Benchmarking Analysis of Danish Municipalities' Integration Policies

by Leif Husted and Eskil Heinesen, March 2004 AKF Forlaget

Summary

Introduction

From 1999 onwards Danish municipalities have had the full responsibility for active policies aimed at integrating new immigrants into the labour market. The group of immigrants for which the municipalities have these obligations are persons who

got residence permit after January 1st 1999,
were 18-61 years of age when they got their residence permit,
are from non-EU and non-Nordic countries, and
are refugees or family reunified.

The general framework for the integration policy is given by national law, but the 275 municipalities have great discretion regarding the administration of the law and the specific implementation of the policies.

The purpose of this report is to specify and estimate an indicator which may be used for comparing or benchmarking Danish municipalities’ success regarding integration of immigrants. We use micro data based on administrative registers at Statistics Denmark for the period 1999-2001. The sample consists of all immigrants fulfilling the four criteria listed above. The benchmark indicator which we estimate is based on the duration from the date an immigrant gets residence permit to the date he enters the labour market, i.e. the date he starts on his first regular job (or education). A short average duration to entering the labour market in a given municipality may indicate a successful municipal integration policy. Contrary, a long average duration may indicate a less successful municipal policy.

Correction for variation in municipal characteristics

However, a long average duration from the date of residence permit to the date of entering the labour market may not necessarily be due to inefficient municipal integration policies. Alternatively, it may be caused by unfavourable general conditions for the municipality, e.g. a high local unemployment rate, or unfavourable characteristics, in terms of labour-market integration, of the immigrants living in the municipality. Therefore, we correct the average duration until entering the labour market for municipal differences in general conditions and (observed) immigrant characteristics.

The corrections are based on estimating a statistical duration model for all immigrants in Denmark fulfilling the four criteria listed above. The individual characteristics of the immigrants which enter the model as explanatory variables (and which are used for calculating corrections) are the following:

gender
basis for residence permit (refugee, or family reunification to a refugee or to a non-refugee)
country of origin
age
single/married (to an immigrant or to a non-immigrant)
having children
year of residence permit
health (number of contacts with general practitioners).

We take account of variation in these individual characteristics at a very detailed level. Thus, we estimate the duration model separately for males and females, we include more than 40 dummy variables for country of origin, and we include interaction terms between age categories and variables related to having children on the one hand and variables for basis of residence permit on the other.

The estimated effects of these individual characteristics on the hazard rate to employment (or education) are in general very significant. We find, e.g., that the hazard rate is significantly smaller for older immigrants than for younger, it is smaller for persons (especially females) with small children, it is smaller for refugees than for persons who are family reunified to non-refugees, and it is very low for immigrants from countries such as Lebanon, Somalia, Afghanistan, Syria, Iran and Iraq compared to immigrants from European countries. Of course, the very large estimated effects of the dummy variables for country of origin may in large part be due to the fact that we have no data on the level of education or labour-market experience from the home country.

The general municipal characteristics included in the duration model as explanatory variables are the following:

the local unemployment rate
the number of immigrants from non-western countries as a share of the population
the number of immigrants fulfilling the four criteria listed above as a share of the population.

The two last variables may reflect the municipality’s experience regarding integration of immigrants. They may also be related to the magnitude of local ethnic networks and firms. According to our estimates, an increase in both variables is associated with a higher hazard rate to employment.

As expected, an increase in the local unemployment rate reduces the hazard rate to employment. We have tried to include several other variables at the municipal level, e.g. related to local labour-market conditions and industrial structure, but they were not significant in the estimated model.

Calculating the benchmarking indicator

The benchmarking indicator for a given municipality is equal to the »observed« mean duration until entering the labour market (based on the non-parametric Kaplan-Meier estimate of the survival function for the municipality) minus the mean expected duration given the estimated parametric duration model and the observed characteristics of the municipality and its immigrants. Due to the short observation period both mean durations are calculated as restricted means at 18 months. The mean expected duration is calculated as the mean of the expected durations for each individual immigrant in the municipality, given the estimated model.

The expected duration is long if, e.g., the local unemployment rate is high or if a large share of the immigrants is refugees, elderly, females with small children, or from countries such as Lebanon, Somalia, Afghanistan, Syria, Iran and Iraq. It is short if, e.g., the local unemployment rate is low or if a large share of the immigrants is young, family reunified to non-refugees and from European countries.

A negative value of the benchmarking indicator for a given municipality implies that the »observed« mean duration until employment is shorter than expected, given the model, i.e. that the municipal integration policy seems to be successful. Contrary, a positive value of the indicator implies a longer observed duration than expected and a less successful integration policy.

Discussion

Compared to other indicators which have been used for benchmarking Danish municipalities, the advantage of the benchmarking indicator calculated in this report is that it corrects the observed duration for important differences in terms of both general conditions and characteristics of the individual immigrants. The correction is important since it implies that the ranking of many municipalities changes a lot (compared to the ranking implied by the observed mean durations without correction). The correction also implies a smaller variation in mean duration between municipalities, although large differences remain.

However, it is important to stress that the benchmarking indicator is not a very precise measure of the degree of success of municipal integration policies. Thus, there are important variables which we are not able to take account of when estimating the indicator: We have only very limited information on health status based on the number of contacts with general practitioners, and we have no information on education or labour-market experience from the country of origin – nor have we information on abilities or motivation. There may also be important municipal characteristics which we have not been able to take sufficiently account of, e.g. regarding local labour-market conditions.

Another reason why the benchmarking indicator is not a very precise measure of the success of municipal integration policies is that it is calculated for a rather short period of time, namely 1999-2001. Especially the immigrants who got residence permit in 2000 and 2001 have only had a very short time to find a job within the sample period. Important parts of the integration measures applied by the municipalities – especially measures targeted towards immigrants with weak qualifications – will have positive employment effects only after several years.

The short data period implies that most municipalities have only rather few immigrants meeting the four criteria listed at the beginning of this summary. For these municipalities the benchmarking indicator is rather imprecise. Therefore, in this report we show the benchmarking indicator only for the 44 municipalities with at least 50 immigrants who got their residence permit between 1 January 1999 and 1 July 2000. Due to the statistical uncertainty in calculating the benchmarking indicator, a large fraction of these 44 municipalities does not deviate significantly from the average for all municipalities. Table 5.1 in this report shows the »observed« mean duration and the benchmarking indicator for the 44 municipalities as well as the ranking according to the indicator and its standard deviation.

If all municipalities are grouped into their respective counties, the analysis indicates that the counties in the metropolitan region of Copenhagen have a rather successful integration policy while several counties in Jutland and the counties of Southern Sealand and Funen are less successful. However, there is considerable variation in the municipal benchmarking indicator within counties.

The fact that the data period is 1999-2001 implies of course that the benchmarking indicator is not affected by the integration policies of the municipalities after 2001. It is important to be aware of this fact since a large number of municipalities have changed policies, organization and implementation measures regarding integration of immigrants within the last couple of years.

Some of the problems of the benchmarking analysis will be less severe in the coming years where the sample period can be extended to include more years. Thus, the statistical uncertainty of the benchmarking indicator will be smaller and the statistical model used to calculate the mean expected duration for each municipality may be extended to include more explanatory variables.

The ranking of municipalities according to the benchmarking indicator may be used as a starting point for other research projects, for instance more thorough qualitative investigations of a few municipalities, some with a high benchmarking score and others with a low score. Such analyses may be used to reveal possible important differences in municipal integration policies which may explain part of the variation in the benchmarking score, thereby creating a basis for an overall improvement of integration policies.