Youth protection and the benefits affair

Publication date: 30-1-2023 11:59Youth protection and the benefits affairQuantitative research into child protection measures among children of victims of the benefits affairAbout this publicationAt the request of the Justice and Security Inspectorate (IJenV), CBS investigated whether families who were victims of the benefits affair have come into disproportionate contact with child protection measures. This looked at integrally treated victims who were duped between 2012 and 2018. The group of victims was first compared with a broad comparison group. This consists of households that used childcare allowance as victims during the same period, but were not affected. This comparison shows that victims of duping were more likely to face child protection measures in the family. The broad comparison group also includes families that differ from the group of affected households in relevant characteristics and therefore already had another chance of coming into contact with a child protection measure, regardless of the benefits affair. That is why, in a second step, a narrow comparison group was put together that, on relevant background characteristics, looks like the group of affected households, but which is not duped. This allows the (possible) effect of duplication to be identified. This study found no evidence that victims came into contact with child protection measures extra often as a result of duping.Erratum (January 30, 2023)In the summary, a subheading has been brought in line with the conclusions of the study. The content and conclusions of the report have not been amended. The fourth subheading in the summary has been amended from “Duping does not increase the risk of child protection measures” to “No evidence found that, on average, duplication led to more child protection measures”.Summary ReasonEarlier, the Central Bureau of Statistics (CBS) published figures on displacements among victims of the benefits affair. These figures prompted the Justice and Security Inspectorate (IJenV) and the Health Care and Youth Inspectorate (IGJ) to investigate how the youth protection chain dealt with families affected by the benefits affair. As part of this study, IJenV asked CBS to identify whether affected families were more likely to face child protection measures than other families. It was investigated whether affected families were more likely to face child protection measures before being duped, and it was investigated whether affected families were more likely to face child protection measures as a result of duping.Research designTo investigate whether the victims of the benefits affair have been disproportionately affected by child protection measures, the affected families are compared with unaffected families. In a first comparison, the affected families are compared with a broad comparison group before they are duped. This consists of households that used childcare allowance as victims during the same period, but were not affected. This allows the question to be answered whether victims of duping come into contact with child protection measures more often than non-victims. However, this broad comparison group also includes families that differ from the group of affected households in relevant characteristics and therefore already had another chance of coming into contact with a child protection measure, regardless of the benefits affair. That is why, in a second step, a narrow comparison group was put together that, on relevant background characteristics, looks like the group of affected households, but which is not duped. This allows the (possible) effect of duplication to be identified.Victims of duping often face child protection measuresA comparison of the victim group with the broad comparison group shows that the victims were more likely to face child protection measures in the family for duping. These differences can be explained by the specific characteristics and circumstances of affected families before they were affected. For example, families who are affected are more likely to be single-parent families and have a lower household income than unaffected families.No evidence found that, on average, duping led to more child protection measuresThe victim group was then compared with the narrow comparison group. This study found no evidence that victims came into contact with child protection measures extra often as a result of duping. After duping, victims are not significantly more likely to face child protection measures in the family than they are duped. Nor do they have to deal with child protection measures more often than the narrow comparison group.Possible follow-up researchThis is the first quantitative study into whether child protection measures are disproportionate among victims of the benefits affair. The conclusion is that, on average, victims often receive child protection in the family, but this study found no evidence that they are extra often imposed on child protection measures in the family as a result of duping. Although the research answers a number of questions, follow-up questions remain:

Due to the available data, not all registered victims could be included in this investigation. This way, only victims whose files have been treated integrally could be included. Only of these victims has provided CBS with a date of duping. If more data becomes available, the analyses could be repeated.
There was also no information available for this study about, for example, the degree of duplication. It is possible that whether a family faces child protection measures matters whether it was a recovery of, for example, 1.5 thousand euros or 50 thousand euros. It can also matter whether and when these “debts” were actually recovered.
It is also possible that the effect of duping depends on the exact situation of a victim. This has been studied for a number of subgroups of victims (low incomes, self and/or parents not born in the Netherlands and single-parent families). For these subgroups, no evidence was found that they were disproportionately affected. The fact that, on average, victims did not have to face additional child protection measures as a result of duping does not preclude the fact that there are individual victims who got into such trouble as a result of duping that child protection measures had to be used.
This study only looked at the relationship between abuse and child protection measures. Other issues that may be the result of duping have not been investigated. This includes effects on mental well-being and stress, physical health, relationships and the financial situation of victims.

IJenV's in-depth qualitative research, which will include interviews, may find starting points for possible follow-up research.1. IntroductionThe Central Bureau of Statistics (CBS) has October 2021 and May 2022 figures published on relocations among victims of the benefits affair. These investigations prompted the Justice and Security Inspectorate (IJenV) and the Health Care and Youth Inspectorate (IGJ) to investigate how the youth protection chain dealt with victims of the benefits affair.1) As part of this investigation, IJenV asked CBS to investigate whether families that were victims of the benefits affair have come into disproportionate contact with child protection measures. A child protection measure is a measure that requires the court to eliminate the threat to a child's safety and development. This can be done by means of undersupervision (the court limits the parental authority, but the parents remain responsible for parenting) or a custody measure (parental custody is terminated and transferred to a guardian). To answer the question whether victims of the benefits affair have come into disproportionate contact with child protection measures, the study compares three groups:

The victim group: households of people who are registered with the Implementation Organization for Recovery of Allowances (UHT) as victims of the benefits affair.
Broad comparison group: households that made use of childcare allowance in the same period as the affected households, but are not registered with UHT as affected by the benefits affair.
Narrow comparison group: a subgroup of households from the broad comparison group that resembles the group of affected households on relevant background characteristics. This group is compiled using a statistical matching technique.

This report describes the results of this research. Both IJenV and the CBS attach great importance to external quality assurance and contradiction when investigating this socially relevant, but also highly sensitive subject. For this research, a guidance committee (with internal and external methodologists and experts) was therefore set up to contribute ideas about the study design, analyses and results during the study. See Annex 1 for the composition of this guidance committee. This study focused on child protection measures in general and not on removals specifically. This is in line with the inspections' research question.2) It was also not possible to investigate disproportionality in removals separately with the available data.3) Out-of-home placements are defined as a child protection measure combined with youth assistance with residence. Replacements are therefore part of this study into child protection measures.4) The next chapter discusses the group of victims. The following chapter describes how the broad comparison group is composed. Chapter 4 describes the characteristics associated with the risk of being duped and with the risk of being imposed or not being imposed a child protection measure. The fifth chapter describes how the narrow comparison group was composed by matching. In chapter 6, the research question is answered. The final chapter contains the conclusions.1) https://www.inspectie-jenv.nl/Publicaties/plannen-van-aanpak/2022/01/17/aankondiging-onderzoekzprogramma-hoe-ging-de-jeugdbescherming-om-met-gezinnen-gedupeerd-door-de-toeslagenaffaire.2)https://www.inspectie-jenv.nl/Publicaties/plannen-van-aanpak/2022/01/17/aankondiging-onderzoekzprogramma-hoe-ging-de-jeugdbescherming-om-met-gezinnen-gedupeerd-door-de-toeslagenaffaire.3) See also https://www.cbs.nl/nl-nl/achtergrond/2022/30/onderzoeken-naar-gedupeerden-toeslagenaffaire.4) For more information about youth protection and the coincidence with youth assistance with residence, see http://opendata.cbs.nl/statline/#/CBS/nl/dataset/85101NED/table?dl=7098F%20.2. Victim groupIn the Netherlands, parents can apply for childcare allowance from the Tax Authority/Surcharges. This is an allowance for the costs of registered childcare. There are several conditions for receiving such a surcharge.5) For example, a person and any partner must work, attend an education or integration course, or be on a journey to work. The applicant and the child for whom an allowance is requested must also be registered with the municipality at the same address.6) Finally, there is a mandatory personal contribution. Part of the costs of childcare must therefore be paid by the applicant himself. It is possible that an applicant has received too much childcare allowance. This can happen, for example, if an applicant's income or household situation changes.7) The overpaid amount of childcare allowance is then recovered by the Tax Authority/Allowances. The Tax Authority/Surcharges checks the requested and provided childcare benefits with the aim of determining whether an applicant is entitled to childcare allowance and has received the correct amount. Due to the processes and practices that the Tax Authority/Surcharges has worked with in the past to detect fraud in childcare benefits, applicants have been wrongly identified as fraudsters. The handling of fraud risk signals was also careless and sometimes not lawful.8) We worked with a “all-or-nothing” approach where, when identifying imperfections (such as a personal contribution not fully paid or a missing signature), the allowance was recovered from the parents for the entire year, with a 80/20 approach so that there was a group approach and could expose innocent parents to intensive controls and with a “intentional or gross negligence” approach as a result of which no appropriate personal payment arrangements were made. (9) These applicants faced a strict recovery policy, which could have major financial consequences for the applicant and could go hand in hand with uncertainty about surcharges and recoveries. This is referred to as the benefits affair.10) This problem has occurred since approximately 2004. Possible victims of the benefits affair can report to the tax authorities. In 2020, the Implementation Organization for the Recovery of Surcharges (UHT) was established to repair what did not go well in the past. UHT provided CBS with a file containing persons who were registered with UHT as victims on 1 July 2022. The privacy of the victims is hereby guaranteed. See appendix 2 for more information about how CBS deals with privacy in this study (and other studies). The file provided by UHT is the basis of this research. This file contains more than 25 thousand identified victims. A parent/applicant is in this file when they have received a formal order from the tax authorities informing them that he/she is a victim and/or has received an amount of 30 thousand euros. In December 2020, the cabinet decided that all affected parents should receive 30 thousand euros. This is referred to as the Cathuis Childcare Allowance Scheme. Individuals are eligible for this amount if they meet a number of conditions:

they reported as victims
the childcare allowance was unfairly stopped between 2005 and 2019, or someone had to unfairly pay back childcare allowance during that period
it involved an amount of at least 1,500 euros.11)

Based on a comprehensive treatment of the victims' files, UHT is investigating exactly what went wrong with the childcare allowance and when. In principle, UHT works in the comprehensive treatment of files on the basis of the order of registration.12) Of the 25 thousand victims who are in the file that UHT provided to the CBS, approximately 7.5 thousand victims have received comprehensive treatment.13) A date is known for the fully treated victims on which a correction with regard to childcare allowance was made to the Tax Authority/Surcharges, which, after full treatment, it appears that these has been unjustified. This date was manually added to the file after full treatment and provides an indication from when the abuse started. (14) This date is not available for victims who have not (yet) fully treated. The only date that is known for this group of victims is the date of the first collection with regard to childcare allowance in the Tax Administration/Allowances system. Without the full treatment of the files, however, it cannot be said whether this date relates to an unjust recovery (and thus the benefits affair) or to a fair correction because, for example, the applicant's income or household situation has changed.15) For victims whose files have not been treated integrally, there is therefore no good information available about when the abuse occurred. In the present study, the main question is whether parents who were victims of the benefits affair subsequently faced a child protection measure more often than comparable unaffected parents. The moment of duping is therefore important information for this investigation. That is why this study only focuses on the integrally treated victims. CBS has data on youth protection since 2011, which means that this study can only look at people who have been affected since 2012. Indeed, the study also looks at child protection measures in the family for duping. Currently, CBS has data on youth protection up to and including 2021. The investigation decided to see whether there was a child protection measure among families 3 years after the duping (see also chapter 6). This means that the last cohort that can be included in this study are people who were duped in 2018 (for these people, it can be determined whether there was a child protection measure in the household in 2019, 2020 or 2021). Of the 7.5 thousand fully treated victims, just under 5 thousand (16) people were affected between 2012 and 2018.Table 2.1. Fully treated victims by year in which they were duped Total 2012 2013 2014 2015 2016 2017 2018Number 4885 1095 840 925 890 530 330 270The table above shows that the number of (integrally treated) victims in the later years (from 2016) is significantly lower than in previous years (from 2012 to 2016). An adjustment to the recovery policy around 2016 will probably have played a role in this. For example, surcharges could be stopped before evidence was requested and reviewed.17) Until mid-2016, for example, the use of a so-called soft stop was used: the tax authorities/Surcharges stopping a surcharge in the current tax year prior to assessing or processing the application in order to prevent payment of possible unfairly awarded surcharges. (18)5) https://www.belastingdienst.nl/wps/wcm/connect/nl/kinderopvangtoeslag/content/kan-ik-kinderopvangtoeslag-krijgen6) This is not necessary if there is co-parenting. Co-parenting occurs when the child lives with an applicant for at least three full days a week and three full days a week with the other parent, or if the child lives with the applicant every other week and the other parent for one week.7) https://www.belastingdienst.nl/wps/wcm/connect/nl/kinderopvangtoeslag/kinderopvangtoeslag.8) onderzoek-pwc-effecten-fsv-toeslagen.pdf (overheid.nl).9) https://www.tweedekamer.nl/sites/default/files/atoms/files/20201217_eindverslag_parlementaire_ondervragingscommissie_kinderopvangtoeslag.pdf.10) This report uses the common term allowance affair. Other names include the childcare allowance affair and the benefits scandal.11) For more information, see: https://services.belastingdienst.nl/toeslagen-herstel/catshuisregeling-kinderopvangtoeslag/.12) While this is the aim, UHT provides in the most recent Progress report on recovery surgery surcharges indicates that this is made difficult due to appeals not taking timely decisions.13) Unfortunately, it is not possible for the CBS to properly investigate whether there is a possible bias due to victims who have been treated integrally and due to victims who have not been treated integrally. The CBS only has data on whether or not you were duped. The CBS has no information about the level of duping. In other words, about the amount for which a parent has been duped. The latest figures show that compensation and compensation amounts realized for more than half of the parents amount to up to 30 thousand euros and less than half of them amount to 30 thousand euros or more. The average amount is about 43 thousand. At the individual level, therefore, no data is available on the CBS. It has been checked whether there are major differences between victims who have been treated integrally and those who have not been treated integrally in terms of origin (main characteristic of duping, see later in this report). There are no major differences; in both groups, approximately 1 in 3 people has a Dutch origin, but among integrally treated people, this is slightly lower (29 percent) than among people who are not treated integrally (32 percent) .14) This is an indication of the start of duress. Indeed, this is the first moment in which a correction was found in the Tax Authority/Surcharges system that, with today's knowledge, requires compensation. Of course, for the victims themselves, dupe potentially only starts after the first letter hits the mat or perhaps after several (rejected) appeals or when the “debts” are actually recovered. In other words, it is impossible to determine the exact moment of abuse per victim based on administrative data. This date provided by UHT is the best available indication that CBS could use for this study.15) In the file provided for the first descriptive publication There was no information about integrally treated victims in this research project. This date was therefore the only available information for CBS at that time about the time of dupering.16) A very small part of the fully treated victims from the years 2012 to 2018 could not be included in the investigation after all. For example, victims who lived in institutional households were removed because the household characteristics of these people could not be meaningfully included in the study. Victims who, for example, could not be found in the CBS household records could not be included in the investigation either.17) https://ophttps://open.overheid.nl/repository/ronl-1c359fdc-4398-44e2-9969-1f353c0c481d/1/pdf/rapport-caf-adr.pdf 18) https://www.tweedekamer.nl/sites/default/files/atoms/files/20201217_eindverslag_parlementaire_ondervragingscommissie_kinderopvangtoeslag.pdf.3. Compiling a broad comparison groupAs indicated earlier, this study looks at a so-called broad comparison group. These are households that made use of childcare allowance in the same period as the affected households, but are not registered with UHT as affected by the benefits affair. The difference with the narrow comparison group (see next chapter) is that the households in the broad comparison group may differ from the victim group on relevant background characteristics. The extent to which these groups have come into contact with child protection measures may therefore differ, without this being a result of the benefits affair. CBS has data on childcare allowance since 2007. This means that data on childcare allowance for the 2012 cohort is available for a total of 6 years: the year of duping (2012) and the five years before duping (2007 to 2011). To compile the broad comparison group, all people who received childcare allowance during this period (t-5 to t) were the first to select. Then, per calendar year, it was checked whether or not people had received childcare allowance. Each person has a certain pattern of receiving childcare benefits (for example, did not receive childcare allowance in 2007 through 2009 and did receive childcare benefits from 2010 through 2012). All non-victims who had the same pattern as at least one victim in these years were selected for the broad comparison group. In this way, people are selected for the broad comparison group who received childcare allowance in exactly the same period as victims. This was then done for all cohort years (2012 to 2018): we looked at who received childcare allowance in the period t-5 to t and when and then all non-victims were selected for the broad comparison group who received childcare allowance in exactly the same years as the victims from that cohort year. In this way, a broad comparison group is compiled per cohort year based on patterns in applying for childcare allowance. With this demarcation, it is possible that people will end up in the broad comparison group over several years. For example, it is possible that a person may have the same pattern of childcare allowance received as a victim in 2012 as well as a victim in 2013, who will initially be selected in the broad comparison group for both years based on the above delineation. Because the assumption of the analysis techniques used (see chapter 4) is that observations are independent (in other words, people only appear once in the file), people in the broad comparison group who occur in multiple cohort years are randomly assigned to one cohort year. The following table shows the size of the groups per cohort year (duping year for the victims and selection year for the unaffected). The above approach results in a broad comparison group of at least 147 thousand people (2015) to a maximum of 286 thousand people (2018). The fact that the size of the broad comparison group has increased in recent years is because the use of childcare allowance has also increased during this period.19)Table 3.1. Number of victims and broad comparison group by dupering/selection yearVictim group 20 Broad comparison group 21 Total 4 100 1 259 825 Years 2012 715 167 650 Years 2013 740 152 250 Years 2014 815 147 720 Year 2015 800 147 015 Year 2016 485 163 880 Year 2017 290 194 900 Years 2018 250 286 41019) https://www.cbs.nl/nl-nl/nieuws/2022/28/ruim-1-miljoen-kinderen-met-kinderopvangtoeslag.20) There are victims who did not apply for childcare allowance during this period (the year of duping and/or the 5 years before duping). There are therefore victims who, for example, were duped in 2012 and who did not receive childcare allowance in the period 2007 to 2012. It is therefore likely that these people were duped as a result of an application made before this period. On average (over all cohort years), this is the case with approximately 16 percent of the victim group. Victims with such a different pattern of applying for childcare allowance were excluded from the analyses. The reason for this is that no good comparison group can be put together for this group.21) Just like the victims (as indicated in the previous chapter), this also included people who, for example, live in institutional households or who did not link to the necessary files to carry out the analyses. It also happens that people who are selected in the above way for the broad comparison group have a partner or other household member who has become a victim of the benefits affair. Such households were also removed from the broad comparison group so that the broad comparison group does not include people who are themselves victims of the benefits affair or whose household is a victim. To do this, all registered victims were looked at (and not just the fully treated registered victims).4. Relevant background featuresTo compile the narrow comparison group, a statistical matching technique is used to search for households in the broad comparison group that look like the group of affected households on relevant background characteristics. This is to ultimately answer the question whether child protection measures are disproportionately common among children of victims of the benefits affair compared to a group of similar, but not affected, households. A background characteristic is relevant to this matching technique if it relates to both the risk of being duped and the chance of coming into contact with a child protection measure. After all, suppose that there are characteristics that are significantly related to the risk of being imposed a child protection measure, but that there is no difference between victims and non-victims for this characteristic. This attribute then does not have to be included in the matching. Indeed, matching ensures that any important differences between the groups (whether or not duped) are removed in order to draw correct conclusions, but in that case, the groups are already the same on that characteristic. The same applies the other way around: suppose that a characteristic is related to the risk of being duped, but not with the risk of being imposed a child protection measure. Then this characteristic is not relevant to include when compiling a good narrow comparison group. Indeed, differences in that characteristic cannot be a reason for any differences between victims and non-victims in child protection measures. With the help of internal and external experts, the following general list of background characteristics that could be relevant was first drawn up (see appendix 1 for the guidance committee of this study). It is important that all characteristics were identified at a time before the time of dupering/selection. Indeed, for a pure comparison, it is important that the characteristics are not already affected by the duplication. Characteristics in 4 different domains are investigated (for the definition and operationalization of these characteristics, see Appendix 3) :1. Demographic characteristics:

Applicant's gender
Applicant's age
Applicant's age at birth of first child
Applicant's marital status
Applicant's country of origin and country of origin
Applicant's residential province
Applicant's municipal residency degree

2. Household situation:

Household type
Number of children in the household
Age of youngest child in the household
Age of oldest child in the household
Application for childcare allowance for a non-legal child
Number of years of applying for childcare allowance
Household relocations
Partner changes

3. Education and socio-economic situation:

Applicant's highest educational level
Children in the household who are early school leavers
Property for sale or rent (yes/no housing allowance)
Household income
Household assets
Main household income source
Household health insurance defaulter
Natural Person Debt Restructuring Act - Household Process

4. Care use:

Mental Health Care Received (GGZ)
Use psychotropic drugs with household members
Use medication for household addictions
WMO household use
Mild mental disability applicant and possible partner
Registration as a household suspect

The above characteristics have been identified whether they are related to, on the one hand, the chance of becoming a victim of the benefits affair and, on the other hand, the chance of being imposed a child protection measure. (22) Whether or not the victim of the affair benefits is derived based on the file of registered victims provided by UHT. Whether or not a child protection measure is imposed, all minor children in the applicant's household in the year before dupering/selection will be considered. It is examined whether at least one of these children was imposed a child protection measure in the year before dupering/selection. The tables in the following sections show the distribution and coherence for all characteristics for people who have or have not been duped and for people who have received a child protection measure and for people who have not. For this descriptive analysis, all dupering/selection years were taken together. In order to show the distribution of characteristics between the different groups (whether or not affected and whether or not a child protection measure in the family), cross-tables were drawn out (the tables show the percentages). In addition, bivariate logistic regressions were carried out (this because the dependent variables, duress and the risk of being imposed a child protection measure in the family) are binary. For more information on logistic regression, see Hosmer Jr et al., 2013) .All differences as shown in the tables are statistically significant (this means that the differences found are unlikely to be random) .23) It should be noted that the groups in this study are large: in total, the analyses were based on more than one million people (more than one million non-victims in the broad comparison group) and more than four thousand people in the broad comparison group group of victims). In general, the larger the research groups, the greater the chance of significant results. In this case, it is therefore more relevant to look at the strength of the coherence than at the significance. Logistic regression analysis does not provide the proportion of explained variance (R2) as defined for interval or ratio variables in a linear model. However, there are several pseudo-R2 measures that are comparable to the R2 from linear regression analysis. In this study, the McFadden R2 (1974) is shown as a pseudo-R2 size. The McFadden R2 can take a value between 0 and 1, with a higher value usually representing a better model (more coherence between the characteristic and the risk of suffering or the risk of being imposed on a child protection measure). There is debate about when a model, based on the pseudo R2, can be characterized as good. It is assumed that when the pseudo R2 is between 0.2 and 0.4, the model is in a good fit with the data (Simonen & McCann, 2008). The tables below include the results of bivariate analyses (so there are only two variables in the model: the characteristic on the one hand and duress or child protection measures on the other). As a result, the pseudo R2 is lower than when multiple features are included in the model (see also later in this chapter). In any case, the closer the pseudo R2 is to 0, the less the characteristic can explain the variance in the risk of abuse or child protection measures and the higher the pseudo-R2, the better. Most applicants for childcare allowance have not become victims of the benefits affair and most applicants will not face child protection measures in their families. As the tables below show, both groups (no duped benefits affair and no child protection measures) are almost the same (also in terms of characteristics).4.1. Demographic featuresAs the table below shows, origin in particular has a clear correlation with the chance of being duped. Of people who were not affected by the benefits affair, 78 percent were born in the Netherlands themselves, as were both parents. Among those affected by the benefits affair, this is 29 percent. It is also known that this characteristic played a role in the benefits affair.24) Surinamese applicants and applicants from the Dutch Caribbean in particular seem strongly overrepresented in the group of victims. This also applies, to a slightly lesser extent, to applicants with a Turkish or Moroccan background. Although origin is very closely related to the risk of being duped, this characteristic is less strongly related to the risk of being imposed a child protection measure in the family. The age at birth of the first (legal) child is also strongly related to dupe. More than half of the victims were relatively young (under 25 years old) when the first child was born. Among non-victims, this is 12 percent. Age at birth of the first child is also quite strongly related to the risk of child protection measures being imposed in the family (people who had a child at a relatively young age have a higher chance of coming into contact with child protection measures). Applicants without a partner (married or unmarried) are also overrepresented in the group of victims and in the group of people who come into contact with child protection measures. Table 4.1 also shows that victims are more likely to be women, are relatively young, more often live in South Holland and in very urban areas. It is important to realize that these are the results of bivariate analyses. These results may be influenced by any correlation between the background characteristics (composition effects). For example, a relatively large number of victims live in South Holland and in highly urban areas. It is known that people with a migrant background, who are overrepresented in the group of victims, live relatively often in such areas. In other words: the fact that relatively many victims live in South Holland/highly urban areas, for example, may be because many victims with a migration background are duped. The analyses as shown in this table are therefore purely descriptive analyses. In order to better understand whether and to what extent characteristics are associated with the risk of suffering and the risk of child protection measures being imposed in the family, multivariate analyses should therefore be considered (see section 4.5). In terms of gender, age and region of residence, there are no major differences between people who have and people who have not come into contact with child protection measures in the family.4.2. Household situationWhen it comes to the household situation, there are also characteristics that are significantly related to the risk of being duped and to the risk of being imposed a child protection measure in the family. This is especially true for household type. Bivariate household type is relatively closely related to the risk of being duped: applicants who are not affected by the benefits affair are often in households that are characterized as “a couple with children” (79 percent) in the year before duping. Victims of the benefits affair are relatively often in a single-parent household (this is the case for about half of the victims). These differences can also be seen when looking at the risk of being imposed a child protection measure. Furthermore, whether or not to apply for childcare allowance for a non-legal child is (relatively) strongly related to the chance of being imposed a child protection measure in the family: 17 percent of applicants who faced a child protection measure in the family have applied for childcare allowance for a non-legal child (for example, for a child of a new partner). Almost none of applicants who were not affected by a child protection measure had applied for childcare allowance for a non-legal child. This characteristic appears to be less relevant to the risk of suffering. Furthermore, victims were less likely to receive childcare allowance for only one year (18 percent) compared to non-victims (31 percent). With regard to the other characteristics, there are no major differences between victims and non-victims and people who have or have not received child protection measures in the family.4.3. Education and socio-economic situationEducation and socio-economic situation (bivariate), as shown in the table below, are clearly related to the risk of being duped and to the risk of being imposed a child protection measure. For example, those affected by the benefits affair are less educated than non-victims. Applicants who face a child protection measure in the family are also more often lower/less educated than applicants who do not have to deal with it. The fact of whether or not people live in a rental home and receive housing allowance is also (bivariate) quite strongly related to the risk of suffering. Just under half of the group of victims studied lives in a rental house and receives housing allowance. This is less than 10 percent among non-victims. In fact, they are more likely to live in their own home (77 percent of the non-victims live in a purchased home compared to 33 percent of the victims). Similar differences are also visible when applicants who did have to deal with child protection measures in the family are compared to applicants who did not. Household income is also strongly related to the risk of being duped and to the risk of having been imposed on child protection measures.25) Both those affected by the benefits affair and those affected by child protection measures are highly overrepresented in the lowest income brackets. In line with this, the table below shows that both victims of the benefits affair and people who have experienced child protection measures in the family relatively often have negative assets and are relatively less likely to have work as their main source of income. Finally, the presence of a health insurance defaulter in the household is quite strongly related to the chance of becoming a victim of the benefits affair: where this is about a quarter among the affected applicants. is, this is 2 percent among non-victims. (26) Such differences are also found, to a lesser extent, in terms of the risk of being or not being imposed on child protection measures in the family.4.4. Finally, when looking at health care use, it can be seen that around half of the applicants who have received child protection measures in the household have received mental health care in the household. For people who have not been in contact with child protection measures, this figure is less than a quarter. Table 4.4 also shows that applicants who have been imposed on child protection measures in the family are more likely to have WMO users in the household (13 percent) than applicants who have not experienced child protection measures (2 percent). Finally, victims are more likely to have people in the household who are suspected of a crime (18 percent) than non-victims (5 percent). These differences are also visible when looking at child protection: people who had child protection measures in the household in the year before dupering/selection have suspects in the household more often (a quarter) than people who have not been in contact with child protection measures (5 percent).4.5. Key features of abuse and child protection measuresThe above mentioned characteristics show whether and, if so, to what extent they are related to the risk of being duped and to the risk of being imposed on child protection measures in the family. The above analyses are based on bivariate analyses. This means that each analysis always includes only two variables (the different characteristics on the one hand and the risk of being duped or being imposed on child protection measures on the other). However, some of the characteristics are highly interrelated. For example, it was seen that victims were more often not born in the Netherlands (as were both parents) and were more likely to live in highly urban areas. However, these characteristics are interrelated: on average, people with a migration background live more often in highly urban areas. As a result, such bivariate analyses can provide a distorted picture. For example, the idea may arise that urbanity is related to the risk of suffering, when in reality it is not because of the degree of urbanity, but because people who are not born in the Netherlands live more often in urban areas. It is also possible that two (or more) characteristics measure more or less the same thing. Think of wealth and household income that are interrelated and are both a measure of financial well-being. It is therefore important that the above analyses are expanded to include models that include multiple characteristics at the same time. In this way, a better selection can be made of characteristics associated with the risk of being duped and with the risk of being imposed a child protection measure (taking into account the other characteristics). In order to make a final choice for characteristics that should be included when compiling the narrow comparison group, in addition to substantive considerations, it is necessary to Akaike information criterion (AIC) used.28) This measure is often used to assess the quality of statistical models (compared to other models on the same data set). The lower the AIC score, the better the model. The big advantage of the AIC measure is that it takes into account the number of variables included in the model (since a model with more variables is usually better than a model with fewer variables). This is because it prefers models that are able to explain as much variance as possible with as few variables as possible. The following tables show the multivariate logistic regression model for duplication (Table 4.5.1.) and child protection measures (Table 4.5.2.) (taking all cohort years together, as in the analyses above). The model with origin, age at birth, first child, number of years of childcare allowance claim, household income and the presence of a defaulter showed the most appropriate multivariate model of household health insurance. In other words, these were the characteristics that were most strongly associated with multivariate duplication (taking into account the other characteristics). People who themselves and/or whose parents were not born in the Netherlands, who were relatively young at the birth of their first child, who applied for childcare allowance for longer, who have a lower household income and who had a household health insurance defaulter in the year before dupering/selection are more likely to be victims of the benefits affair.The above model of characteristics associated with the risk of duping reflection describes (the consequences of) the risk model classification was created by the Tax Authority/Surcharges were used to detect fraud. Child protection measures may have other determinants. That is why a separate model is being drawn up for this. (29) Child protection measures are used when a children's judge believes that the child's development and/or safety at home is threatened. The multivariate logistic regression model shows that applying for childcare allowance for a non-legal child is strongly related to the chance of being imposed a child protection measure. The presence of a suspect of a crime in the household and mental health use in the household also significantly increases the chance of being imposed on child protection measures. Furthermore, single-parent families (compared to couples with children), applicants with more children in the household, applicants who applied for one year of childcare allowance instead (of several years), low-educated applicants and applicants with a lower household income are more likely to come into contact with child protection measures.30)22) The analyses were performed with the statistical package R (version 4.1.3.) .23) All chi2 tests had a p-value < 0.001.24) See, for example, the report from the Data Protection Authority, which describes how (dual) nationality has been processed by the Tax Authority/Surcharges in the past for applicants for childcare allowance. Also Amnesty Hand it College for Human Rights have published about this. The Secretary of State concerned has also recognized that there has been institutional racism at the Tax Authority/Supplements.25) This feature is also characteristic known that the Tax Authority/Surcharges looked at this and that in the model used, low incomes received a higher risk score and higher incomes received a lower risk score.26) This characteristic also includes known that, as of January 2014, it was included in the risk classification model used by the Tax Authority/Surcharges. After the first quarter of 2019, this indicator is no longer in use.27) This attribute is only available for people who were duped/selected after 2016 (WMO use in the dupering/selection year could only be considered for this group) .28) More than 100 regression models were compared for both duping and child protection measures. The general procedure that has been followed is that, for both duplication and child protection measures, the characteristics that were bivariously related were selected (for duplication, for example, origin and, for child protection measures, application for childcare allowance for a non-legal child). The other features were then added in different combinations, after which a choice was made to add a feature and then several combinations were explored, etc. This was done together for all cohort years and then the regression models were run separately for all cohort years.29) As mentioned earlier, this study ran, assessed and compared many different models to arrive at this model. Of course, all characteristics were considered here. For example, origin has also been considered for child protection measures. However, this was not a determining variable for child protection. Compared to the model shown in this study, the explanatory power (MacFadden R2) increased by 0.002 if this feature was included. This is in line with the fact that the relationship between child protection measures and the origin of the bivariate was relatively limited (see Table 4.1.). The purpose of these models is not to explain child protection as optimally as possible, but to identify the most important characteristics in order to establish a narrow suitable comparison group.30) All analyses were carried out per cohort year. For the latest cohorts (from 2016), Wmo could also be included in the models. This characteristic also appeared to be important for child protection: people who had Wmo in the household for dupering/selection are more likely to also have child protection measures in the family. For the matching model (see next chapter), this characteristic appeared to have too little added value to be included in the matching process.5. Compiling a narrow comparison group5.1. Introduction to matchingMethodologically, the ideal way to establish a causal relationship (as in this case between abuse and child protection measures) is a classic experiment. This means that people random be assigned (by chance) to an experimental group (in this case, the victims) or to a control group (in this case, the non-victims). The idea behind this is that differences in the dependent variable (in this case child protection measures) between the groups are then only due to the intervention (in this case due to duress) and not due to any other differences between the groups. In this case, that would mean that you would or would not have let people become victims of the benefits affair on the basis of chance. Of course, this is not in line with practice. It is known that the Tax Authority/Surcharges used certain characteristics in the risk selection model during the benefits affair. The previous chapter describes how the victims differ from those who have not been affected. For example, victims are more likely to have a lower household income. As shown in the previous chapter, such characteristics are also related to the risk of child protection measures being imposed in the family. In order to find out whether the benefits affair played a role in the risk of child protection measures being imposed (i.e. whether child protection is disproportionately common among those affected by the benefits affair), it is therefore not useful to compare the victims with all non-victims. You want to compare the victims with similar unvictims. Only then can meaningful conclusions be drawn about whether child protection measures are disproportionately common among children of victims of the benefits affair. A common way of comparing groups when a classic experiment is not possible is matching (Rosenbaum & Rubin, 1983). Using this technique, you select non-victims who look like the victims. In other words, you put together a control group of non-victims who are similar to those affected in terms of relevant characteristics. In this way, it is still possible to make a quantitative statement about whether child protection measures are disproportionate among those affected by the benefits affair. In this study, this control group is called the narrow comparison group (in contrast to the broad comparison group that includes all non-victims). More information about this technique can be found in Caliendo & Kopeinig (2008), among others.5.2. Matching in this study5.2.1 Selection characteristicsIn order to put together a good narrow comparison group by means of matching, it is important that the correct characteristics are looked at. There are several conditions attached to this. For example, it is important to look at the characteristics for dupering/selection. In other words, both groups (the victims and the narrow comparison group) must be similar in terms of dupering/selection characteristics. Indeed, the ultimate goal is to be able to make statements about whether the benefits affair played a role (in the risk of child protection measures being imposed) and in order to do this, both groups must be similar in characteristics that the benefits affair has not (yet) had any influence on. This is also the reason that the previous chapter looked at characteristics in the year before dupering/selection. In addition, it is important to look at characteristics that are related to both the risk of being duped and with the risk of being imposed a child protection measure. Only these characteristics are relevant when setting up a good narrow comparison group (see also previous chapter). Furthermore, it is important that all characteristics that are potentially relevant are included in the research. If characteristics that are important for the risk of suffering and the risk of child protection measures being imposed are not included, a less good narrow comparison group can be drawn up.31) The previous chapter described which characteristics (in the multivariate logistic regression model) are most strongly related to dupe and which to the risk of child protection being imposed. Based on these insights, the characteristics that are included in the matching were selected to create a suitable narrow comparison group. More specifically, this study includes the following characteristics in order to establish a suitable narrow comparison group. In terms of demographic characteristics, the origin and age at birth of the first child are considered. With regard to household situation by household type, applying for childcare allowance for a non-legal child and the number of years that childcare allowance has been applied for. In terms of socio-economic position, household income, applicant's education and the presence of health insurance defaulters in the household are included. Finally, mental health use in the household and the presence of suspects of crime in the household are included in the model for compiling the matching group. The following table shows the model.5.2.2. Matching methodThere are several ways in which these attributes can be used to create a suitable narrow comparison group. One way to do this is by using exact matching. This means selecting someone from the broad comparison group for the narrow comparison group who has exactly the same characteristic as a victim. For example, there are a relatively large number of victims with a Surinamese origin. Bee exact matching ensure that you also select a non-victim with a Surinamese origin for each victim with a Surinamese origin. Another way to include the features is by using propensity score matching. The propensity score of an applicant shows the chance of being duped given their characteristics. There are then also various methods to use, based on the propensity scores selecting an unaffected person for each victim. A commonly used method is nearest neighbour matching. This means that from the broad comparison group, the person whose propensity score is as close as possible to the propensity score of the victim. In other words, for each victim, someone from the broad comparison group is selected who has as much as possible (but is not) the same chance of being duped, given the characteristics that are being looked at. A disadvantage of nearest neighbour matching is that the person with the closest propensity score, may still be a person who does not really look like the victim. This is because this person does have the closest propensity score, but the difference between the two scores can be quite large. One way to deal with this is caliper matching. With this technique, you determine in advance how big the difference between the propensity scores may be between the victim and a person from the broad comparison group to select this person for the narrow comparison group. Once again, the person is chosen with the propensity score closest to the victim's score. One advantage of this method is that it can provide better matches. A disadvantage of this method is that fewer matches may be found. Indeed, it is possible that a victim cannot find anyone in the broad comparison group that has a propensity score has it sufficiently close to the propensity score The victim lies in selecting this person for the narrow comparison group. This would mean that this victim cannot be included in further analyses (to investigate whether child protection measures are disproportionate among those affected by the benefits affair). The examples above always assume the situation where you select one person per victim for the narrow comparison group. However, it is also possible to select several similar people per victim for the narrow comparison group. For example, that you select multiple Surinamese non-victims per Surinamese victim (in the case of exact matching) or that you select multiple people who are the same propensity score have as the victim (in the case of nearest neighbour matching) or that you select multiple people who are in the interval that you set as the maximum distance between the propensity scores (in the case of caliper matching, this will then be radius matching mentioned). Finally, it is also possible to combine the above techniques (for example, exact matching on a number of variables and other variables). propensity scores use to match). In short, there are several methods for compiling a narrow comparison group from the broad comparison group by matching.32) Because all methods have advantages and disadvantages, several (combinations of) methods were explored in this study. The table below shows the methods that were explored in this study. Because of the importance of origin for duplication (both statistically and theoretically), exact matching is used for this characteristic. This also applies to the number of children in the household. This attribute is not included in the model above to determine the propensity scores to be calculated because it is too closely related to other characteristics (there is multicollinearity with, for example, the type of household). Given the importance of this attribute, this is included in this way (this is therefore the eleventh characteristic that is included in the matching in addition to the 10 characteristics as shown above). Finally, the analyses are carried out separately per cohort year. This is because the relationship between characteristics and duress can change over time, because the tax authorities/allowances recovery policy has changed over time (see, for example, chapter 2).Table 5.2.2. Overview of matching methods UsedModelSelect 1 or more people per affected Matching method 1 1 Exact by origin + number of children + cohort year nearest neighbor2 1 Exact by origin + number of children + cohort year caliper3 Multiple Exact by origin + number of children + nearest neighbour cohort year 4 Multiple Exact by origin + number of children + cohort year radius5.3. Matching resultsThe first model (one-on-one matching with nearest neighbour) was ultimately used in this study to compile the narrow comparison group. The reason for this is that this method made it possible to find a suitable match for all victims and that for all characteristics included in the model, there were no more significant differences between the victims and the non-victims in the narrow comparison group. This also applies to all cohort years separately. As an illustration, the differences between the groups are shown below for origin (most decisive for duping), type of household (determining child protection measures) and household income (important for both) .33) As shown in the figures below, there are clear differences in these characteristics between the victims and the broad comparison group (also see previous chapter), but there are no significant differences in these characteristics between the victims and the broad comparison group (see also previous chapter) the group of victims and the narrow comparison group. This applies to all characteristics that are included in the matching process. In other words, by means of matching, a narrow comparison group was selected from the broad comparison group that is similar to the victims on important characteristics.31) As shown in the previous chapter, based on the available data, it is possible to reasonably predict the risk of being duped and being imposed on child protection measures (MacFadden R2 > 0.2) .32) For more information about this matching method (and other matching methods not used in this study) see Caliendo & Kopeinig, 2008. This study uses logistic regression as estimation method. This is a commonly used method and is often also sufficient for dichotomous variables. Furthermore, selection is carried out without delay. This is because the broad comparison group is large enough to find suitable matches without back.33) The standardized mean difference was studied for all characteristics. Other checks were also carried out to assess the quality of the matching (s group). For example, the common support figures were studied for the various methods.6. Relationship duping and child protection measuresThis chapter describes whether the victims of the benefits affair have been disproportionately faced with child protection measures. To do this, it was identified how many victims had to face a child protection measure after duping and how many before duping. These results are compared with the narrow comparison group. This group is similar to the group of victims in terms of relevant characteristics (see previous chapter). For information, the broad comparison group also shows how much this group has experienced child protection measures. Every year, around 1.2 percent of children face a child protection meas.34) In this study, the analyses are not done at the child level, but at household level. As a result, the percentages in this study are somewhat lower for the broad comparison group (several children can receive a child protection measure in one household). This analysis looked at the minor children who belonged to the applicant's household in the year prior to dupering/selection. These do not have to be legal children, but can also be children of a (new) partner, for example. However, since they live at the same address, it can be expected that these children fall under this applicant's daily life (and that this applicant's situation will therefore have affected these children in terms of duplication and/or related problems). Figure 6.1.1 shows the progress of child protection measures in the three different groups over time.35) It shows what percent of the households in the different groups per year had to deal with a child protection measure in the household. Year t is here the year of dupering/selection (1 January to 31 December). A total of 5 years will be considered (the year before dupering/selection (t-1) to 3 years after dupering/selection (t+3)). As this graph shows, victims have more child protection measures in the family than non-victims in the broad comparison group (around 3 percent and less than 1 percent, respectively). However, this does not seem to be due to duping, but because of the characteristics of these victims. Indeed, even before duping (t-1), victims have more child protection measures in the family than people in the broad comparison group. There is also no significant increase for the victims over time (the percentage of child protection measures in the family varies between 2.5 and 3 percent over time). When compared with the narrow comparison group (which has the same characteristics as the victim group), victims do not have to deal with child protection measures more often both before and after duping; the percentage is around 3 percent for both groups. In the figure, there appear to be small differences between the victims and the narrow comparison group, but these are not statistically significant. For all individual years (t-1 to t+3), the victims do not have child protection measures in the family more (or less frequently) than the narrow comparison group.36) The fact that there were no significant differences between these groups even at t-1 shows, like the results from the previous chapter, that an appropriate comparison group was put together by matching.Source: CBS, t-1tt+1t+2t+3012346.1.1. Child protection measures over time by groupSmall comparison groupDuplicates Broad comparison group%6.1.1. Child protection measures over time by GroupCategory”, “Narrow comparison group”, “Victims”, “Broad comparison group t-1", “2.68", “2.56", “0.63 t”, “2.83", “2.73", “2.98", “0.71 t+2, “2.68", “3.00", “2.73", “2.98", “0.71 t+2”, “2.68", “3.00", 75 t+3", “2.88", “2.73", “0.77Narrow comparison group (%) Victims (%) Broad comparison group (%) t-1 2.68 2.56 0.63t 2.83 2.73 0.77t+1 2.73 2.98 0.71t+2 2.68 3.00 0.75t+3 2.88 2.73 0.77Source: CBS, The graph above shows the percentage of households per group that experienced with child protection measures in a shown for a certain year. The following graph shows how many percent of the households in the three groups experienced household child protection measures at any time in the 3 years after dupering/selection. This therefore concerns the total number of households with child protection measures in the family sometime in the 3 years after dupering/selection.Source: CBS,012346.1.2. Child protection measures after duping/selection (t+1 to t+3) by groupNarrow comparison groupsSubjects Broad comparison group%6.1.2. Child protection measures after duping/selection (t+1 to t+3) by groupCategory”, “Narrow comparison group”, “Victims”, “Wide comparison group “, “3.95", “3.98", “1.07Narrow comparison group (%) Victims (%) Broad comparison group (%) 3.95 3.98 1.07Source: CBS, Here it is also clear that victims are more likely than the broad comparison group comes into contact with child protection measures (4 percent and 1 percent, respectively). Compared to the narrow comparison group, child protection measures among victims are not more common after dupering/selection (approximately 4 percent for both groups). Significance tests also show that child protection measures after dupering/selection are no more common among victims than in the narrow comparison group compared to before dupering/selection. On average, the fact that someone has been duped does not seem to have further increased the chance of a child protection measure compared to comparable non-victims. This does not rule out the possibility that there are certain (subgroups) victims who, as a result of the benefits affair and possible accumulation of problems, have more often had to deal with child protection measures. That is why exploratory analyses were carried out to see whether the effects of the benefits affair could be found for certain subgroups of victims. More specifically, we looked at having a low household income, single parents and whether or not they have a migration background. No significant interaction effects were found here that suggest that this effect is true for certain subgroups (see Appendix 4 for an overview of the most important tests performed) .34) http://opendata.cbs.nl/statline/#/CBS/nl/dataset/85099NED/table?dl=6ED2E.35)Waarbij we look at the children who were in the household for duping. Over the various years, it was identified whether or not these children were subject to a child protection measure. It was decided not to look at the children who currently live in the household every year, because differences in the number of children between groups after being duped in the household could then possibly influence the results. The number of children in the household for duping was included in the matching and is therefore comparable for the narrow comparison group and the victims, so that a pure comparison can be made.36) The lines in the graph show the average. If the confidence intervals are taken into account, it can be seen that each year the line/value of the victim group falls into the confidence interval of the narrow comparison group (so there are no significant differences).7. Conclusions• This is the first multivariate quantitative study to investigate the characteristics of those affected by the benefits affair. This research confirms the view that people who themselves and/or whose parents were not born in the Netherlands are more likely to be duped. Of people who were not affected by the benefits affair, 78 percent were born in the Netherlands themselves, as were both parents. Among those affected by the benefits affair, this figure is 29 percent. • The group of victims was also already vulnerable to duping. For example, victims are strongly overrepresented in the lowest income brackets: 44 percent of the victims have a household income that is among the lowest 20 percent in the Netherlands and among unaffected applicants for childcare allowance, this is 11 percent. • Victims also face child protection measures more often than non-victims. In the year before duping, around 3 percent of the victims had a child protection measure in their family, and among non-victims, this figure is less than 1 percent.• This study found no evidence that victims were more likely to have child protection measures imposed in the family as a result of duping. There is no significant increase in child protection measures after duping and victims do not have child protection measures more often after duping compared to a group of non-victims with the same characteristics. • On average, the duping has therefore not provided additional child protection measures for victims. This study also found no evidence that certain subgroups of victims (e.g. people with a low household income, single parents or people with a migrant background) did come into disproportionate contact with child protection measures after duping as a result of duping. This does not mean that it can be ruled out that there are individual victims who got into such trouble as a result of the benefits affair that child protection measures had to be used. • Currently, only information is available to CBS about whether a person is registered as a victim of the benefits affair or not. For example, there is no information available about the extent to which people have been duped. It is possible that people who had a very high “debt” with the tax authorities got into such trouble that, after duping, they more often had to deal with child protection measures. Only if additional information becomes available could CBS investigate this.• The CBS does not have the files that underlie the child protection measures. The CBS therefore does not know the reason for such measures. • This study only looked at the relationship between abuse and child protection measures. In follow-up research, attention could also be paid to other problems that may be the result of becoming a victim of the benefits affair. This includes financial problems that affected families may have experienced.Appendix 1 Composition of the Supervisory CommitteeProf. dr. Casper Albers, University of GroningenMr. dr. Alexander Hoogenboom, College for Human Rights/Maastricht UniversityProf. dr. Tobias Klein, Tilburg UniversityProf. dr. Frank Pijpers, Central Bureau of Statistics/University of AmsterdamDepartment of Methodology, Central Bureau of StatisticsAppendix 2 PrivacyAs with all CBS studies, the privacy of people studied is also central to this study. UHT provided a file containing the social security numbers (BNSs) of the victims via a secure environment. After the file was received, the anonymisation process started. This means that all identifying variables are removed from the file. An anonymous coupling key was then added (and the BSNs removed) based on the BSns. This is a personal number that all persons in the Basic Registration of Persons (BRP) have with the CBS and that is meaningless outside the CBS (it is impossible to trace this number to specific persons). Only a limited number of people within the CBS are involved in this anonymisation process. For the other people within the CBS, and therefore also for the researchers of this project, the BSNs were not visible in the files that worked with. All necessary files were linked together by means of the anonymous link key (for example the UHT file with the BRP). Here, too, only a limited number of people within the CBS have access to all files. Researchers must always submit an analysis plan that clarifies which variables are requested for which research and for what reason. This involves a check for purpose binding: is the file in question really necessary to answer the research question? For some data sources, explicit permission must also be requested from the external supplier of the data. Under the CBS Act, CBS has the task of conducting statistical research for practice, policy and science by the government and making the statistics compiled on the basis of such research public. The CBS therefore only uses the available data if it is for the purpose of an investigation carried out by CBS at the request of government organizations, in this case the Justice and Security Inspectorate. (37) In addition to the legal consideration whether CBS may carry out a request for additional statistical research, the ethical aspects of a statistical request are also considered. That's why CBS has an ethics committee consisting of a number of content experts. The ethics committee considers whether the social benefits of a study outweigh the burdens (including the emotional burden for those involved). This ethics committee was also asked for advice for this study. The ethics committee then gave positive advice because of the social importance of the research in combination with the available data and capabilities. In addition, all CBS studies ensure that the research results can never be traced back to individuals. If a certain breakdown results in numbers that are too low (and may therefore be traceable to specific individuals), the results will not be published. The numbers in the tables are also rounded (in this case to at least five). Furthermore, CBS never provides recognisable personal data to third parties. The Justice and Security Inspectorate therefore has no access to the files used for this investigation. More information about how CBS takes privacy into account in all investigations can be found on the CBS website.37) Only exceptionally, CBS conducts research for private parties: https://www.cbs.nl/nl-nl/onze-diensten/maatwerk-en-microdata/aanvullend-statistisch-onderzoek.Annex 3. Definitions of background characteristics researchedDemographic characteristics• Gender of applicantIt is being investigated whether the gender of the childcare allowance applicant (male or female) is related to the risk of being duped and to the risk of being imposed a child protection measure on one or more children in the household.• Applicant's pageThe applicant's age is also included in the analyses. The following categories are used for this purpose: Under 30 years old Between 30 and 40 years old Between 40 and 50 years 50 years and older This looked at the age in the year for dupering/selection.• Age at birth first childIt also looks at the age at which the applicant for the childcare allowance first became the parent of a legal child. For this purpose, this study distinguishes the following categories: Under 25 years old (Between 25 and 35 years) Between 35 and 55 years (Applicant has no legal child Unknown). The age groups are based on the average age at which people in the Netherlands have a child for the first time. (38)• Applicant's marital statusThis concerns marital status as derived from commitment data in the BRP. The following categories are distinguished:Married or partnered. Legal commitment to the coexistence of two persons or marital status that occurs after entering into a registered partnership.39) Unmarried. Marital status that indicates that a person has never entered into a marriage or entered into a registered partner.Other or unknown. This includes widowed (marital status resulting from the dissolution of a legal marriage or registered partnership due to the death of the partner) and divorced (marital status resulting from the dissolution of a marriage due to divorce or the dissolution of a registered partnership other than the death of the partner). Excluding legal persons as they remain formally married/registered partners.) This concerns the applicant's marital status on December 31 of the year before dupering/selection.• Origin and country of origin (applicant 40)The new CBS format is used to identify origin. 41) This first looks at whether someone was born in the Netherlands or abroad and then where the parents were born. These results in the following categories:- Born in the Netherlandsboth parents born in the Netherlands, one parent born abroad, two parents born abroad- Not born in the Netherlandsboth parents born in the Netherlands, one parent born abroad, two parents born abroad, then the country of origin will be determined. For people born abroad, the country of origin is their own country of birth. For people born in the Netherlands, the country of origin is determined by the country of birth of the parents. When both parents were born abroad, the mother's country of birth is leading in determining the origin. This is because the mother's birth details are known more often than those of the father. If the mother was born in the Netherlands or the mother's country of birth is unknown, the father's country of birth is used. For country of origin, in line with the new CBS layout, the following categories are distinguished: The Netherlands, Europe (excluding the Netherlands), Turkey, Morocco, Suriname, Dutch Caribbean and Other Outer Europe.• Residential province of the householdIn order to identify the regional spread (42) of abuse and the chance of being imposed a child protection measure, we will look at the province in which the applicants (according to the BRP) lived on December 31 of the year before dupering/selection: Drenthe-Flevoland-FrieslandGelderland-Groningen/Limburg-Noord-Brabant/Noord-Holland/Overijssel/UtrechtZeeland-South Holland.• Urban level of household residenceAs a final demographic characteristic, we look at the urbanity of the municipality where people lived in the year before dupering/selection. The division of municipalities by urbanity is based on the municipality's environmental address density.43) First of all, the address density of an area with a radius of 1 km around that address was determined for each address within a municipality. The environmental address density of a municipality is the average value for all addresses within that municipality. The following classes are distinguished:Very urban (environmental address density of 2,500 or more); Highly urban (environmental address density of 1,500 to 2,500); Moderate urban (environmental address density of 1,000 to 1,500); Low urban (environmental address density of 500 to 1,000); Non-urban (environmental address density of less than 500) .UnknownHousehold situation • Household typeThis is a description of a household based on the relationships between the people within a household. This is based on the basic registration of persons (BRP). The household of applicants was examined on December 31 in t-1. This study distinguishes the following categories: Couple with children: Two people who have one or more children living at home. These can be married couples (two people who are married to each other or who have entered into a registered partnership) and unmarried couples (are in a cohabitation relationship but are not married to each other or have no partnership registration). Another member (person who is part of a private household other than as a partner or child living at home) can also belong to this type of household, for example a resident grandma. Single-parent household: Private household consisting of one parent with one or more children living at home. Another member (person who is part of a private household other than as a parent in a single-parent household or as a child living at home) can also belong to this type of household.Other household: Private household that consists exclusively of other members (persons who are part of a private household other than as a partner, parent in a single-parent household or as a child living at home). This includes, for example, two brothers who together form one household. People for whom the household is unknown also fall into this category. Single-person households (private household consisting of one person), couples without children (two people who have a cohabitation relationship and have no children living at home at the time of reference) are also included in this category for the purposes of this study.44)• Number of children in the householdFor this characteristic, we looked at the address where people live according to the BRP. To deduce the number of children in the household, we look at all minor children in the household where the applicant lives (children under 18). These are not only the applicant's legal children, but can also be children of the (new) partner, for example. In this study, a distinction is made between 0, 1, 2, 3 and 4 or more children in the household. This also concerns the household situation on December 31 of the year before dupering/selection. As a result, the number of children at that specific time can be 0 (even though the persons have previously received childcare allowance and therefore had children in the household or were co-parents with the other parent registered in the BRP).• Age of children in the householdThe age of the above children (people under 18 living in the same household as the applicant in the year before dupering/selection) was also considered. In line with CBS publications on youth protection, the following age groups are distinguished:0 to 4 years4 to 8 years8 to 12 years12 to 18 yearsUnknownThis variable is created for both the age of the youngest child and the age of the oldest child in the household. The moment of reference for this is December 31 of the year for dupering/selection. This concerns the age of children who lived in the applicant's household, regardless of whether or not childcare allowance was applied for.• Application for a non-legal childThe study also looked at whether the child for whom childcare allowance was requested (in the period t-5 to t) is the applicant's legal child. For a woman, there is a legal child if the child is born to the woman or if the woman has officially adopted the child. For men, a legal child exists if the man is married to the mother at birth, or if he has recognized the child or unborn child, or if paternity has been determined by a judge or when he has adopted the child.45) This variable is assigned a value of 0 to an applicant if all children for whom he/she has applied for childcare allowance are his/her legal children. If at least one child is not the applicant's legal child, the applicant has a 1 on this variable. This can occur, for example, in so-called composite families.• Number of years of applying for childcare allowanceAs indicated in chapter 3, we looked at people who received childcare allowance in the period t-5 to t (i.e., 2012 looked at applicants for childcare allowance from 2007 to 2012). We then looked at how long people applied for childcare allowance during this period. The minimum value is once (because everyone selected for this analysis has applied for childcare allowance at least once during this period) and the maximum score is six times (a person received childcare allowance every year in the period t-5 through t).• Household relocationsThe study also looked at whether applicants have moved. To do this, we look at the applicant's address (again according to the BRP) on December 31 of the year before dupering/selection and whether this address was the same in the three years (including December 31) before (yes or no).• Change of partnerIn the household situation domain, the last thing we looked at was whether there have been changes of partners (if they were present in the year before duping). To do this, an applicant's partner is considered. This is explicitly about partners living together. So partners who live at the same address. This can include either married or unmarried partners. The applicant's partner on December 31 of the year before dupering/selection will be examined and then it will be examined whether or not the applicant had the same cohabiting partner in the 3 years before (including December 31). If people did not live with a partner in the year before duping, these persons are classified in the “no change” category.Education and socio-economic situation• Applicant's highest educational levelTo do this, the applicant's highest educational level is considered. This looked at the level of education in the year before dupering/selection. The following categories are distinguished:LOW: This includes education at the level of primary education, VMBO, the first 3 years of HAVO/VWO and the entrance course, the former assistant course (mbo1) .Secondary: This includes the upper level of HAVO/VWO, basic vocational training (mbo2), vocational training (mbo3) and middle management and specialist courses (mbo4) .High: This includes education at the higher professional or university level.Unknown: The educational level is not available to everyone in the entire population; as people are older, the coverage rate of delete the source files.• Children are the household who drop out of schoolThis looked at whether someone has left (funded) education and does not have a starting qualification. Having a starting qualification means that someone has at least completed HAVO or VWO education, basic vocational training (MBO level 2) or an old course of comparable level. Children in the household are being looked at for dupering/selection and these children are identified whether at least one child in the year before dupering/selection had dropped out of school (yes or no).• Property for sale or rent (yes/no housing allowance)The investigation also looked at whether the applicant lives in their own home or rental property on December 31 of the year before dupering/selection and if he/she lives in a rental property whether or not they receive housing allowance. More specifically, the following categories are distinguished:Private HomeRented property with rent benefitRental property without rent benefitInstitutional or unknown household• Household incomeAnother characteristic of the socio-economic situation that was looked at in this study is disposable household income in the year before dupering/selection. A household's disposable income consists of gross income less paid income transfers such as alimony from the former spouse, income insurance contributions such as social security contributions, national insurance and private insurance related to unemployment, disability and old age and survivors, health insurance contributions, and taxes on income and wealth. To make a meaningful comparison between different types of households, we look at the standardized household income. This means that the incomes of households of different sizes and compositions are made comparable. For this purpose, equivalence factors are used that take into account the number of adults and children (by age) in a household. This study works with quintiles:20 percent households with the lowest incomesSecond 20 percentThird 20 percentFourth 20 percent highest-income householdsUnknown• Household assetsA household's wealth is the balance of assets and debts in the year before dupering/selection. For household assets, the total value of bank and savings and securities, bonds and stocks, home ownership, business assets and other assets of a household is considered. The debts include mortgage debt for your own home, student debts and other debts such as for consumer purposes, financing stocks, bonds or rights to periodic benefits, debts for financing the second home or other real estate. This research distinguishes the following categories: Less than 0 euro (more debts than assets) 0 to 50 thousand euro50 thousand euro or moreUnknown• Main household income sourceThis looks at the source from which a household received the most income in the year before dupering/selection. The following categories are distinguished here: Wage as an employee Income from your own companySocial assistance and/or other social benefits 46) Unemployment benefit/disability benefit47) Unknown income or other income (property income, student loans or pension benefits).• Household health insurance defaulterOn September 1, 2009, the Health Insurance Defaulters Act came into force. Under this law, defaulters are detected. To do this, health insurers report to the CAK (until 2016, this was to the Zorginstituut Nederland) of their insured persons with a contribution delay of at least six months. The CAK then collects an administrative contribution (i.e. an increased premium) by deducting the salary or benefit (withholding tax). If withholding tax is not (entirely) possible, the premium is collected with the help of the Central Judicial Collection Office (CJIB). Defaulters of the Health Insurance Act are persons who have not paid a premium for their basic insurance for at least 6 months, are registered in the Basic Registration of Persons (BRP) on the reference date, are registered with the CAK, are in the administrative contribution regime, and are 18 years or older. This study looked at whether the applicant himself and/or his/her household members were registered as defaulters of the health insurance Act in the year before dupering/selection (yes or no).• WSNP processThis looks at whether or not people had completed a process under the Natural Person Debt Restructuring Act (WSNP). These are debt restructuring processes pronounced by the court that offer people in a problematic debt situation the opportunity to become debt-free. During the debt restructuring process, the debtor pays off as much of the debts as possible under the supervision of an administrator and in accordance with a strict regime. This enforces the cooperation of creditors. There are several conditions. For example, no new debts may be incurred. If the debtor has complied with the agreements, the court can grant a so-called blank slate after three years, after which residual debts are no longer due and payable. It is examined whether or not someone in the applicant's household (including the applicant himself) had completed such a process in the year before dupering/selection.Healthcare use• Receiving mental health careThis looked at whether people have received one or more forms of curative (medical) mental health care (mental health care) under the Health Insurance Act (Zvw) (care covered by the basic insurance) .48) If the insured person receives a prescription for, for example, a medicine, this care falls outside mental health care but under pharmaceutical care. This study looked at the applicant and any household members who were in the applicant's household before being duped in the year before being duped. These persons (applicant plus any other household members) have been identified whether or not they have received mental health care in the 3 years for dupering/selection.• Use psychotropic drugs with household membersIn order to gain more insight into people's mental well-being (and the presence of any problems in this area in the household), we are looking at the use of psychotropic drugs. Psychotropic drugs are drugs that are used to treat psychiatric disorders and psychological problems. More specifically, it looks at whether people are taking (or at least have been prescribed and reimbursed) the following medications: Antipsychotics, anxiolytics, hypnotics and sedatives, antidepressants and psychostimulants (ADHD and nootropics). It will be examined whether these drugs were provided in the 3 years before dupering/selection to household members who were in the applicant's household and/or the applicant himself (yes or no) in the year before duping.• Use medication for addictionsThe last characteristic of drug use is whether or not people use drugs that are associated with addictions (drugs for alcohol addiction and opioid addiction). This also looks at whether people who lived in the same household as the applicant and/or the applicant themselves in the year before dupering/selection received these medicines in the 3 years before dupering/selection (yes or no).• WMO household useThe Social Support Act (Wmo) makes municipalities responsible for supporting the self-reliance and participation of people with disabilities, chronic mental or psychosocial problems. It includes support provided under the Wmo 2015 in the form of a product or service that is tailored to an individual's wishes, personal characteristics, opportunities and needs. WMO customized services can be divided into the following main groups:Home support (guidance, personal care, short-term stay, other support focused on the individual or household/family, day care, other group-oriented support and other customized arrangements) .Household assistance.Accommodation and care (protected living, childcare, emergency care and other protected living and care) .Tools and services (housing services, transport services, wheelchairs, transport facilities, housing, etc.) facilities and other aids) .For this study looked at whether at least one household member who was in the applicant's household in the year before being duped (including the applicant himself) made use of a WMO facility for duping in the year. Data about Wmo has been available since 2015. This study looks at characteristics before duplication. This means that Wmo can only be included for the cohort years 2016 to 2018.• Mild intellectual disability (LVB) registration applicant and possible partnerThe study also looked at whether the applicant and/or any partner has a mild mental disability. According to the operationalization used, there is a mild intellectual disability when a person meets at least one of the following requirements: receives disability benefits with a diagnosis of a mild mental disability or has a WSW indication.49) This characteristic also looks at the year for dupering/selection.• Registration as a household suspectFinally, it was examined whether household members who were in the applicant's household before dupering/selection in the applicant's household (including the applicant himself) were registered for dupering/selection as suspects of committing a crime (yes or no) at some point in the 3 years. This concerns people who are registered in the Basic Police Enforcement Registration System. A person is registered by the police as a suspect if there is a reasonable suspicion of a crime. (38) For the average age at which people in the Netherlands have a first child, see: https://www.cbs.nl/nl-nl/visualisaties/dashboard-bevolking/levensloop/kinderen-krijgen (cbs.nl). In exceptional cases, the age at which someone had a child is unlikely, according to the administrative sources used (for example, under 10 or over 55). These persons are classified as unknown. (39) A marriage-like relationship between two persons laid down in a Civil Status Deed.40) The CBS uses an assessment framework to determine whether a migration background can be included in investigations. This is only possible if there are substantive and methodological considerations for doing so. Due to the fact that nationality played a role in the benefits affair, the internal expert group allowed the use of migration background for this study.41) https://www.cbs.nl/nl-nl/nieuws/2022/07/cbs-introduceert-nieuwe-indeling-bevolking-naar-herkomst.42) The Netherlands is divided into 42 youth regions to promote supra-local cooperation in the field of youth assistance. The amount of available data does not allow a breakdown to such a large number of regions, which is why a breakdown into the 12 provinces was chosen. 43) The municipality in question was derived on December 31 at t-1. Data from 2021 was looked at for both the municipal layout and the degree of urbanity.44) This study concerns applicants for childcare allowance. This study looked at the situation of applicants at one time (in order to make a good comparison). Namely, the moment as close as possible to the year of dupering/selection (in this case, December 31 at t-1). It is therefore possible that people had no children in the household at this specific time (for example because the children have moved or because relationships have ended) .45) Partly, this also applies to parents of the same sex.46) As far as assistance is concerned, this is a benefit under the General Assistance Act (ABW) or the Work and Assistance Act (WWB). Other social benefits include: benefits Act on income provision for older and partially incapacitated unemployed workers (IOAW) benefits Act on income provision for older and partially incapacitated former self-employed workers (IOAZ) allowance Act on income provision for older unemployed Decree benefits on self-employed assistance (Bbz) benefits Act on Work and Income Artists (WWIK) benefits Act (WWIK) benefits Act (WWIK) benefits Act on Disability Benefits for Young Disabled Persons (Wajong) Benefits Act War and Resistance Pensions Act. (47) Here include several benefits such as: the Disability Insurance Act (WAO), the Disability Provision Act for Young Disabled Persons (Wajong), the Self-Employed Disability Insurance Act (WAZ), the Work and Income by Capacity Act (WIA) 48) Due to a different layout of mental health care, different dates/measurements were used for different years. Until 2013, the operationalization of mental health care (GGZ) looked at the costs incurred and reimbursed for a person, under the Health Insurance Act (Zvw), for secondary mental health care (GGZ). The costs for secondary mental health care include: costs of diagnostic treatment combinations (dBCs), mental health care with stay (with or without treatment), costs of DBC's mental health care without a stay (with institutions or independent residents), costs of personal budgets (PGB), mental health care and costs of other mental health care. Since 2014, we have looked at the costs for a generalist basis and specialized mental health care that are made and reimbursed for a person, under the Health Insurance Act (Zvw), which looks at both the treatment of patients with mild to moderate, non-complex mental problems or people with stable chronic problems and at treatment instead of patients with serious or complex mental problems. As of 1 January 2015, this item also includes costs incurred in long-term mental health care. (49) An indication under the Social Employment Act. Someone with a physical, mental or mental disability cannot simply apply for a job in the social work facility. This person must first register for research. The UWV Werkbedrijf is carrying out the investigation. If the result is positive, the person will receive a WSW indication. The standard CBS definition for mild mental disability also looks at whether a person has a Wlz Indication (Long-Term Care Act) in the care profile with a mild mental disability. However, the Wlz has only been in effect since 2015. In order to keep the measurement of this characteristic the same across the different cohorts, this part was not included in this study.Appendix 4: Difference in difference analysesThis appendix contains the results of difference into difference tests performed for this study.50) To do this, the group of victims (n = 4 100) is compared to the narrow comparison group (n = 4 100). Family child protection measures for dupering/selection (t-1) are considered compared to child protection measures after dupering/selection (t+1 to t+3) .51)50) If the same people are compared at multiple times, you can choose to work with clustered standard errors. Otherwise, there may be (too) small standard errors (due to clustering). In these analyses, there are only 2 clusters and this was not chosen. The most important coefficients that were the reason for these models were already not significant either. The substantive conclusions drawn from these analyses with regard to the interactions will therefore not change if they were opted.51) Child protection measures in year t (the dupering/selection year) are not considered. There is an exact date of duplication and a date of commencement of child protection measures. However, the date surrounding duplication is when a first correction was found in the administration of the tax authorities. However, this is not necessarily the time when the duplication actually starts for the victims. This is probably only when the letter about this has been received. Or even later when the “debts” are actually recovered. In order to have a more pure before and after dumping, the year of duping is therefore not included.Appendix 5. ReferencesCaliendo, M., & Kopeinig, S. (2008). Some practical guidance for the implementation of propensity score matching. Journal of economic surveys, 22 (1), 31-72. Hosmer Jr, D. W., Lemeshow, S., & Sturdivant, R. X. (2013). Applied logistic regression (Vol. 398). John Wiley & Sons.McFadden, D. (1974). Conditional logit analysis of qualitative choice behavior. Pp. 105—142 in P. Zarembka (ed.), Frontiers in Econometrics. Academic Press. Rosenbaum, P. R., & Rubin, D. B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrics, 70 (1), 41-55.Simonen, J. & McCann, P. (2008) Firm innovation: The influence of R&D cooperation and the geography of human capital inputs. Journal of Urban Economics, 64 (1), 146-154. Read more aboutyouth protection youthfulness childcare allowance

Date

20 April 2024

Author (s)

research

Source

No items found.

Readers' comments

No items found.