Immigration and crime: a time-trend analysis of self-reported crime in Sweden, 1999–2017

ABSTRACT

This study investigates the direction and rate of change in self-reported crime over time, based on immigrant status and region of origin. The study is based on eight nationally representive school surveys conducted by the National Council for Crime Prevention between 1999 and 2017, with a sample of 50,657 adolescents. Results in this study showed a decreasing trend in self-reported offending among both first and second generation immigrant youth, and also among immigrant youths from different regions of origin. The results also show that offending has declined at a faster rate among first generation immigrants by comparison with native Swedes.

KEYWORDS:

• Immigration and crime
• crime-trends
• self-report
• youth-crime
• Sweden

Background and aim

Trends in youth crime are a topic that has been the subject of much discussion over the years. One central question in the crime policy debate is that of whether youth crime is increasing or decreasing. In both the US and several European countries studies indicate a declining trend in rates of crime and delinquency for several types of offences (Arnett, 2018; Balvig, 2011; Fernández-Molina & Bartolomé Gutiérrez, 2018; Grucza et al., 2018; Moss, Santaella‐Tenorio, Mauro, Keyes, & Martins, 2019). In general, the Swedish research shows a similar picture, indicating that both official and self-reported crime rates have decreased over time for several types of crime; youths from a range of different backgrounds are tending to report lower rates of participation in crime (Shannon, Bäckman, Estrada, & Nilsson, 2014; Svensson & Ring, 2007).

However, while crime rates appear to have declined in recent decades, questions still remain as to whether the decline is specific rather than general (Nilsson, Estrada, & Bäckman, 2016). Few studies have examined whether and how youth crime has changed among different immigrant groups in comparison to their native-born counterparts. Although the vast majority of Swedish studies suggest that immigrants are overrepresented in crime by comparison with the native born population (see Kardell, 2010 for a review), no studies have yet examined whether this gap is increasing or decreasing and whether the direction and rate of change in crime trends differs across immigrant groups and immigrant generations. Immigrants and their children are characterized by considerable heterogeneity in terms of socioeconomic status, levels of social integration, and cultural factors – implying that the dynamics of crime may not be identical for all immigrants (Skardhamar, Aaltonen, & Lehti, 2014). However, immigrants have generally been studied as a single group, resulting in little knowledge about potential changes in the delinquent involvement of different immigrant groups over time.

With regard to youth crime, the self-report method has become a valuable tool for the measurement and analysis of delinqency trends (Kivivuori & Bernburg, 2011). Since 1999, the Swedish National Council for Crime Prevention has administered a nationally representative self-report study of crime among youth in their final year of compulsory education. The survey constitutes an important data source for studying the state of youth crime in Sweden and allows us to produce a picture of the crime that is not reported to the police. Most European studies on crime trends have been based on official crime statistics, i.e. register data. Registered delinquency often primarily includes more serious crimes, and is therefore often considered to be of limited use for estimating the prevalence of crime in general (Svensson & Ring, 2007).

Time trend regression analysis is a commonly used procedure for estimating and comparing rates of growth, or rates of change, across aggregate groups in repeated cross-sectional studies (e.g. Scholes et al., 2012). By fitting a time trend regression line, we are able to estimate how rapidly any systematic change in offending is occurring as time passes. Compared to the traditional approaches used in Nordic crime trend research (for a review see Kivivuori & Bernburg, 2011), this approach has the advantage of providing a direct and simple way to compare rates of change across immigrant subgroups via the inclusion of a group-by-time interaction. This type of analysis may provide valuable information on how fast social problems are increasing or decreasing among different immigrant groups in relation to native swedes, and to clarify whether there are issues that need to be addressed from a theoretical or policy perspective.

On the basis of nationally representative school survey data from 1999 to 2017, the aim of this study is to investigate whether and how self-reported delinquency has changed over time among youths with different immigrant status (i.e. first- or second generation) and from different regions of origin (i.e. region of birth), and whether the direction and rate of change differs in comparison to native Swedes.

Method

Participants

The study is based on eight waves of a nationally representive school survey of year nine youth, (aged 15), conducted by the Swedish National Council for Crime Prevention between 1999 and 2017. The survey was conducted every second year between 1999 and 2005, and thereafter every third or fourth year. The surveys are based on systematic samples of schools with year-nine classes. The data have primarily been collected in December, and the questionnaires are completed during lesson time.

For the surveys conducted between 1999 and 2017, the response frequency varied between 87% and 69%. To compensate for non-response and to make the data as representive as possible, calibration weights were used for the surveys conducted in 2015 and 2017.

In each survey, between 4,200 and 7,400 students have anonymously completed a questionnaire containing questions about their social situation, family, school, peer group, and leisure activities. The eight subsamples combined produce a total sample of 50,657 adolescents at 1,065 schools.

Since the youths have given informed consent¹ and completed the questionnaire anonymously, we see no general ethical problems linked to this study. We would nonetheless like to emphasize that studies such as this, which are based on differentiating between different groups of youths, are always associated with a risk of stigmatizing certain groups. We are aware of this risk, but feel that it must be balanced against the benefits that may accrue from improved knowledge in this area. The present study was approved by the Regional Ethics Committee at Lund University (Dnr 2015/784).

Measures

Definition of immigrant background. The relevant questions included in the survey are: In which country were you born? In which country was your mother born? In which country was your father born? The response alternatives are: Sweden/Another country. If the respondent answers that he/she or one/both parents were born in another country, the respondent is asked which country. Different definitions of immigrant background have been employed in different research contexts, based on whether individuals themselves and one or both of their parents were born in Sweden or abroad. We will be using the following division:

1. Born in Sweden to two Swedish-born parents

2. Born in Sweden to one Swedish-born and one foreign-born parent

3. Born in Sweden to two foreign-born parents

4. Born abroad

In the total sample, 74.5% of the respondents were born in Sweden to two Swedish-born parents, 9.2% were born in Sweden to one Swedish-born and one foreign-born parent, 6.5% were born in Sweden to two foreign-born parents, and 6.2% of the respondents were born abroad.²

Information on the respondents’ place of birth was collected for the surveys conducted between 1999 and 2011. Because many immigrant groups are small and not suitable for separate statistical analysis, reporting crime rates by single countries of origin is not possible. Four subgroups were created, based on the region of the world in which the respondent was born. The subgroups were ‘born in Sweden’ (including second generation immigrants), ‘western countries’, ‘other European countries’, and ‘non-European countries’.^3,4 In the total sample for the surveys conducted between 1999 and 2011 (n = 41,579), 92.9% (n = 38,639) of the respondents were born in Sweden, 1.2% (n = 512) in western countries, 2.1% (n = 856) in other European countries, and 3.8% (n = 1570) in non-European countries.

Self-reported offending measures how many times during the past 12 months the respondents report having committed various offences from a wide range of items measuring violence, theft, and criminal damage. For the descriptive analysis, we will present prevalence data, i.e. whether the youths have engaged in one or more of the relevant offence types, using the following scales: an overall offending scale including all of the 23 items with an Alpha of .89, violence subscale (6 items) with an Alpha of .67, theft subscale (13 items) with an Alpha of .85, and criminal damage subscale (4 items) with an Alpha of .70. It is important to note that the formulations used in some of the items on offending have changed over time. The same questions were used during the period 1999–2011. From 2015, the questions have been subject to some minor adjustments.

For the main analysis, we will be analysing differences in the direction and rate of change and we will use the overall offending scale transformed into a variety scale measuring how many different types of offences the youths report having committed. Although ease of interpretation often leads researchers to employ prevalence and frequency scales, variety scales are the preferred method for summarizing changes in criminality over time (Sweeten, 2012). Variety scales have the advantage that they are not biased towards less serious crimes that occur more often, and they have been shown to have higher internal consistency, higher stability over time, to produce larger group differences, and to be more strongly correlated with conceptual variables than the more commonly used frequency scales (Bendixen, Endresen, & Olweus, 2003).⁵

The time trend variable was created by transforming the years of the survey into an ordered set of numbers with uniform intervals, t = (1, 2, 3, 4 … 8), which measures the time span between observations. This means that even if there is some irregularity in sampling years, we treat the values as regularly sampled, assuming that the responses would be valid if sampled on regular 2.5 yearly basis.

Findings

Descriptive statistics on prevalence

The proportion of youths reporting involvement in theft, criminal damage and violence during the observation period (1999–2017), by immigrant status and region of origin, are presented in Tables 1 and 2. In general, the proportion of youths who report having committed at least one of the acts included in the studied crime categories appears to have decreased since 1999 in all groups. It should also be mentioned that only a small number of young people have reported involvement in the more serious crime types. Analyses have also been conducted focusing on frequency rates for the different crime types. Here the pattern is the same; there appears to be a decrease over time in both mean and median values for different crime types and this pattern is also found for the different immigration groups (results not shown).

Table 1. Proportion (per cent) of youths reporting that they have engaged in different types of offences over the past 12 months, by immigrant status (1999–2017).

	1999	2001	2003	2005	2008	2011	2015	2017
By immigrant status
Swedish parents
Total	63.1	59.8	58.6	56.3	58.4	50.5	42.7	42.4
Theft	57.8	53.4	53.5	50.8	52.9	45.5	39.1	39.3
Criminal damage	35.9	34.2	31.2	29.3	30.9	24.3	16.6	16.9
Violence	17.3	15.6	16.1	13.9	12.9	9.9	9.7	10.1
One immigrant parent
Total	66.5	65.3	66.8	61.8	64.6	54.7	49.7	49.4
Theft	60.3	60.1	62.4	56.1	59.1	50.2	46.2	45.1
Criminal damage	42.2	39.5	39.2	35.1	38.7	27.0	22.8	24.1
Violence	22.6	21.9	23.7	19.0	19.4	13.0	12.4	13.7
Two immigrant parents
Total	65.0	66.2	64.7	59.0	62.5	44.8	42.9	39.1
Theft	61.0	58.5	55.5	53.1	53.8	37.5	36.8	35.0
Criminal damage	37.6	40.5	40.6	37.9	37.3	26.8	23.4	20.4
Violence	26.4	31.5	28.1	20.6	23.5	16.8	15.9	14.1
Born abroad
Total	62.3	60.1	58.6	59.1	61.9	43.6	32.5	33.5
Theft	54.8	53.2	49.7	50.2	55.4	35.6	26.7	28.0
Criminal damage	35.5	37.7	36.1	32.3	30.9	26.6	14.7	15.8
Violence	24.0	27.5	26.5	22.1	21.9	12.6	11.8	10.3

Table 2. Proportion (per cent) of youths reporting that they have engaged in different types of offences over the past 12 months, by region of origin (1999–2011).

		1999	2001	2003	2005	2008	2011
By region of origin
Sweden
Total		63.6	60.8	60.0	57.0	59.4	50.6
Theft		58.2	54.4	54.6	51.6	53.6	45.4
Criminal damage		36.8	35.2	32.7	30.4	32.2	25.0
Violence		18.5	17.2	17.8	14.9	14.6	10.9
Western countries
Total		67.6	61.3	56.8	60.7	73.0	53.8
Theft		64.3	56.1	56.5	50.8	66.7	50.8
Criminal damage		31.3	39.1	37.1	30.6	44.0	27.5
Violence		18.1	25.3	28.9	30.6	26.0	11.1
Other European countries
Total		64.9	62.9	53.6	60.5	63.2	42.9
Theft		54.9	53.8	42.8	52.9	61.8	35.9
Criminal damage		32.8	36.9	34.8	29.8	33.3	24.6
Violence		27.6	28.6	25.0	19.4	23.3	4.7
Non-European countries
Total		62.9	59.4	61.2	59.1	57.9	44.5
Theft		57.0	52.8	54.1	51.1	50.9	36.8
Criminal damage		39.0	35.4	35.0	32.3	28.1	22.8
Violence		25.2	24.7	23.4	21.2	17.4	13.9

Differences in the rate of change in self-reported crime between natives and immigrant youths

For this analysis, a negative binomial regression model was used to examine whether there are any differences in the direction and rate of change over time in self-reported crime between native and immigrant youth. Time trends were reported as the average percentage change in self-report offending for each 2.5 year increase in time, based on the assumption that the change was constant from assessment to assessment. Even if this assumption is not true, the estimates may be viewed as an approximate measure of the average change or trend (McNamee, Carder, Chen, & Agius, 2008). All regression analyses were conducted without weights.

The main analysis, focusing on immigrant status, region of origin, and change in self-reported offending over time, began by analysing the relationship between the time trend and self-reported offending for each and every category, and for boys and girls separately. The results are presented in the first section of Table 3 and show that self-report offending decreases over time in all groups of youths, with the exception of boys and girls born in ‘western countries’ and girls born in ‘other European countries’, where no change over time is found.⁶ The average percentage decrease in self-report offending ranges from 4.3% (CI: −5.3, −3.2, girls born in Sweden) to 19.6% (CI: −21.7, −17.3, boys born abroad) for each 2.5 year increase in time.

Table 3. Average percentage change in self-reported offending, as estimated by negative binomial regression, for each 2.5 year increase in time, by immigrant status (1999–2017), and region of origin (1999–2011).

	Total		Boys		Girls
	% Change	95% CI	% Change	95% CI	% Change	95% CI
Analysis 1. Time trends
Immigrant status, 1999-2017
Swedish parents	−9.6***	(−10.2, −9.1)	−11.1***	(−11.8, −10.3)	−7.4***	(−8.2, −6.5)
One immigrant parent	−9.1***	(−10.6, −.7.6)	−11.0***	(−13.0, −.9.9)	−7.3***	(−9.5, −5.0)
Two immigrant parents	−11.0***	(−12.8, −15.1)	−10.6***	(−13.1,-10.1)	−10.9***	(−13.5,-8.2)
Born abroad	−15.1***	(−16.8, −13.5)	−19.6***	(−21.7, −17.3)	−10.2***	(−12.9, −7.5)
Region of origin, 1999-2011
Born in Sweden	−7.0***	(−7.7, −6.3)	−8.5***	(−9.4, −7.5)	−4.3***	(−5.3, −3.2)
Western countries	−6.2*	(−11.7, −0.3.)	−8.0	(−15.9, 0.6)	−0.2	(−8.4, 8.7)
Other European countries	−9.2***	(−14.2, −3.9)	−13.9***	(−20.5, −6.6)	−4.5	(−12.0. 3.5)
Non-European countries	−10.0***	(−13.1, −6.9)	−9.6***	(−13.8, −5.2)	−9.7***	(−14.2, −4.9)
Analysis 2. Time by immigrant status (reference group: native Swedes)
One immigrant parent	0.6	(−1.2, 2.4)	0.1	(−2.3, 2.6)	0.1	(−2.5, 2.1)
Two immigrant parents	−1.5	(−3.6, 6.0)	0.5	(−2.3, 3.5)	−3.8*	(−6.7,-0.8)
Born abroad	−6.1***	(−8.0, −4.1)	−9.5***	(−12.9, −6.1)	−3.1	(−6.1, 0.0)
Analysis 3. Time by Region of origin (reference group: youth born in Sweden)
Western countries	0.8	(−5.2, 7.3)	0.5	(−8.2, 9.9)	4.2	(−4.4, 13.6)
Other European countries	−2.3	(−7.7, 3.4)	−5.9	(−13.2, 2.1)	−0.3	(−8.1, 9.2)
Non-European countries	−3.3	(−6.6, 0.2)	−1.2	(−5.9, 3.7)	−5.6*	(−10.5, −0.5)

The first section presents an analysis of linear trends, here presented as the average percentage change in self-reported offending for each 2.5 year increase in time, for each and every immigrant category, and separately for boys and girls. The second section presents percentual differences in the rate of change by immigrant staus. The third section presents percentual differences in the rate of change by region of origin.

*P ≤ .05; **P ≤ .01; ***P ≤ .001.

Differences in the rate of change were investigated by regressing the dependent variable – self-reported crime – on the time-trend, the immigrant status variable, and an interaction term between immigrant status and the time trend. A statistically significant interaction between the time trend and a given immigrant status group indicates that the rate of change in self-reported crime over time differs for the group in question by comparison with the reference group (native Swedes, i.e. youths born in Sweden to two Swedish-born parents). The results are presented in the second section of Table 3 and show that the time trend for first generation immigrant youths is negative and significantly different from zero: compared to native born youths, self-reported crime among first generation youths decreases by 6.1% (CI: −8.0, −4.1) more for each 2.5 year increase in time than for native Swedes (no significant difference in trends were found for girls in this category). The analyses have been conducted with the final years of the time series both included and excluded. A comparison of the results shows that the significantly larger relative decline in offending among first generation immigrant youth in relation to native Swedes primarily occurs during the later years of the period examined.

Differences in the rate of change by region of origin were also investigated by regressing the dependent variable – self-reported crime – on the time-trend, the country of origin variable, and an interaction term between region of origin and the time trend. These results are also presented in Table 3 and show that the there are no differences in the rate of change between Swedish-born-youth and immigrants from different regions.⁷

Finally, sensitivity analyses were performed comparing the outputs of regression models with weighted and unweighted data. The weighted analyses were mostly consistent with the main analyses, and the statistical significance of the results varied only in three cases.⁸

Discussion

Generally speaking, in Sweden, as in many other countries, self-reported crime rates have decreased over time for several different types of offending (Shannon et al., 2014; Svensson & Ring, 2007). The current study, which is based on an 18-year series of self-report surveys (1999–2017), largely confirms this picture. Our data reveal that a decreasing trend in self-reported offending is found among native Swedes, and both first and second generation immigrant youth, and also among immigrant youths from different regions of origin. This tendency appeared more pronounced among boys than girls, although no formal statistical testing was performed on gender differences in this study

The study’s results also show that self-reported crime among second generation immigrants decreases at a similar rate to that of native Swedes, while the decrease among first generation immigrants youth is faster than among youth born in Sweden, indicating an decreasing gap in offending over time. An investigation of differences in offending over time based on region of origin yielded no significant findings, suggesting that no one region is driving the overall findings from this analysis.

Previous Scandinavian and European studies on self-reported crime trends among youths have indicated that crime trends among immigrants may be declining (Balvig, 2011; Fernández-Molina & Bartolomé Gutiérrez, 2018). In one Spanish study, the results show that, following a period of decline among immigrants as well as natives, self-reported crime rates among immigrant youths have stabilized, while rates among Spanish juveniles continue to fall (Fernández-Molina & Bartolomé Gutiérrez, 2018). However, in all these studies immigrants have been treated as a single group, which has produced little knowledge about potential changes in the delinquent involvement of different immigrant groups over time.

Several explanations have been suggested for the decreasing trend in youth crime, such as changes towards a generally less tolerant attitude towards crime, increasing motivation among youths to work hard in school, increases in the level of formal control exerted over young people, and general changes in leisure time activities among young people (e.g. Svensson & Ring, 2007). In North America, studies have suggested that decreasing crime trends overlap with an increase in immigration to such a degree that the latter could be considered responsible for the decline (Sampson & Bean, 2006). Many immigrant groups are associated with having strong community ties, arguably providing protection against the criminogenic effects of social disorganization, strain, and economic and social marginalization (Wadsworth, 2010).

However, the situation in Sweden is different and how these explanations might apply to different groups of immigrants has as yet not been empirically studied in this context. The current study is primarily descriptive, and no adjustments have been made for explanatory variables. Obviously, more research is needed, from different methodological perspectives, focused on evaluating and explaining crime trends across immigrant and non-immigrant groups, as well as the potential effects of immigration on the crime drop. To our knowledge, this is the first Scandinavian self-report study to describe how crime trends differ across immigrant groups and immigrant generations.

Methodological considerations

Self-reports of delinquency have been criticized on several grounds, including the over- and underreporting of criminal behaviour and selective non-response. Regarding over- and underreporting of criminal behaviour, previous studies have indicated that youths born abroad may be more likely to under-report their crimes in crime surveys (e.g. Laajasalo et al., 2016). It is difficult to estimate the extent of these phenomena, but it is reasonable to assume that the pattern should be similar across the eight waves of the survey conducted since 1999, and thus over- and underreporting ought not to affect the validity of the findings as regards comparisons over time.

On the subject of selective non-response, it has been argued that truants may have a higher propensity to commit more offences than others, and that it is difficult to know whether the results would have been different if students absent on the day of the survey had participated (e.g. Cernkovich, Giordano, & Pugh, 1985; Shannon, 2006). However, when the total level of missing data is low, this concern may be exaggerated.

A critical question in all studies is the representativeness of the sample studied. Up until 2011, the survey was characterized by low levels of non-response in comparison with many other studies, and the representativeness of the study sample may thus be viewed as good. For the surveys conducted in 2015 and 2017, the data were weighted for non-response and the high level of agreement between unweighted and weighted analyses provides some reassurance that changes over time in the composition of the sample are unlikely to affect our findings in a major way.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Swedish ethical guidelines states that children aged 15 years or over should themselves decide whether to consent to participate in research of this kind.

2. In this study, the proportion of youths with an immigrant background is reasonably consistent with the numbers found in the general population of Swedish ninth grade youth. According to the Swedish National Agency for Education, the proportion of ninth grade youth with an immigrant background (either born abroad or born in Sweden to two foreign-born parents) in the Swedish population has increased from 14.5% in 1999 to 24.5% in 2017. The corresponding proportions in the study sample were 11.2% and 22.2%, respectively.

3. Our definition of western countries includes all the Nordic countries (except Sweden), the US, Canada, Australia, New Zealand, Belgium, the Netherlands, Luxemburg, Germany, Austria, Switzerland, France, Greece, Spain, Italy, Israel, Cyprus, Portugal, England, Scotland, and Ireland. Our definition of other European countries includes Albania, Bulgaria, Romania, Estonia, Georgia, Latvia, Lithuania, Moldova, Poland, Russia, Czech Republic, Slovakia, Hungary, Belarus, and all countries from the former Yugoslavia. The definition of non-European countries includes Latin America, Africa, Middle-Eastern countries (including Turkey) and the remaining Asian countries.

4. In the non-European category, there is a large heterogeneity in cultural and socioeconomic conditions. A sensitivity analysis was therefore performed without the East Asian countries (Japan, China, Korea, Mongolia, Hong Kong, Taiwan) included in the category; the overall result did not change.

5. As a robustness test for the main analysis, we also created a frequency score of the overall offending scale, which overall yielded the same findings.

6. The results for the western countries region may be underpowered, due to the relatively small number of subjects per time point.

7. A complementary analysis was performed without second generation immigrants included in the ‘born in Sweden’ category and the results did not change.

8. The weighted analysis showed that for each 2.5 year increase in time, youths with two immigrant parents presented a decrease in self-reported crime that was 2.7% greater than that found among natives, while among boys with one immigrant parent, the decrease in self-reported crime was 2.7% less than that found among native boys. For girls born abroad, the decrease in self-reported offending was 6.3% greater than for native girls. These three findings were not significant in the unweighted analysis.