Table 1. Descriptive statistics

Characteristics of children and household characteristics / No criminal involvement of parents / Criminal involvement of parents / Differences of means
Mean / Std. Dev. / Mean / Std. Dev. / ∆ / Std. error
Parents born in the Netherlands / 0.98 / 0.11 / 0.96 / 0.18 / 0.02*** / 0.01
Gender of parents by criminal involvement status (mother if 1) / 0.60 / 0.49 / 0.32 / 0.47 / 0.27*** / 0.03
Age of the parent / 59.14 / 10.14 / 55.63 / 8.96 / 3.51*** / 0.66
Single-parent household / 0.16 / 0.25 / 0.16 / 0.26 / -0.01 / 0.02
Income (Euro) / 1497.46 / 1253.21 / 1512.39 / 980.12 / -14.93 / 81.62
Number of household members / 2.32 / 1.21 / 2.71 / 1.44 / -0.39*** / 0.08
Number of children / 2.71 / 1.17 / 2.66 / 1.33 / 0.04 / 0.08
Girls / 0.50 / 0.50 / 0.45 / 0.50 / 0.05* / 0.03
Age (child) / 31.92 / 10.28 / 28.32 / 9.58 / 3.60*** / 0.67
Urbanized residence area / 0.63 / 0.48 / 0.63 / 0.48 / 0.01 / 0.03
Parents’ educational attainment
(5 categories)
(1) Primary Education / 3.10
0.13 / 1.45
0.34 / 2.87
0.23 / 1.54
0.42 / 0.23***
-0.10** / 0.10
0.02
(2) Lower Secondary Education / 0.36 / 0.48 / 0.32 / 0.47 / 0.04 / 0.03
(3) General Secondary Education / 0.07 / 0.25 / 0.03 / 0.18 / 0.03** / 0.02
(4) Intermediate Vocational Education / 0.18 / 0.38 / 0.18 / 0.39 / -0.01 / 0.03
(5) Higher Education / 0.27 / 0.44 / 0.24 / 0.43 / 0.03 / 0.03
Children’s educational attainment
(5 levels)
(1) Primary education / 0.35
0.11 / 1.40
0.32 / 2.77
0.21 / 1.37
0.41 / 0.58***
-0.10*** / 0.09
0.02
(2) Lower Secondary Education / 0.23 / 0.42 / 0.29 / 0.45 / -0.06** / 0.03
(3) General Secondary Education / 0.15 / 0.36 / 0.14 / 0.35 / 0.01 / 0.02
(4) Intermediate Vocational Education / 0.21 / 0.41 / 0.22 / 0.41 / -0.01 / 0.03
(5) Higher Education / 0.29 / 0.46 / 0.14 / 0.34 / 0.16*** / 0.03
Observations by criminal involvement status / 5580 / 242
Total / 5822

*/**/*** denotes statistical significance at the 10/5/1 percent level, respectively.