“I Suck at Everything”: Crime, Arrest, and the Generality of Failure

Abstract

In this paper we advance the idea that getting arrested amounts to “failing at crime.” And akin to the notion of the generality of deviance—where those who engage in any given form of criminal behavior are also likely to engage in a wide array of other problematic behaviors—we examine whether failing at crime (getting arrested) is associated with other forms of life failure. Using data from multiple waves of the National Longitudinal Study of Adolescent to Adult Health, our results reveal that, independent of one’s level of self-reported criminal behavior and other key potential confounders (IQ and self-control), being arrested is a significant predictor of a host of life failures related to education, employment, relationships, and health. The key implication of our study is that it highlights the need to develop a theory of the “generality of failure.”

Most crimes are pretty easy to commit (Gottfredson and Hirschi, 1990). This fact might seem disappointing when compared to popular portrayals like the highly cohesive group of daring professional thieves pulling off a complex heist that made for a good movie in Ocean’s Eleven, or the genius methamphetamine kingpin who consistently outsmarted the competition in television’s Breaking Bad. But the far less glamorous reality is that it requires little skill or planning to swipe someone’s iPhone; no acquired expertise is needed to knock a mailbox off of its perch with a baseball bat from the back of a fast-moving El Camino; and no staggering intellect is necessary for big people to hit little people (Felson, 1996).

So if committing crime is not all that hard to do, it might logically follow that one should find it is pretty easy to get away with most criminal activity. This idea is confirmed by research demonstrating that clearance rates for a wide variety of crimes—even really serious ones like armed robbery and homicide—are far from 100 percent (Doerner and Doerner, 2012; Roberts, 2014). Even for crimes like driving while intoxicated—an offense that carries extremely stiff penalties all around the country (Yao, Johnson, and Beck, 2014)—research has indicated that the odds of getting caught can be upwards of 1 in 200 (Beitel, Sharp, and Glauz, 2000). Put simply, that someone would successfully commit a crime in this country should not come as a big shock to anyone.

But of course, not everybody gets away with it. People doing bad things do, in fact, get caught all the time (Barnes, 2014). In this sense, getting arrested for one’s criminal behavior can be seen as a form of failure. Indeed, unlike the heroin dealer whose transaction goes undetected, the burglar who gets arrested has failed at crime. And if we were to view arrest as a form of failure, it begs the question of whether this kind of failure is somehow unique. It probably isn’t. If someone fails at something as simple as crime, it is entirely possible that they will fail at things that are a lot more difficult to do like getting and keeping a job or maintaining a stable relationship. If that is the case, the long-standing notion of the “generality of deviance” (Hirschi and Gottfredson, 1994)—which describes the tendency for criminals to also engage in a host of sketchy and questionable behaviors (Osgood et al., 1988; Reisig and Pratt, 2011; Turanovic and Pratt, 2014)—might be extended to include the “generality of failure.”

To address this issue, we examine the following question: independent of one’s level of self-reported offending (as well as key potential confounding variables), does getting arrested—conceived of as failing at crime—predict a range of life failures related to education (e.g., having to repeat a grade; not graduating from high school), employment (e.g., unemployment, job instability, getting fired), social relationships (e.g., never getting married, getting divorced), and health (e.g., getting a sexually transmitted disease, abusing alcohol)? To do so, we analyze multiple waves of data from the National Longitudinal Study of Adolescent to Adult Health (Add Health, see Harris, 2009). Our broader purpose is to shed light on the utility of thinking about criminal behavior—and the arrests that can stem from it—in the context of a wider set of negative life outcomes.

THE GENERALITY OF DEVIANCE AND FAILURE

Historically, the field of criminology has been rather fond of offering up various kinds of typologies for offending and offenders. A short list of these would include those created to categorize different kinds of juvenile offenders (Cohen and Short, 1958) and different youthful offending trajectories (Moffitt, 1993); different kinds of adult offenders depending on the crimes they committed (e.g., violent versus property offenses, see Gibbons, 1965); and even specific typologies for sex offenders (Lussier et al., 2012) and burglars (Vaughn et al., 2008). The explicit assumption being made with these taxonomies is that offenders in one category differ in important ways from those in the other categories. While that may or may not be true (depending upon the taxonomy in question), the more problematic assumption is the one being implicitly made that offenders tend to exhibit a certain amount of specialization within their designated group (see, e.g., the discussion by Sullivan et al., 2006).

The problem is that these categories do not match up very well with offenders’ actual behavioral patterns. The reality is that offenders are not all that picky when it comes to their misbehavior (McGloin et al., 2007). Those who steal things are also more likely to use illegal substances and to threaten other people with violence, and those who drink and drive are also more likely to abuse their children and to assault their spouses (Mazerolle et al., 2000; Yonai, Levine, and Glickson, 2013). Recent research also shows that offenders are even more likely to “drunk dial” their friends and to use profanity in public settings (Reisig and Pratt, 2011). This is the full picture of offending patterns that led to the notion of the “generality of deviance” (Hirschi and Gottfredson, 1994).

Research also highlights the generality of the consequences of criminal behavior. To be sure, criminal offending has been linked to a wide array of other negative life outcomes (see, e.g., Brame et al., 2012), including long stretches of unemployment and job instability (Bushway, 1998; Nagin and Waldfogel, 1995; Western, Kling, and Weiman, 2001), educational deficits (Sweeten, 2006; Weisner, Kim, and Capaldi, 2010), welfare dependence (Makarios et al., 2015), violent victimization (Averdijk and Bernasco, 2015; Berg et al., 2012; Turanovic, Reisig, and Pratt, 2015), poor health (Piquero et al., 2007; Reingle et al., 2014), and even early death (Laub and Valliant, 2000; Piquero et al., 2014). Put simply, engaging in criminal behavior is generally not a recipe for leading a happy and healthy life.

But not all of those who break the law suffer these negative consequences. What, then, separates those who experience such adverse outcomes from those who do not? The answer to that question likely lies in the complex interplay between one’s personal characteristics (de Ridder et al., 2012), access to quality social ties (Cullen, 1994; Turanovic and Pratt, 2015), and the structural conditions in which they reside (McManus et al., 2010; Sampson, 2012). And while all of that complexity needs to be considered, an additional promising place to look is the role that arrest plays in the processes that link offending to other negative life outcomes. Those who offend and are able to avoid being arrested for their crimes are, in essence, successful criminals who may be more adept at navigating a lifestyle that contains certain unconventional elements. It is possible that their level of success with criminal behavior may translate to success in other areas of their lives—for example, being able to acquire and maintain legitimate work when it becomes available, and playing an important role in family life (Laub and Sampson, 2003; Rose and Clear, 1998).

Conversely, those who get arrested as a result of their misbehavior have essentially failed at crime. And failing at crime—something that Gottfredson and Hirschi (1990:xv) have noted is comprised of “mundane, simple, trivial, easy acts aimed at satisfying the desires of the moment”—might be a harbinger of a special kind of incompetence. Indeed, if someone is unable to commit and get away with an act that “may be invented by even the most unimaginative and untutored among us” (Hirschi, 2015:90), how likely is it that they will be able to get a job and keep it? How likely is it they will attract a mate and put in all the work necessary to keep a successful marriage going? How will they be able to live a life that will leave them physically and mentally healthy? It is likely that they will have difficulty with all of these things. It is therefore possible that one’s arrest may serve as an indicator of a broader pattern of life failure—a possibility that has yet to be considered by criminologists.

RESEARCH STRATEGY

The purpose of the present study is to examine whether arrest—conceived of as a form of failure—is associated with a wide range of other life failures. If that is what we end up finding, there are a couple of explanations that would need to be considered. First, it is possible that there could be a labeling effect that occurs with being arrested. In this case, the offender is stigmatized as a result of his/her involvement with the criminal justice system. This label, in turn, limits his/her ability to succeed in other life domains (Pager, 2007; Uggen, et al., 2014). This is not an explanation that we can examine directly, yet if there is a labeling effect at work what we would likely see in our data is that arrest would be associated with a limited set of outcomes (e.g., employment and education) but would be unrelated to others (e.g., acquiring an STD; alcoholism).

A second explanation would be that there is some other personal factor that would lead someone to get arrested and would also lead them to fail in other life domains. For example, intelligence has been linked to a wide variety of life domains, including education, employment, welfare dependence, parental efficacy, and criminal behavior (Herrnstein and Murray, 1994; Wilson and Herrnstein, 1985). If this is the case, controlling for IQ should fully account for the effect of arrest on our outcomes. In addition, Gottfredson and Hirschi (1990) have long held that those who lack self-control will offend more often and will also fail more often in other life tasks where those with higher self-control will succeed. This contention of theirs has garnered considerable empirical support over the years (de Ridder et al., 2012; Pratt and Cullen, 2000). And if they are right, then controlling for self-control should account for the effect of arrest on our outcomes.

It is possible, however, that we will find that arrest is associated with a wide array of negative outcomes (a more diverse range of life failures than would be expected from a labeling perspective).¹ It is equally possible that we may find that the “arrest effect” is not accounted for by self-control or IQ. If so, then something else could be at work. And while we do not yet have a clearly specified theory of life failure, our results might suggest that we need one.

METHODS

Data

The data for this study were drawn from the National Longitudinal Study of Adolescent to Adult Health (Add Health; Harris, 2009). The Add Health data have been described at length elsewhere (e.g., Harris et al., 2009) so we offer only a brief overview here. The Add Health is a large, prospective longitudinal study of American youth who were enrolled in middle or high school during the 1994–95 school year. The study began by identifying 132 middle and high schools using multi-stage probability sampling methods. All students attending these schools were invited to take part in the study and this initial round of surveys, which is often referred to as the school-based survey, netted responses from more than 90,000 youth. Forming the prospective longitudinal cohort was a subsample of these 90,000 respondents. Specifically, 20,745 youth drawn probabilistically from the in-school sample were enrolled in the longitudinal cohort and were given an extended interview that took place in their home. These interviews are referred to as wave 1. During the wave 1 interviews, respondents ranged between 11 and 21 years of age with a mean age of 15.70. A wide variety of topics spanning domains such as the youth’s health, social life, and involvement in delinquent activities were covered in the interviews.

The second wave of data collection occurred approximately one year later. Due to the relatively short time interval between the first two waves, the wave 2 survey instrument was nearly identical to the wave 1 questionnaire. Wave 3 interviews took place approximately five years after the wave 2 interviews. Most of the respondents had aged into young adulthood by the time wave 3 interviews took place. As a result, the survey instruments were updated to account for the aging cohort. Finally, the most recent wave of data collection took place between 2007 and 2008, when most of the respondents were between 24 and 32 years old. The wave 4 interviews were conducted with 15,701 of the original respondents. During wave 4 interviews, each respondent was asked a variety of questions about their overall wellbeing, their relationship history, their educational history, their labor market status, and their involvement with the criminal justice system.

The present study will draw on information gleaned from wave 1 and wave 4 of the Add Health data. In general, the dependent variables and the key independent variable analyzed in this study were drawn from the wave 4 interviews. A host of control variables were taken from the wave 1 interview to help rule out potential confounding influences. After eliminating cases with missing data on any of the included variables, the analytic sample size was 12,589.²

Key Independent Variable: Arrest

The key independent variable was a dichotomous indicator of the respondent’s arrest history. Specifically, during wave 4 interviews, all respondents were asked the following question: “Have you ever been arrested?” Responses were coded so that no = 0 and yes = 1. It is important to note that a few respondents (n = 73) were not prompted with this question because their interview took place in prison. All of these cases were coded as “1”. Descriptive statistics for this and all other variables utilized in the analysis can be found in Table 1.

Table 1.

Descriptive statistics.

Variables	Mean	S.D.	Min	Max
Key Independent Variable
Arrest	0.270	--	0	1
Outcome Variables: Life Failures
Number of Jobs	3.223	2.107	0	10
Number of Times Fired	0.424	0.755	0	3
Working 10+ hrs/week	0.827	--	0	1
Repeated a Grade (W1)	0.195	--	0	1
High School Dropout	0.067	--	0	1
Ever Married	0.505	--	0	1
Marriage Ended	0.041	--	0	1
Any STD	0.233	--	0	1
Alcoholism	0.300	--	0	1
Key Confounding Variables
Low Self-Control (W1)	28.036	7.450	0	72
IQ (W1)	101.476	14.086	14	146
Control Variables
Drug Use (W1)	1.573	1.384	0	6
Delinquency (W1)	4.307	5.234	0	47
Criminal Behavior	0.384	1.273	0	20
Concentrated Disadvantage	−0.034	0.888	−1.492	5.749
Age	28.526	1.751	24	34
Male	0.452	--	0	1
White	0.650	--	0	1
Black	0.214	--	0	1

Note: All variables are measured at wave 4 unless otherwise indicated.

N = 12,589.

Outcome Variables: Life Failures

Number of Jobs

During the wave 4 interview all participants were asked to report the number of jobs they had held since 2001. Respondents were prompted to include any job that lasted 9 weeks or more and included at least 10 hours of work each week. Participants originally provided responses in discrete count intervals that ranged between 0 and 50 (where 50 included all responses that were over 50), but more than 96% of the cases reported between 0 and 10 jobs. As a result, the upper end of the range was capped at 10 for the analysis in order to limit the influence of outliers. Also, respondents were told to exclude military service. Referring to other parts of the survey allowed us to identify respondents who were active duty military. Responses to the Number of Jobs question were augmented by adding a value of “1” to the original response value if the subject was active duty military. For example, someone who originally reported no jobs (“0”) was now coded as having had one job (“1”).

Number of Times Fired

Any respondent who reported having a least one job since 2001 was asked a follow-up question. Specifically, these participants were asked how many times they had been fired, let go, or laid off from a job. Originally, responses ranged between 0 and 50, but the vast majority of respondents (95%) reported values between 0 and 3 so all cases were capped at 3 for the analysis. Subjects who had never held a job or who had not held a job since 2001 were coded as 0.

Working 10+ Hours

Each of the wave 4 participants was asked whether s/he was currently working for pay at least 10 hours a week. Responses were coded as no = 0 and yes = 1. As before, respondents were specifically told not to include military service. We considered military service as currently working 10+ per week and, thus, coded active duty military as “1”.

Repeated a Grade

During the wave 1 interview, each respondent was asked whether they had ever repeated a grade or had been held back in school. Responses were coded so that no = 0 and yes = 1. We recognize that the temporal ordering of events is out of alignment for this variable. Specifically, our key independent variable was drawn from wave 4 but the present outcome variable was taken from wave 1. It is important to reiterate the present study is not intended to identify a causal relationship between arrest and the various outcomes, but rather to identify whether individuals who experience an arrest are at heightened risk for other life failures. Thus, drawing an outcome variable from wave 1 becomes less of a concern due to the exploratory nature of the analysis.

High School Dropout

Wave 4 respondents were asked to report the highest level of education they had achieved to date. Anyone reporting less than a high school education was coded “1” and all others (responses ranged between high school graduate and post baccalaureate education) with valid responses were coded “0.”

Ever Married

During the wave 4 interviews respondents were asked to report the number of persons to whom they had been married. Responses were originally recorded as a discrete count variable that ranged between 0 and 4, but fewer than 50 respondents indicated they had been married more than twice. For the present analysis, responses were coded so that never married = 0 and married at least once = 1.

Marriage Ended

Although the Add Health data do not include a measure of divorce, it is possible to identify respondents whose marriage(s) ended by drawing on responses to the Ever Married question. Specifically, a new variable was constructed that was coded so that respondents who were never married or were married only once were coded as “0” and anyone reporting more than one marriage were coded as “1.”

Any STD

During the wave 4 interviews, respondents were asked whether a doctor, nurse, or other health professional had told them they had any of a list of sexually transmitted diseases (STD). From this list, we identified STDs that are not sex-specific and that had case counts above 50. Specifically, participants who indicated having been diagnosed with any of the following STDs were coded “1” and all others were coded “0”: chlamydia, gonorrhea, trichomoniasis, syphilis, genital herpes, genital warts, hepatitis B (HBV), human papilloma virus (HPV), and any other sexually transmitted disease.

Alcoholism

The wave 4 interviews included a host of questions aimed at identifying respondents with substance abuse problems. To begin, respondents were asked whether they had ever had a drink of beer, wine, or liquor more than two or three times. Nearly 80% of all respondents gave an affirmative response. These participants were asked several follow-up questions, many of which can be considered indicators of alcoholism. Specifically, each respondent was asked: 1) how often drinking had interfered with their responsibilities at work or school; 2) how often they had been under the influence of alcohol when they could have hurt themselves or others, or they had put themselves or others at risk, including unprotected sex; 3) how often they had legal problems because of drinking, like being arrested for disturbing the peace or driving under the influence of alcohol, or anything else; and 4) how often they had problems with family, friends, or people at work or school because of their drinking. Each of these four variables were originally coded so that never = 0, one time = 1, and more than one time = 2. We combined information from these four questions by creating a new variable that was coded as “1” if the respondent was coded as “1” or “2” on any of the four focal variables. All other respondents—including non-drinkers—with valid information were coded as “0.”

Key Potential Confounding Variables

Two potential confounding variables, IQ and low self-control, are included in the analyses. IQ was assessed using each respondent’s age-normed Add Health Picture Vocabulary Test (PVT) score. Add Health PVT scores come from a shorter, computerized version of the Peabody Picture Vocabulary Test (Revised) that was administered to adolescents during the wave 1 interviews. PVT scores are normally distributed with a mean of approximately 100 and a standard deviation of approximately 15.

Low self-control

The wave 1 questionnaire included 19 items that have been identified as tapping into various domains of self-control such as impulsivity, preferring physical tasks over mental tasks, and having a short temper (Gottfredson and Hirschi, 1990). When the 19 items were combined together, higher values on the scale indexed lower levels of self-control (α = .741). This scale is consistent—albeit a slightly shortened version to preserve case counts—with prior research using the Add Health data (Miller et al., 2011).

Additional Control Variables

Drug Use

The wave 1 interview included a series of drug use questions that asked whether the participant had ever smoked cigarettes, drank alcohol, smoked marijuana, used cocaine, used inhalants, or used any other type of drug. Each question was coded so that no = 0 and yes = 1. Each of these questions was combined together by summing across the six items so that the resulting scale ranged between 0 and 6.

Delinquency

The wave 1 questionnaire included 17 questions about the respondent’s involvement in delinquency. The items referenced covered property crimes such as breaking into a building and violent crimes such as using a weapon in a fight. When combined together, higher values reflected a greater involvement in delinquency (α = .832).

Criminal Behavior

The wave 4 interviews included 12 questions that referenced involvement in property crimes, financial crimes, drug crimes, and violent crimes. When combined together the criminal behavior sale was coded so that higher values indicated a greater involvement in criminal activities (α = .674).

Concentrated Disadvantage

Census tract data for the neighborhoods in which each of the respondents resided was available at wave 4. A measure of concentrated disadvantage (Sampson et al., 1997) was generated by combining information from the following five items: proportion of the population below the poverty level in the past 12 months; proportion of households receiving public assistance income in the past 12 months; proportion of families with own children headed by a female householder; the unemployment rate for all persons 16 years and over; and the proportion Black or African American alone. Each of the five items was measured using the 2009 Census estimates. The scale was created by extracting the first factor from an exploratory factor analysis (eigenvalue = 2.495, α = .727). Higher values indicated greater concentrated disadvantage.³

Lastly, several demographic variables were included. These control variables were the respondent’s Age at wave 4 (ranging between 24 and 34 years), Male (where male = 1 and female = 0), and Race. Specifically, race was included as two dummy variables that identified participants as White (=1) or Black (=1). All other races were included as the reference category.

Analytic Strategy

Two generalized linear models (GLM) were estimated to analyze the statistical relationships between Arrest and the outcome variables tapping life failure: the negative binomial model and the logistic regression model. The negative binomial model is a count model that is part of the Poisson family of regression models that is used when the outcome variable is overdispersed (the conditional variance is greater than the mean, see Cameron and Trivedi, 2013), which was observed in the count variables analyzed here (i.e., Number of Jobs and Number of Times Fired). The negative binomial model can be expressed algebraically as:

$$ln({\mu }_i)=b_0+b_1({\mathit{arrest}}_i)+\sum _{q=1}^Q{b}_q(C_{qi}),$$ (Eq. 1)

where the left-hand side of the equation is the natural log of μ_i, which is the expected value of the outcome variable for person i given his/her values on the predictor variables; b₀ is the intercept term; b₁ provides the maximum-likelihood parameter estimate for the effect of Arrest on the outcome; and ${\displaystyle \sum _{q=1}^Q}{b}_q(C_{qi})$ represents the cumulative effect of the control variables on the life-failure of focus.

The second GLM that will be estimated is the logistic regression model. The logistic regression model is a form of the GLM that uses a logit link function to align the left-hand side of the regression equation with the right-hand side predictors:

$$ln\left(\frac{P_i[\mathit{lifeFailure}=1]}{1-P_i[\mathit{lifeFailure}=1]}\right)=b_0+b_1({\mathit{arrest}}_i)+\sum _{q=1}^Q{b}_q(C_{qi}),$$ (Eq. 2)

where the left-hand side of the equation represents the log odds of a particular life failure for individual i; b₀ is the intercept term; b₁ provides the maximum-likelihood prediction of the effect of Arrest on the life-failure of focus; and ${\displaystyle \sum _{q=1}^Q}{b}_q(C_{qi})$ still represents the cumulative effect of the control variables on the life-failure of focus. The logistic regression model will be estimated for all of the dependent variables coded dichotomously.

The coefficient estimates are provided in their original metric—the natural logarithm of the expected count of a life failure in the negative binomial model and the natural logarithm of the odds of life failure in the logistic regression model. Recognizing that logged counts and logged odds are less intuitive than other metrics, we will convert the coefficient estimates to incidence rate ratios (IRR) in the negative binomial model and to odds ratios (OR) in the logistic model for any findings discussed in the text.⁴

In addition, the coefficient estimates and standard errors are potentially biased if the complex sampling design of the Add Health study is not taken into account (Chen and Chantala, 2014). As a result, all of the coefficients were weighted using the sampling weights provided by the Add Health staff (Harris, 2009). Standard errors were also corrected for the clustering of individuals within the 132 schools at wave 1 and for the stratification processes that were carried out during the sampling phase.

RESULTS

Tables 2 and 3 provide the results from the negative binomial and logistic regression models that assess the effects of arrest on multiple life failures. Beginning with Table 2, it can be seen that the effect of Arrest as a predictor of life failure is statistically significant in all of the models. Specifically, in model 1 of Table 2, the incident rate ratio (IRR) of arrest is 1.148, meaning that respondents who reported an arrest also reported approximately 15% more jobs, on average, than respondents who did not report having been arrested. As for the Number of Times Fired outcome in model 2 of Table 2, arrestees reported 63.4% more firings on average than non-arrestees (IRR = 1.638). In the remaining models we can see that, compared to non-arrestees, arrestees were 27% less likely to be working more than 10 hours per week (e^−.322 − 1), 34% more likely to have repeated a grade in school, and 244 times more likely to have dropped out of high school.

Table 2.

Negative binomial and logistic regression models for employment and education life failures.

Variables	Number of Jobsa		Number of Times Fireda		Working 10+ hrs/weekb		Repeated a Gradeb		High School Dropoutb
	Model 1		Model 2		Model 3		Model 4		Model 5
	b	(SE)	B	(SE)	B	(SE)	b	(SE)	b	(SE)
Arrest	.138**	(.019)	.491**	(.050)	−.322*	(.098)	.290**	(.078)	.895**	(.147)
Low self-control (W1)	.004*	(.001)	.008*	(.003)	−,.022**	(.006)	.016*	(.005)	.034**	(.008)
IQ (W1)	.004**	(.001)	−.006*	(.002)	.011**	(.003)	−.053**	(.004)	−.064**	(.004)
Drug use (W1)	.007	(.008)	.009	(.018)	−.010	(.032)	.040	(.033)	.159**	(.042)
Delinquency (W1)	.003	(.002)	.006	(.004)	.008	(.009)	.009	(.008)	.009	(.009)
Criminal Behavior	.027**	(.006)	.063**	(.010)	−.011	(.022)	−.012	(.028)	−.012	(.035)
Conc. Disadvantage	.004	(.011)	.108**	(.025)	−.164**	(.039)	.148*	(.047)	.350**	(.072)
Age	−.049**	(.005)	−.026	(.012)	.006	(.022)	.274**	(.027)	−.158**	(.039)
Male	.051**	(.015)	.382**	(.046)	.776**	(.084)	.594**	(.072)	.421**	(.103)
White	.00004	(.029)	.058	(.069)	−.166	(.114)	.286	(.155)	.321	(.163)
Black	.089*	(.034)	.304**	(.077)	−.067	(.134)	.440*	(.158)	−.346	(.172)

Note. All estimates corrected and standard errors adjusted for survey design features of the Add Health (N = 12,589); b = Unstandardized regression coefficient; SE = standard error.

^aNegative binomial regression model.

^bLogistic regression model.

^*p < .01,

^**p < .001 (two-tailed test).

Table 3.

Logistic regression models for marriage and health life failures.

Variables	Ever Married		Marriage Ended		Any STD		Alcoholism
	Model 1		Model 2		Model 3		Model 4
	b	(SE)	b	(SE)	b	(SE)	b	(SE)
Arrest	−.363**	(.071)	.141	(.145)	.527**	(.079)	.905**	(.074)
Low self-control (W1)	−.002	(.004)	.011	(.011)	.009	(.004)	.016*	(.005)
IQ (W1)	−.007	(.003)	−.011*	(.004)	.005	(.003)	.031**	(.002)
Drug use (W1)	.023	(.027)	.062	(.052)	.234**	(.027)	.234**	(.026)
Delinquency (W1)	−.004	(.007)	−.011	(.012)	.005	(.008)	.004	(.006)
Criminal Behavior	−.141**	(.030)	−.022	(.052)	.111**	(.024)	.138**	(.028)
Conc. Disadvantage	−.222**	(.039)	−.016	(.083)	.071	(.050)	−.100	(.044)
Age	.229**	(.024)	.206**	(.044)	−.056	(.022)	−.109**	(.019)
Male	−.331**	(.060)	−.496*	(.148)	−1.376**	(.068)	.193*	(.060)
White	.487**	(.087)	.819**	(.207)	−.123	(.129)	.238	(.110)
Black	−.401**	(.104)	−.369	(.310)	1.077**	(.147)	−.642**	(.134)

Note. All estimates corrected and standard errors adjusted for survey design features of the Add Health (N = 12,589); b = Unstandardized regression coefficient; SE = standard error.

^*p < .01,

^**p < .001 (two-tailed test).

A similar pattern emerges in Table 3, where the Arrest variable was a statistically significant predictor of all but one of the outcome variables. The only one that it did not predict was Marriage Ended. As for the remaining equations, Arrest was a statistically significant predictor of the life failures in the expected direction: arrestees reported more life failures. In particular, respondents who had been arrested were 30% less likely to have been married, 169 times more likely to have reported an STD, and 147% more likely to have been coded as having problems related to alcoholism.⁵

In short, the GLMs in Tables 2 and 3 reveal that the association of arrest with other life failures appears to be rather “general.” Indeed, our results reveal significant links between arrest and a wide array of negative life outcomes in various domains such as education, employment, relationships, and health. This pattern of findings makes sense given that, even independent of arrest, these form of life failure are all related to each other (e.g., additional analyses demonstrate that when these forms failure are extracted into a common factor, that factor is significantly correlated with each of the forms of life failure that we examine). And second, our results show that including controls for IQ and self-control does not account for the relationships between arrest and our outcomes of interest.

DISCUSSION

The idea that offenders vary in how successful they can be in their criminal enterprises is not new. Works such as Conwell and Sutherland’s (1937) “professional thief” and Chambliss’s (1973) “saints and roughnecks” have long cautioned criminologists against completely abnormalizing crime. To be sure, treating crime/delinquency, in and of itself, as an indicator of the failure to live a conventional life obscures how much criminal behavior can mirror that of other more conventional behavioral domains. We have no trouble believing, for example, that some youngsters are better at focusing in on academic tasks than others (i.e., that there are good students and not-so-good students), or that some people are better at meeting the demands of honest employment than others (i.e., that there are good employees and lousy ones). Yet in much the same way, some people are simply more successful at engaging in crime than others (see generally, Barnes et al., 2015; Brame et al., 2012).

It was with this backdrop in mind that we introduced the concept of arrest as being a form of “failure.” Our key objective, then, was to examine whether that kind of failure would be associated with other forms of life failure, such as in the areas of education, employment, and relationships. In short, we asked: just like there is the “generality of deviance,” is there also a “generality of failure”? Based on the analyses we presented here, the answer to that question is a resounding “yes.” And as a result of arriving at that answer, three additional issues warrant further consideration.

First, the association of arrest to other forms of life failure extends well into adulthood. This is important since the negative consequences of arrests for juveniles have been well documented, at least with respect to how being arrested predicts subsequent offending and rearrest (Liberman, Kirk, and Kim, 2014), as well as dropping out of school (Kirk and Sampson, 2013). In moving beyond this critical yet narrow range of outcomes, our results reveal that being arrested is also linked with several other types of life failure as well. This is important since recent estimates suggest that 30% of Americans—and roughly 40% of males—will be arrested by the time they reach middle-adulthood (Barnes et al., 2015; Brame et al., 2012). And since the association between failing at crime and failing at other life tasks does not appear to be confined to any particular stage of life, we could be looking at an extremely large population of people who are at risk for a host of negative life outcomes. Nevertheless, as Turanovic (2015) recently noted, “the life course does not end with wave IV of Add Health. People still have interesting and important life experiences well after their thirtieth birthday” (p. 157). It therefore remains to be seen whether and how these relationships hold when assessed in the latter stages of the life course—stages that criminologists are just beginning to pay serious attention to (Holtfreter et al., 2015; Mears et al., 2014; Reisig and Holtfreter, 2013; Wolfe, 2014).

Second, the causal processes linking arrest to these various negative life outcomes still need to be determined. It is entirely possible, for example, that one’s arrest may be both a consequence of previous life failures as well as a cause of future failures. It may also be the case that one’s arrest is not a cause of other life failures at all, but is instead a manifestation of other enduring traits or circumstances that result in both being arrested and in not being very good at other things in life (see, e.g., de Ridder et al., 2012; Gottfredson and Hirschi, 1990). It is also possible that the degree to which these relationships may or may not be causal could be age-graded. Either way, thinking about arrest as failure might provide scholars conducting future research with a new way of assessing the role of arrest in the context of various outcomes over the life course.

Third, it is important to note that the arrest effect was not confined to those outcomes that could be attributed to the potential stigma associated with being arrested. Put simply, our results are too robust across a wide range of life failures for labeling theory to provide a credible explanation of them. In addition, the arrest effect could not be explained away by controlling for certain individual traits like IQ or self-control. Thus, those who focus on the potential negative consequences of particular kinds of personal characteristics (see, e.g., Gottfredson and Hirschi, 1990; Herrnstein and Murray, 1994) will not be able to explain our pattern of findings either.

There are, however, a couple of avenues that may be more promising to explore in an effort to develop a “theory of failure” that could explain our results. The first concerns the cluster of life skills that appear to be necessary for successfully navigating the social world. Outside of the correctional rehabilitation literature, criminological scholarship has generally ignored the issue of problem solving skills. But we know these skills are important in other areas of criminological theory and research (e.g., “coping skills” in response to negative life events like victimization, see Agnew, 2006; Turanovic and Pratt, 2013). It may therefore be time to look more closely at these skills with respect to life failures. And second, it may be useful to explore how these forms of failure are experienced and internalized by those doing the failing. In particular, does repeated failure result in a loss of self-efficacy or the creation of what Maruna (2001) refers to as a “condemnation script” (see also LeBel et al., 2008; Hallett and McCoy, 2015; Fader and Traylor, 2015)? Put differently, do those who fail across a wide range of life domains adopt a defeated inner voice that says: “I suck at everything”? Any theory of the generality of failure may benefit greatly from looking closely at this literature about how personal identities get formed, reinforced, and maintained (see also Giordano, 2010).

And with that in mind, we wish to caution scholars against attempting to explain the generality of failure as a strictly “individual” problem. That someone might get arrested and that they might also become an alcoholic, chase away a romantic partner, and lose their job is not preordained at birth. All of these problems instead have roots in early developmental and family processes (Giordano, 2010; Wright et al., 2015), in the social and institutional contexts that structure our daily routines and opportunities (Bursik and Grasmick, 1993; Wright et al., 2013), and in the unique experiences that we all have as we live our lives through time (Sampson and Laub, 2005).

In the end, one’s arrest is unlikely to be the lone blemish on an otherwise spotless record of personal and professional triumphs. Thus, our study points to a rather simple yet sad conclusion: people who fail at crime also fail at a lot of other things too. What is not so simple, however, is the notion that the problems of crime, arrest, and life failure are all likely to be the consequence of the complex relationships that exist between individual characteristics and social context—all of which are structurally embedded (Sampson, 2012). This reality has important implications for a criminal justice system—and for the scholars who study its dynamics—that focuses heavily on failure (e.g., failure to complete treatment; failing a urinalysis; failure to comply with conditions of supervision; failure to appear in court; and so on). We hope that our work may provide fresh insight into these very problems as researchers continue the critical task of understanding what keeps people from leading successful, happy, and healthy lives.

Biographies

Travis C. Pratt is a Fellow with the University of Cincinnati Corrections Institute. His research focuses primarily on criminological theory and correctional policy. He is the author of Addicted to Incarceration: Corrections Policy and the Politics of Misinformation in the United States and his work has appeared in Criminology, Journal of Research in Crime and Delinquency, Journal of Quantitative Criminology, and Crime and Justice: A Review of Research.

J.C. Barnes is Associate Professor in the School of Criminal Justice at the University of Cincinnati. He is a biosocial criminologist whose research seeks to understand how genetic and environmental factors combine to impact criminological phenomena. He has published in outlets such as Behavior Genetics, Criminology, Developmental Psychology, and PLoS ONE.

Francis T. Cullen is Distinguished Research Professor Emeritus and a Senior Research Associate in the School of Criminal Justice at the University of Cincinnati. His recent works include Challenging Criminological Theory: The Legacy of Ruth Rosner Kornhauser, Sisters in Crime Revisited: Bringing Gender into Criminology, and Reaffirming Rehabilitation (30^th anniversary edition). His current research focuses on the organization of criminological knowledge and on rehabilitation as a correctional policy. He is a Past President of both the American Society of Criminology and the Academy of Criminal Justice Sciences.

Jillian J. Turanovic is Assistant Professor in the College of Criminology and Criminal Justice at Florida State University. Her current research examines victimization and offending over the life course, and the collateral consequences of incarceration. Her recent publications have appeared in Criminology, Journal of Quantitative Criminology, Justice Quarterly, and Journal of Pediatrics.