LABOR MARKET TRANSITIONS OF IMMIGRANTS WITH EMPHASIS ON MARGINALIZATION AND SELF-EMPLOYMENT* Kræn Blume, Centre for Research in Social Integration and Marginalization (CIM) and AKF, Nyropsgade 37, DK-1602 Copenhagen V. Email:
[email protected]. Mette Ejrnæs, CIM, CAM and University of Copenhagen, Studiestraede 6, DK-1455 Copenhagen K. Email:
[email protected]. Helena Skyt Nielsen, CIM, IZA and University of Aarhus, Building 322, DK-8000 Aarhus C. Email:
[email protected]. Allan Würtz, CAM and University of Aarhus, Building 322, DK-8000 Aarhus C. Email:
[email protected]. Date: January 2007 Abstract In developed countries, immigrants are more likely to be non-employed and self-employed compared to natives. Based on register data of male immigrants in Denmark, we perform a detailed investigation of the immigrant-native difference in transition patterns across labor market states. We find that a high proportion of immigrants from non-western countries tend to be marginalized relative to natives, and they tend to use self-employment to escape marginalization. Keywords: discrete competing risks, duration, mixed logit model, panel data, selfemployment, unobserved heterogeneity. JEL classification: C23, C25, C41, J23
*
We thank Lars Muus, George Neumann and Stephen Jones for helpful discussions, and we appreciate comments from two anonymous referees, seminar participants at McMaster University, York University, University of Illinois at Chicago, University of Copenhagen, University of Aarhus, Aarhus School of Business, Copenhagen Business School, IZA, the ESPE Conference and the CEPR/TSER Workshop. The usual disclaimer applies. This project was supported financially by the Danish National Research Foundation (the FREJA grant) and the Institute of Local Government Studies, AKF. In addition, Helena Skyt Nielsen was supported by the Social Science Research Council, and Mette Ejrnæs and Allan Würtz acknowledge support from the Center for Applied Microeconometrics (CAM). CAM’s activities are financed by a grant from the Danish National Research Foundation.
1
1. Introduction In many developed countries immigrants are much more likely than natives to be nonemployed. For example, the data used in this paper on 30-50 year-old male immigrants in Denmark shows that 43% were non-employed in 1997 compared to 18% of natives. In addition, immigrants from a number of less developed source countries were also more likely to be self-employed than natives. Among Pakistani and Turkish males, who are the two largest immigrant groups from less developed countries in Denmark, 22% and 16%, respectively were self-employed in 1997. The similar figure for natives was 9%. In this paper a framework is developed to investigate both facts simultaneously. Motivated by the huge differences in the probability of being non-employed and selfemployed, we investigate how the transition patterns between labor market states reveal the reasons for the differences. We propose a framework that takes all transitions between labor market states into account,1 and characterizes each immigrant as a transitional type. A transitional type could, for instance, be an immigrant who, compared to a native with the exact same characteristics, is more likely to enter non-employment from wage-employment and self-employment, and more likely to enter self-employment from non-employment. Such a transition pattern is consistent with a situation where the immigrant faces barriers against entering wage-employment and, as a consequence, chooses self-employment. Among the eight transitional types defined, we find a high proportion of this type of immigrant. No clear definition of marginalization exists. However, studies of marginalization often focus on the duration of unemployment (see the survey by Machin and Manning (1999)) and on the potential scarring effect of unemployment as discussed by Arulampalam (2001). If we regard marginalization as being closely related to the expected time spent in unemployment over a lifetime, it depends not only on the expected duration of an unemployment spell, but also on the likelihood of entering unemployment, which is closely related to the focus of our paper. We gain additional insight on marginalization via a focus on transition patterns rather than a focus on the duration of single spells. Since self-employment and wage-employment are significantly different economic activities, it is expected that the probability of exiting to unemployment differs among the two. Therefore, the three labor market states: wageemployment, self-employment and non-employment, are included in the analysis.
1
This is in line with Carrasco (1999), who demonstrates that it is important to investigate both transitions from unemployment and from wage-employment to gain insight on the reasons to become self-employed.
2
Self-employment is more important for a number of immigrant groups than it is for natives. There are many general explanations why people become self-employed,2 and many specific reasons why immigrants choose self-employment. For example, self-employment may be 3
used as a steppingstone to wage-employment, or it may be a tradition from the country of 5
origin4 and ethnic enclaves. In the empirical analysis, we control for all these reasons. In addition, there may be unobservable barriers for immigrants to enter wage-employment, and therefore they choose self-employment. We investigate the presence of such barriers using the framework of transitional types suggested in the paper. The transition probabilities are estimated by a discrete competing risks model. The empirical model allows for duration dependence and unobserved heterogeneity. Therefore, the model may be used to find the probability of being in a state. This is done by solving for probabilities, which resembles equilibrium probabilities in a discrete Markov chain model. They are denoted steady state probabilities.6 The data is register-based panel data sets for 1984-97 for 40% of all male immigrants in Denmark and a sample of 2% of the male native population. The data set has several advantages: It is a long panel of 14 years; it is known to be very reliable and precisely measured; and finally, the sample is by construction a random draw from the register data over the entire population. In total, 240,466 observations are used. The empirical analysis reveals very different transition patterns for immigrants of different origin. The model rightfully predicts that immigrants are much more likely to be selfemployed and non-employed than natives, and the gap persists in steady state. The analysis reveals that many immigrants are concluded to be a transitional type, which indicates that self-employment is being used as a last resort to avoid non-employment. This finding is 2
Many studies find a positive self-selection of supposedly innovative individuals into self-employment (Evans and Leighton, 1989; Evans and Jovanovic, 1989; Taylor, 1996; Fairlie and Meyer, 1996).Contrary to this finding, several papers find that push-factors such as previous unemployment experience, involuntary job loss or low labor income prompt self-employment (Carrasco, 1999; Moore and Mueller, 2002; Alba-Ramirez, 1994; Evans and Leighton, 1989). In some sectors there is a considerable need for capital to start a business, and therefore part of the literature is concerned with liquidity constraints (Lindh and Olsson 1996; Blanchflower and Oswald, 1998; Evans and Jovanovic, 1989; Taylor, 2001). To loosen the liquidity constraints, public selfemployment support has been introduced, though the success has been questioned (Pfeiffer and Reize, 2000). 3
See Light (1984).
4
See studies by Yuengert (1995), Hammarstedt (2001) and Hout and Rosen (1999).
5
Borjas (1986) and Borjas and Bronars (1989) argue that ethnic enclaves are a significant explanation of selfemployment among immigrants, because it is easier to attract customers and employees in an area with inhabitants of similar ethnic origin. The empirical evidence, however, is mixed. Contrary to the prediction of the theory, Clark and Drinkwater (2000) find a negative effect of enclaves on self-employment, whereas Aldrich and Waldinger (1990), Yuengert (1995) and Bager and Rezaei (2001) find no effect of ethnic enclaves. 6
A similar analysis is performed by Constant and Zimmerman (2005) although their focus is on the relationship between labor market dynamics and the business cycles.
3
perfectly in line with the conclusions drawn by Constant and Schultz-Nielsen (2004) in their comparative study of self-employment among Danish and German immigrants. In section 2, the transitional types are defined based on transition probabilities. The transition probabilities are estimated using a discrete competing risks model outlined in section 3. In section 4, the data for the empirical analysis is presented. In section 5, the results are discussed, and section 6 concludes the paper.
2. Transitional Types In this section, eight transitional types are defined in order to summarize differences in transition patterns between immigrants and natives. Each transitional type corresponds to frequently posed hypotheses on why immigrants tend to have that particular transition pattern compared to natives. Two of the transitional types are consistent with marginalization of immigrants in the labor market. 2.1. Definition of Transitional Types The transitional types are based on a comparison of the probability of leaving a state. The basic transition probability is the probability of entering a state conditional on the event history and characteristics; P ( St = j | St −1 = i, S t-2 , X t ) where St denotes the state at time t, St-2 is the event history until time (t-2), and Xt is a vector of characteristics. To investigate the cause of exit from a state, we condition on leaving the state: P ( St = j | St −1 = i, St ≠ i, S t-2 , X t ) This probability compares to relative cause-specific hazard rates in continuous duration models. The comparison of immigrants to natives assumes the same conditioning on characteristics: (1) PIM ( St = j | St −1 = i, St ≠ i, S t-2 , X t ) - PNA ( St = j | St −1 = i, St ≠ i, S t-2 , X t )
4
where subscripts IM and NA refer to an immigrant and a native, respectively. The sign of (1) determines whether an immigrant with the same characteristics and event history as a native is more likely to enter state j upon exit from state i. The expression in (1) can be extended to account for every state, and it forms the basis of the transitional types. When (1) is positive, state j is said to be a relatively dominant transition out of state i for the immigrant. Excluding ties, there is at least one relatively dominant transition out of a state. With three states in total, there are two states possible for exit and, thus, only one is a relatively dominant state. With a total of three states it is possible to define eight transitional types based on relatively dominant transitions, and they are listed in Table 1. For example, an immigrant of Transitional Type 1 has non-employment as the relatively dominant transition from both self-employment and wage-employment, and self-employment as the relatively dominant transition from non-employment. This means that, compared to a native with the same characteristics and event history, an immigrant of Transitional Type 1 is more likely to enter non-employment upon exit from self-employment and from wageemployment, and to enter self-employment upon exit from non-employment. Thus, the transitional types summarize a transition pattern between all the states.
Table 1 Transitional Types of Immigrants Transitional Type No
Relatively dominant transition Characterization Out of Out of Out of Attractor Escape SE WE NE state state
1
NE
NE
SE
NE
SE
2
NE
NE
WE
NE
WE
3
NE
SE
SE
SE
NE
4
WE
SE
SE
SE
WE
5
WE
NE
WE
WE
NE
6
WE
SE
WE
WE
SE
7
NE
SE
WE
-
-
8
WE
NE
SE
-
-
Description
Relatively marginalized
Relatively entrepreneurial
Relatively employed
Steppingstone
For six of the eight transitional types, one state is the relatively dominant transition out of all other states. For example, Transitional Type 1 has non-employment as a relatively dominant transition from the other two states. We denote such a state an attractor state because
5
immigrants are relatively more likely to make transitions into this state compared to natives. If an attractor state exists, we define an escape state as the relatively dominant transition state from the attractor state. In case of Transitional Type 1, the escape state is self-employment since this is the relatively dominant transition from the attractor state: non-employment. The characterization of states as attractor and escape states will be useful in highlighting differences between immigrants and natives. 2.2. Discussion of Transitional Types Transitional Types 1 and 2 can reveal new insight on marginalization of immigrants. The word marginalization is typically used in connection with being non-employed. In our framework, we denote immigrants with the attractor state, non-employment, as relatively marginalized. This happens when i) PIM ( St = NE | St −1 = SE , St ≠ SE , S t-2 , X t ) - PNA ( St = NE | St −1 = SE , St ≠ SE , S t-2 , X t ) >0 ii) PIM ( St = NE | St −1 = WE , St ≠ WE , S t-2 , X t ) - PNA ( St = NE | St −1 = WE , St ≠ WE , S t-2 , X t ) >0 Two transitional types of relatively marginalized immigrants exist depending on the escape state from non-employment. The definition of relatively marginalized differs from the usual definition of marginalized (which is often denoted discouraged workers or long-term unemployed) because the focus is on the transition pattern rather than on the duration of nonemployment. The role of self-employment to avoid marginalization is reflected in Transitional Type 1, which is a relatively marginalized immigrant who uses self-employment to escape nonemployment. A Transitional Type 1 satisfies besides conditions i) and ii) the following condition: iii) PIM ( St = SE | St −1 = NE , St ≠ NE , S t-2 , X t ) - PNA ( St = SE | St −1 = NE , St ≠ NE , S t-2 , X t ) >0 This transitional type is more likely than natives to fall back to non-employment when leaving one of the other states and to enter self-employment when leaving non-employment. If a relatively marginalized immigrant uses self-employment as escape state (Transitional Type 1), it may be because he faces barriers to entering wage employment. The immigrant is compared to a native with exactly the same observable individual characteristics except for 6
country of origin. Hence, a difference in transitional probabilities between immigrants and natives can arise from two sources: country of origin and different impact of the observed characteristics, that is, functional form of P( ). The two sources can again arise from barriers in the labor market, e.g. discrimination, lack of informal networks or knowledge of the local labor market. In particular, the variable country of origin may serve as a proxy for language proficiencies and cultural and religious differences. The transitional types provide new evidence on a number of hypotheses suggested in the literature concerning self-employment among immigrants. For example, the steppingstone hypothesis claims that immigrants use self-employment as a step towards getting a job. This is consistent with Transitional Type 8, who is more likely than natives to exit non-employment to self-employment, self-employment to wage-employment and finally back from wageemployment to non-employment. The mechanism is that an immigrant obtains skills, for instance language proficiency, while self-employed, which improve the chances of becoming wage-employed. Another example of a hypothesis suggested in the literature is a selfemployment tradition. Such a tradition would induce an immigrant to seek self-employment relatively more often than a native with otherwise similar characteristics due to comparative advantages unobserved from the viewpoint of the researcher. This is the case for Transitional Types 3 and 4, who are the relatively entrepreneurial types.
3. Discrete Competing Risks Model The discretely measured durations in the three states available in our data lead us to specify a discrete competing risks model. The specification includes a correction for unobserved individual heterogeneity. The event history of an individual is modeled conditional on characteristics X n and individual specific effects cn . The probability of an event history can be written as a product of the transition probabilities:
P ( SnT = snT ,.., Sn1 = sn1 | sn 0 , sn,−1 , sn,−2 , sn ,−3 , X n , cn ) = Tn
∏ P( S t =1
nt
= snl | sn ,t −1 , sn ,t − 2 ,.., sn ,−3 , X n , cn ) ,
where Tn is the number of observations on individual n.
7
We specify the transition probabilities as a first-order Markov process but add further duration dependence. A high-order Markov model with many characteristics implies that many parameters need to be estimated. In practice, this is infeasible. The probability of leaving a labor market state, however, may well depend on durations longer than the one year implied by the first-order Markov process. Therefore, we add variables on the duration in the current state. Let D(snt) be the duration, at time t, in the state snt. The specification of D(snt) is made flexible by using a vector of five dummy variables indicating 1, 2, 3, 4 or 5+ years in the current state. Furthermore, among the individual characteristics we included variables that are likely to capture the higher-order Markov effects e.g. total experience as a wage-earner, indicators for eligibility of unemployment benefit and self-employment support. The transition probabilities can then be written: P ( Snt = snl | sn ,t −1 , sn ,t − 2 ,.., sn,−3 , X n , cn ) = P( Snt = snl | sn,t −1 , D( sn,t −1 ), X nt , cn ) , where Xnt are the relevant characteristics for the transition at time t. The transition probabilities are assumed to follow a logistic distribution:
P ( Snt = j | S n ,t −1 = i, Dnt (i ), xint , cni ; β ij , γ ij ) =
exp( β ij ´xint + γ ij ´dint + cijn ) J
∑ exp(β
im
,
´xint + γ im ´dint + cimn )
m =1
where i,j are the states (1,2 or 3) and cni = (ci1n , ci 2 n , ci 3n ) ' are the individual specific effects relevant when the current state is i. Hence, the vector of individual specific effects is the 9dimensional vector cn = (c1n ', cn2 ', cn3 ') ' . We assume that the individual specific effects are random effects. Then it is necessary to eliminate the random effects by integrating them out of the probability of the event history. (1) P ( SnT = snT ,.., Sn1 = sn1 | sn 0 , sn , −1 , sn , −2 , sn ,−3 , X n ) = T
= ∏ ∫ P ( S nt = snt | sn ,t −1 , D( sn ,t −1 ), X nt , c) P (c | sn ,t −1 , D( sn ,t −1 ), X nt )dc t =1
8
The necessity to specify P (c | sn ,t −1 , D( sn,t −1 ), X n ) is known as an initial condition problem. We assume that the random effect is independent of the characteristics X n : P(c | sn ,t −1 , D( sn,t −1 ), X n ) = P(c | sn ,t −1 , D( sn,t −1 ))
The random effect, however, is not independent of the current state, sn,t-1 if the random effect influences the transition probabilities. For a similar reason, the random effect may not be independent of the duration, D(sn,t-1). We specify a finite support of cn . The cn cannot be identified and a normalization is * needed. A possible normalization is to set cijn = cijn − ciin . Then transitions out of, say, state 1, * * * H L c1* n = (0, c12 n , c13 n ) . We further assume that each cijn , i ≠ j , may equal two values µij and µij .
A B Hence, the support of c1* n consists of the four points supp {(0, µ12 , µ13 ) ', A, B ∈ {H , L}} . We
include a constant in the model. This implies that the points of support are not identified and a normalization is needed. We choose the normalization µ12H = µ13H = 0 . Therefore, we can identify the following support of c1n :
⎧⎛ 0 ⎞ ⎛ 0 ⎞ ⎛ 0 ⎞ ⎛ 0 ⎞ ⎫ ⎪⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎪ supp ⎨⎜ 0 ⎟ , ⎜ 0 ⎟ , ⎜ µ12L ⎟ , ⎜ µ12L ⎟ ⎬ . ⎪⎜ 0 ⎟ ⎜ µ L ⎟ ⎜ 0 ⎟ ⎜ µ L ⎟ ⎪ ⎩⎝ ⎠ ⎝ 13 ⎠ ⎝ ⎠ ⎝ 13 ⎠ ⎭
Similarly, the two parameters µ21L , µ23L and µ31L , µ32L can be identified in the support of cn2 and cn3 , respectively. With each support, we can identify four corresponding probabilities i i i i , π HL , π LH , π LL π HH . In principle there is no problem of adding further points in the support but
it turned out infeasible to estimate more points of support and their probabilities. It is necessary to specify the probability of the support conditional on the state and duration. We choose a multinomial logit model given by: ⎛ ⎞ ⎛ µik1 ⎞ ⎜ ⎟ ⎜ k⎟ i i* (2) π AB | D( snt −1 ) = Pr ⎜ cn = ⎜ µi 2 ⎟ | D( snt −1 ) ⎟ = ⎜ ⎟ ⎜ µk ⎟ ⎝ i3 ⎠ ⎝ ⎠
4
exp (δ ik + α ik ' D( snt −1 ))
∑ exp(δ im + α im ' D(snt −1 )) m =1
9
, k = 1,.., 4 ,
where k indicates the k’th point in the support. If we further assume that c1n , cn2 and cn3 are mutually independent then equation (1) can be written as a product of three components: P ( S nT = snT ,.., S n1 = sn1 | sn 0 , sn ,−1 , sn ,−2 , sn ,−3 , X n ) = Ln ,1 Ln ,2 Ln ,3 where Ln ,i =
T
∏ ∫ P( S
t =1, sn ,t −1 = i
nt
= snt | sn ,t −1 = i, D( sn ,t −1 ), X nt , c i ) P(ci | sn ,t −1 , D( sn ,t −1 ))dc i
This implies that we can estimate the three transition probabilities from each state separately. We test the influence of duration in the conditional distribution of the random effects by testing that the coefficient vectors α ik equal the zero-vector. This is done using a Lagrange multiplier test. The advantage of the LM test in our context is that it is sufficient to estimate the model under the null hypothesis. The transition probabilities are used to estimate the probability of being in a state. Suppose there are no characteristics in a Markov model (i.e. a homogeneous Markov model). Then equilibrium probabilities of being in a state can be derived. The derivation solves the equation: p* =Γ´p*, where Γ is the transition matrix, and p* is the equilibrium distribution over the states.7 A standard interpretation of p* is the distribution of individuals over the states when they have experienced (infinitely) many periods in the system. This interpretation is more troublesome when the transition matrix depends on characteristics that change over time, for example, age and experience. Keeping the characteristics fixed, we denote the solution to the equation p* =Γ´p* the steady state probability for individuals with those particular characteristics.
4. Data
The empirical study is based on longitudinal data sets from Danish administrative registers. One data set contains information on 40% of all immigrants in Denmark aged 15 and above
8
in the period 1984-1997. Furthermore, we exclude 2nd generation immigrants. In order to compare with natives, a second data set is used, which is based on a 2% panel sample of the whole Danish population aged 15 and above. Only men aged 30-50 are included, because they have completed their education, and they are not yet eligible for early retirement. Only annual 7The
adding up constraint in the transition matrix implies that a solution to the eigenvalue 1 exists. Then p* is the eigenvector corresponding to that eigenvalue. 8 For
a further description of immigrants in Denmark and the data set applied, see e.g. Husted et al. (2001).
10
observations are available on all important variables. The estimation period is 1988-1997. The observations 1984-87 are used for conditioning on spell durations.9 The final sample of immigrants contains 118,838 observations from 22,243 immigrant men, and the sample of natives contains 121,628 observations from 18,723 native men. All variables included are described in Appendix A. In the following, we discuss some of them. 10
11
The labor market states are self-employment (SE), wage-employment (WE), and nonemployment (NE). The literature debates whether unemployment and non-participation need to be treated as two distinct states. On the one hand, Clark and Summers (1979) find it unimportant to distinguish the two states because of the large number of transitions between them. On the other hand, Flinn and Heckman (1982) claim that the two are distinct, based on the fact that the unemployed search more intensively for jobs and hence have a higher 12
transition into employment than non-participants. Using a sample of youths, Goldsmith et al. (1995) find that joblessness (whether in terms of non-participation or unemployment) and the subsequent externality to the labor market have severe personal psychological effects, which are indistinguishable between the two states. Using a sample of mature males, Winkelmann and Winkelmann (1998) find that this conclusion falls as the psychic costs of unemployment are significantly higher than the psychic costs of non-participation. In Denmark during the sample period, the period for which UI members could gain eligibility for UIB was almost unlimited (> 7 years). This means that a fair number of unemployed were literally nonparticipants without real search activity. Smith (1998) finds that only 60% of the registered unemployed would also be defined to be unemployed according to the ILO-definition of unemployment, which requires search activity within the last four weeks before the survey and availability for a job within the next 2 weeks after the survey. Therefore, we choose the definition of states in accordance with Goldsmith et al. and Clark and Summers. Moreover, among our individual characteristics we include variables like indicators for eligibility of unemployment benefit, which might control for some of the differences between unemployed and non-participants. When an individual experiences more than one state in a year, the predominant state is chosen. Thus, individuals who are observed in the state NE are necessarily unemployed or 9
Furthermore, information for some of the individual characteristics are only available from 1988 and onwards.
10
Including agriculture (roughly 20% of native self-employed).
11
Including part-time employment, which is a small number.
11
outside the labor force for more than four months during a year, which means that we disregard temporary lay-offs and the main part of voluntary search unemployment. For some individuals, the variable describing the main activity is missing for some periods. If possible, the state is based on several labor market indicators, such as degree of individual unemployment, receipt of different public income transfers and unemployment insurance membership. The individual characteristics include indicator variables for eligibility and expiration of unemployment benefits. Eligibility and expiration of self-employment support are also available. In addition to entitlement and expiration indicators, the data also includes educational attainment, labor market experience, an indicator for living in a big city and finally an indicator for being a property owner. For immigrants, age at migration, country of origin and immigrant status (refugee/non-refugee) are observed. Included is also the concentration of immigrants in a local area and whether the individual is a Danish citizen and cohabits with a native. Further details are given in Appendix A. 4.1. Descriptive Statistics and Transitional Types Based on the Raw Data Table 2 shows the distribution of labor market states for natives and immigrants from 13
different countries of origin, consisting of the following groups: Scandinavia, EC-12, the Former Yugoslavia, other developed countries (DCs), Turkey, Pakistan, Vietnam, Iran, no citizenship (Palestine) and other less developed countries (LDCs). In 1997, the proportion of 14
non-employment was 17.8% for natives compared to 43% for immigrants. Among native males, the proportion of self-employment was 9% in 1997, which places Denmark at the bottom compared to the rest of the OECD. According to Blanchflower (2000) only Luxembourg, Norway and USA have lower rates of self-employment. Like most other OECD countries, the proportion of self-employment in Denmark has decreased during the last ten years (Blanchflower, 2000). For immigrants from LDCs, the proportion of self-employed is higher than for natives in 1997, except for the residual category of other LDCs. This tendency is confirmed for immigrants in other OECD countries (Fairlie and Meyer, 1996; Borjas, 1986; Yuengert, 1995). From 1988 to 1997 there is a steep increase in the proportion 12
In some sense Jones and Riddell (1999) represent a compromise between the two viewpoints, as they use transition probabilities to identify a new labor market state of marginal attachment to the labor force. 13
EC-12 comprises the 12 EU member states prior to the expansion in 1997.
14
This number is consistent with other statistics and conceals the effect of 10% non-participation, to which should be added unemployment that peaked at 13% in 1993.
12
of self-employed among immigrants from most LDCs. The steep increase in the proportion of self-employment among Turks and Pakistanis was coupled with an increase in nonemployment; for Iranians and immigrants without citizenship, it was coupled with an increase in wage-employment. This suggests that immigrants from different countries use selfemployment for different reasons. However, the pattern in Table 2 conceals variations due to changing background factors and individual characteristics.
Table 2
The Distribution on Labor Market States for Males Aged 30-50. Percent
Natives Immigrants Scandinavia EC-12 Former Yugoslavia Other DCs Turkey Pakistan Vietnam Iran No Citizenship Other LDCs
Self-employment¤)
Wage-employment
1988
1988
1992
1997
1992
Non-employment 1997
1988
1992
1997
11.5
10.5
9.0
74.2
70.6
73.2
14.3
18.9
17.8
9.8
10.7
11.1
58.0
49.8
45.7
32.2
39.6
43.1
9.7
9.8
8.6
62.8
59.1
62.9
27.5
31.2
28.4
11.3
12.4
12.4
67.4
61.6
60.5
21.3
26.0
27.1
6.0
5.8
3.3
68.2
58.4
35.9
25.8
35.8
60.9
10.4
12.0
10.7
60.0
53.3
55.6
29.5
34.7
33.7
5.8
10.8
15.9
58.0
45.6
44.3
36.2
43.6
39.7
15.7
21.0
22.3
53.7
38.3
38.4
30.6
40.6
39.3
9.4
9.9
10.7
50.9
40.2
49.7
39.7
49.9
39.6
4.4
6.3
16.0
17.4
34.6
33.0
78.1
59.1
51.0
5.2
7.0
14.4
14.0
21.4
17.1
80.8
71.5
68.5
9.7
9.1
8.7
49.5
42.5
38.9
40.8
48.3
52.4
¤) Including agriculture.
The differences between immigrants and natives can also be seen in some of the transition probabilities. In Table 3, raw transition probabilities are reported. Noticeable differences between immigrants and natives are seen in the transition probabilities out of selfemployment and wage-employment into non-employment. These probabilities are larger for immigrants than for natives, although many of the transition probabilities are similar for the two groups.
Table 3
Raw Transition Probabilities Natives
SEt
WEt
NEt
13
SEt-1
0.875
0.089
0.037
WEt-1
0.015
0.944
0.041
NEt-1
0.026
0.180
0.794
SEt-1
0.820
0.080
0.100
WEt-1
0.020
0.853
0.127
NEt-1
0.033
0.169
0.797
Immigrants
The definition of transitional types in section 2 summarizes transitional patterns. The transitional types ignoring individual characteristics are shown in Table 4. The first three columns refer to the transition probabilities used in conditions i) to iii). Column four shows the transitional type of the immigrant group as defined in Table 1. All the immigrant groups are of the relatively marginalized type.
Table 4
Identification of Types Ignoring Individual Characteristics Transition probabilities used in condition
Type
i)
ii)
iii)
Natives
0.294
0.732
0.126
Immigrants
0.556
0.864
0.163
1
Scandinavia
0.433
0.814
0.182
1
EC-12
0.427
0.817
0.134
1
Former Yugoslavia
0.324
0.920
0.069
2
Other DCs
0.517
0.816
0.173
1
Turkey
0.726
0.923
0.124
2
Pakistan
0.660
0.860
0.224
1
Vietnam
0.859
0.922
0.184
1
Iran
0.744
0.880
0.292
1
No Citizenship
0.866
0.890
0.360
1
Other LDCs
0.629
0.895
0.147
1
Note. The numbers in the columns i) to iii) are i) P ( S t = NE | S t −1 = SE , S t ≠ SE , S t-2 ) ii) P ( S t
= NE | S t −1 = WE , S t ≠ WE , S t-2 )
iii) P ( S t
= SE | S t −1 = NE , S t ≠ NE , S t-2 )
14
One purpose of estimating the discrete competing risks model in the next section is to determine how much of the results in Tables 2 and 4 are due to different (human capital) characteristics.
5. Results
In this section we discuss the estimation results of the discrete competing risks model in various ways to illustrate how the transition pattern of immigrants differs from that of natives. Subsection 5.1 discusses the estimates of the model. Subsection 5.2 illustrates how to identify the transitional type of a representative immigrant. Subsection 5.3 calculates steady state probabilities for the population of immigrants and natives, and analyzes the distribution of transitional types in the population of immigrants. 5.1. Estimates of Discrete Competing Risk Models In this subsection we comment on the estimated coefficients in relation to the different theories for self-employment. The estimated coefficients can be found in Tables B1 and B2 in Appendix B. We find a significant effect of living in an ethnic enclave on transition probabilities. A higher concentration of immigrants in the area of residence exerts a negative effect on all transition probabilities out of a state. For example, a higher concentration of immigrants lowers the probability of exiting self-employment, but it also lowers the probability of entry into self-employment from the two other states. The overall effect cannot be calculated directly. Another effect stems from liquidity constraints. Several variables control for this effect. One is Public Self-employment Support (PSS), which facilitates the process of establishing a business. One would expect the variables indicating entitlement to public self-employment support to increase the transition probabilities from non-employment to self-employment. One would also expect the variables indicating the expiration or non-existence of entitlement to public self-employment support to increase the transition probabilities from self-employment. The results in Table B2 confirm both expectations. Self-employment can be used to obtain entitlement to Unemployment Insurance Benefits (UIB). If self-employment is chosen for that reason, coefficients on entitlement to UIB should be positive for exit from self-employment to non-employment. This, however, applies to neither immigrants nor natives (see Tables B2 and B1, respectively).
15
There is evidence that the mixed logit specification is preferable to a multinomial logit specification. Unobserved heterogeneity plays a role in transition probabilities out of selfemployment and wage-employment. This could be a result of unobserved entrepreneurial characteristics of an individual. The duration dependence is negative in all states for both natives and immigrants. This is seen by the decreasing values of the coefficients to the duration dummies, γij. The negative duration dependence implies that the probability of making a transition out of a state decreases with the duration in that state. However, as is clear from the reported estimates and standard errors in Appendix B, this decrease is not always statistically significant. 5.2. Transitional Type of a Representative Immigrant In this subsection, we illustrate how to identify the immigrant types defined in section 2. The types are determined from the transition probabilities of an immigrant compared to a native conditional on the same characteristics and duration. For a representative individual, transition probabilities are calculated conditional on the 15
characteristics and the country-specific dummy variables.
The same representative
individual is used for each immigrant group. Therefore, the only difference in predicted transition probabilities among immigrants arises from the country-specific dummy variables, whereas differences from a native with the same characteristics can also arise from different estimates on parameters related to duration dependence and explanatory variables. We focus on the following ethnic origins: Turkey, Pakistan, Vietnam, Iran and no citizenship. Figure 1 shows the transition probabilities conditional on leaving the state for representative immigrants from each country as a function of the duration. The transition probability from self-employment to non-employment conditional on exit (see panel (a)) shows that immigrants and natives have about 90% probability of finishing short spells (less than two years) of self-employment with a transition into non-employment. At elapsed durations above two years, immigrants have a substantially higher probability of exit to nonemployment compared to natives. After an elapsed duration of five years, natives have 50% probability of exit to non-employment and 50% probability of exit to wage-employment, whereas the corresponding probabilities for immigrants are 75% and 25%.
15
The characteristics of the representative individual are mean of X’s. As regards the unobserved components, the individual is assumed to have points of support equal to the weighted average ( Ψi ' π i ).
16
The transition probabilities from wage-employment (see panel (b)) show that for both natives and immigrants, the most likely transition is to non-employment, namely with probabilities between 86% and 98%. There are, however, differences between immigrants. Turkish and Vietnamese immigrants always have higher probabilities of exiting to nonemployment compared to natives, whereas the opposite is the case for Iranians and individuals with no citizenship.
Figure 1
Transition Probabilities Conditional on Exit of State for Representative Individuals 1.00
1.00
0.98 0.96 Prob. of exit t NE
Prob. of exit to NE
0.90 0.80 0.70 0.60 0.50
0.94 0.92 0.90 0.88 0.86 0.84 0.82
0.40 1
2
3
4
5
0.80 1
DurationinSE
2
3
4
Duration in WE
(a)
(b)
0.4 0.35
prob of exit to SE
0.3 0.25 0.2 0.15 0.1 0.05 0 1
2
3
4
5
Duration in NE
(c) Labels: ■ Native, ∆ Turkey, × Pakistan, t Vietnam, ● Iran, + No state
17
5
The representative immigrants all have a higher probability than natives of becoming selfemployed when exiting non-employment (see panel (c)). For natives there is about a 10% probability of an exit to self-employment and a 90% probability of wage-employment after a spell of non-employment. The corresponding figures for immigrants are 20% exit to selfemployment and 80% exit to wage-employment. The decreasing duration dependence implies that the conclusion about the transitional type of an immigrant may change over the duration of a spell.16 For instance, the representative individual from Vietnam is concluded to be Transitional Type 1 regardless of elapsed duration, whereas a representative individual from Turkey is concluded to be Transitional Type 1 for most durations, but for elapsed duration in self-employment below 3 years, the individual is concluded to be Transitional Type 8. There is no simple conclusion for the representative individuals from the other countries of origin. Large differences show up when we compare the conclusions about transitional types from Figure 1 to those in Table 4, where transitional types were identified while ignoring individual characteristics. When ignoring individual characteristics, all immigrants are concluded to be relatively marginalized (Transitional Type 1). Control for individual characteristics gives a much more complex picture. Two important implications can be inferred. First, the fact that most immigrants were characterized as relatively marginalized is partly caused by differences in individual characteristics between immigrants and natives. Second, even when controlling for individual characteristics, the existence of relatively marginalized individuals among certain immigrant groups persists. Analysis based on a representative individual does not necessarily allow us to rule out that the phenomenon of relative marginalization is economically important. For instance, the fact that the Pakistani representative individual is not relatively marginalized might conceal the fact that ‘only’ one third of the Pakistani immigrants are relatively marginalized. Therefore, we proceed with analysis of the entire population. 5.3. Analysis of Steady State Probabilities and Transitional Types in the Population In this subsection we calculate the steady state distribution of the population across labor market states and derive the distribution of transitional types in the population.
16
Panel a, b and c corresponds to conditions i), ii) and iii), respectively (see section 2). Whenever the graph for an immigrant group lies above that of the natives, the condition is fulfilled.
18
To calculate the transition probabilities for an immigrant, we need to choose the length of duration in each state. Typically, it is not possible to do this based on the actual transition history of an immigrant because most immigrants only experience one or two transitions in the sample period. It is possible, however, to calculate the distribution of durations for any immigrant given his characteristics. The probability of a duration of length d in a state j at time t is: d −1
P( S ∏ τ =1
n ,t −τ
= j | Sn ,t −τ −1 = j , Dn ,t −τ ( j ) = d − τ , xnj ,t −τ ) = 1 − P( Sn ,t = j | S n,t −1 = j , Dnt ( j ) = d , xnjt )
The steady state distribution across the three labor market states for the population based on the competing risks model is reported in the first three columns in Table 5.17 In general, the steady state probabilities exhibit the same pattern as the marginal probabilities reported in Table 2. Again, we find that immigrants are more likely than natives to be self-employed or non-employed, and this tendency is more pronounced for immigrant groups from Former Yugoslavia, Turkey, Pakistan and for immigrants with no citizenship. The fact that the steady state probability does in fact mirror the marginal distribution indicates that the model gives a good description of the data.
Table 5
Steady State Probabilities for the Population in 1997 SE
WE
NE
Natives
0.058
0.818
0.124
Scandinavia
0.097
0.678
0.225
EC-12
0.134
0.695
0.172
Former Yugoslavia
0.117
0.561
0.322
Other DCs
0.125
0.683
0.193
Turkey
0.176
0.464
0.361
Pakistan
0.224
0.447
0.329
Vietnam
0.252
0.506
0.242
Iran
0.340
0.394
0.267
No Citizenship
0.343
0.251
0.406
17
The steady state probabilities are the elements of p*, which is found by solving p* =Γ´p*, see section 3.
19
Other LDCs
0.106
0.589
0.305
We proceed by identifying the transitional type of each individual based on the model to be able to derive the distribution of transitional types in the entire population. The identification of transitional types is based on a comparison of transition probabilities, for instance, as stated in conditions i) to iii) for Transitional Type 1. Since a condition may be either satisfied or not, this leads to the eight different transitional types. To satisfy, say, condition i), there may be a small or a large difference between the transition probabilities for the immigrant and the corresponding (fictional) native. If the difference is small, the identification of a type is weak in the sense that a small change in the transition probability can change the type. Suppose the transition probabilities in each of the three conditions are close to those of natives. Assuming independence between transition probabilities from different states, there is about a 1/8 probability of observing each type. After identifying the type of an immigrant, we calculate the proportion of each type for all immigrant groups. Table 6 shows the results. The first column refers to relatively marginalized immigrants who use self-employment as an escape state, and it shows that more than 50% of the immigrants from Turkey, Vietnam, Iran and persons with no citizenship are of the self-employed marginalized type (Transitional Type 1). The proportion of Transitional Type 1 among Scandinavians and immigrants from other DCs is about 1/8, which means that they behave much like natives. Transitional Type 2 also refers to relatively marginalized individuals, but they use wage-employment as an escape state for non-employment. This type seems to be over-represented among immigrants from Vietnam and other LDCs. When we move on to the relatively entrepreneurial type (Transitional Type 3 and 4) and relatively employed type (Transitional Type 5 and 6), there is a remarkable difference between immigrants from Western countries and immigrants from non-Western countries. Immigrants from non-Western countries are much more likely to be relatively entrepreneurial than immigrants from Western countries. The opposite is true for the relatively employed. However, the table shows that the dominating transitional type among immigrants from nonWestern countries is Transitional Type 1. These results are in accordance with SchultzNielsen’s findings. Based on interviews with 232 self-employed Danish immigrants from nonWestern countries, Schultz-Nielsen (2001) found that 18% became self-employed because they were unable to get a job, and she reports even higher figures for Sweden, whereas Constant and Schultz-Nielsen (2004) find that the self-employed immigrants seem
20
disadvantaged, since they earn much lower wages than the wage-employed immigrants. This is in contrast to immigrants in Germany.
Table 6
The Distribution of Transitional Types in Population of Immigrants 1997 Transitional Types 1
2
3
4
5
6
7
8
Scandinavia
0.185
0.096
0.030
0.286
0.151
0.127
0.000
0.121
EC-12
0.073
0.071
0.010
0.254
0.252
0.289
0.000
0.048
Former Yugoslavia
0.000
0.020
0.000
0.01
0.804
0.157
0.000
0.006
Other DCs
0.163
0.073
0.110
0.355
0.074
0.184
0.010
0.034
Turkey
0.639
0.117
0.188
0.05
0.003
0.007
0.010
0.001
Pakistan
0.361
0.000
0.356
0.279
0.001
0.000
0.000
0.001
Vietnam
0.639
0.153
0.202
0.000
0.000
0.000
0.010
0.000
Iran
0.517
0.000
0.352
0.129
0.000
0.000
0.000
0.001
No Citizenship
0.613
0.000
0.383
0.000
0.000
0.000
0.000
0.000
Other LDCs
0.284
0.337
0.108
0.147
0.029
0.069
0.018
0.008
Origin
Differences in human capital cannot explain why a large proportion of immigrants are categorized as relatively marginalized. In Table 4, section 4, immigrants are labeled a certain transitional type based only on country of origin. Except for immigrants from the Former Yugoslavia and Turkey, everyone else was concluded to be relatively marginalized. When controlling for individual characteristics, it is seen in Table 6 that a large proportion of the immigrant population from Turkey, Vietnam, Iran and persons with no citizenship is still concluded to be of the relatively marginalized type. The robustness of our results in Table 6 is investigated by calculating the density of the differences between transition probabilities of immigrants and their corresponding (fictional) natives. The densities are calculated from the left-hand side of conditions i) to iii). Figure 2 shows the densities for these left-hand sides for each country. For Vietnam, the country of origin with the highest proportion of Transitional Type 1, the density of the left-hand side in condition i) is located on the positive part of the axis. This implies that condition i) is satisfied for all Vietnamese immigrants. The left-hand sides of condition i) are quite large, for instance larger than 0.25 for half of the Vietnamese immigrants. In other words, a Vietnamese immigrant always has a higher probability than a native of entering non-employment from 21
self-employment. For Turkey, Iran and individuals with no citizenship, which constitute the other countries of origin with a high proportion of Transitional Type 1, a similar picture is seen. The graph also reveals why we do not find any Transitional Type 1 among immigrants from the Former Yugoslavia, since conditions i) and iii) are concentrated on the negative part of the axis. For Scandinavians and other DCs, the countries with about a 1/8 proportion of Transitional Type 1, the densities are concentrated about 0. The figure also reveals that there is quite some variation in the transition probabilities from self-employment to nonemployment. The overall conclusion is that the determination of transitional types for the countries with a high proportion of Transitional Type 1 is robust since the densities of the differences of the left-hand sides of conditions i) to iii) are not concentrated about 0. Figure 2
Distributions of Left-hand Sides in Conditions i) to iii).
EC 12
ExYugo
Iran
Other DC
Scandinavia
No Citizenship
Pakistan
Turkey
Other LDC
22
Vietnam ____: Lhs condition i) (from SE), - - - -: Lhs condition ii) (from WE), ------: Lhs condition iii) (from NE)
6. Conclusion
In many developed countries, immigrants are observed to have both a large proportion of nonemployment and self-employment. We propose a framework to investigate both observations simultaneously. In contrast to the rest of the literature, we test the hypothesis by scrutinizing transition patterns rather than focusing on durations or the signs of a few key coefficients. We propose a method to identify the marginalized by using transition probabilities between selfemployment, wage-employment and non-employment for immigrants compared to natives. We find evidence that immigrants have a very different transition pattern between labor market states compared to natives. In general, immigrants are more likely than natives to end in non-employment, but also in self-employment. Furthermore, a closer investigation of the transition pattern suggests that some immigrant groups are marginalized in the Danish labor market. The empirical analysis indicates that especially many Turks, Vietnamese, Iranians and immigrants with no citizenship are marginalized. For those groups, more than half of the immigrants are characterized as marginalized, and they are more likely to use selfemployment as an escape route out of non-employment. This indicates that self-employment is indeed used as a last resort for these groups. Besides documenting huge differences in the transition pattern between immigrants and natives, we also seek to explain why these differences arise. Our analysis suggests that for the Turks, Vietnamese, Iranians and immigrants with no citizenship the differences in human capital (measured by education and experience) cannot explain the difference in the transitional pattern. This indicates that there might be barriers for particular immigrant groups in the Danish labor market. In the ordinary labor market an immigrant only finds employment if the value of his marginal product exceeds the minimum wage (or equivalently: the social assistance). For an unskilled immigrant, this might be a binding constraint that represents an
23
entry barrier. In self-employment no similar constraint exists. When identifying selfemployed marginalized immigrants, we control for a number of individual-specific characteristics that are related to the human capital characteristics and time-invariant unobserved characteristics. Hence, the barriers arise from other unobserved characteristics (e.g. language proficiency or other time-varying qualifications) or discrimination. None of these barriers would persist after starting an immigrant business with customers from the same ethnic origin as the owner. One policy instrument is to change the level of education among immigrants. However, we find that the impact of more education on steady state probabilities of a representative immigrant is low. Providing representative immigrants with 5 additional years of education leads to lower steady state probabilities of non-employment as would be expected from the coefficients on education, but the effect is rather modest. Similarly, eligibility and expiration of unemployment benefits and self-employment support are also potential policy instruments, but the effects are modest. According to our study, politicians should focus on equalizing behavior rather than on formal education. That would include tearing down entry barriers due to lacking language proficiency, lack of informal skills or discrimination. Our approach to analyzing marginalization differs from earlier studies, which usually focus on the survival probability in non-employment. We identify marginalized immigrants based on their transitions in and out of all the states compared to natives. In addition to shedding light on marginalization among immigrants, our method also casts new light on effects that prompt self-employment. For instance, our method identifies immigrants with a selfemployment tradition and immigrants who use self-employment as a steppingstone to wageemployment. For future research, it would be very interesting to apply this new method to other countries to investigate the existence of self-employed marginalized groups. A cross-country comparison may also help identify the specific barriers on the labor market that cause immigrants to be relatively marginalized. Although we have only used the method in this study, we believe the method can be applied in other contexts.
24
Appendix A
Description of the Variables
The data is register-based with annual observations. Our definition of self-employment relies on two variables concerning employment status. The primary variable is a pure register-based variable and relates to the dominating employment status during the year, while the secondary variable is constructed on the basis of several other variables. We only use the secondary variable in cases where the primary variable is missing. If the secondary variable is also missing, we cannot determine the state of employment, which will therefore be missing. Furthermore, if a person is registered with employment other than self-employment, but receives public self-employment support, we treat him as self-employed. Table A1 Explanatory Variables Natives Variable name
Description
mean
γ1
Elapsed duration 1 year in the current state
γ2
Elapsed duration 2 years in the current state
γ3
Elapsed duration 3 years in the current state
γ4
Elapsed duration 4 years in the current state
γ4+
Elapsed duration >4 years in the current state
Immigrants std
mean
std
0.308
0.462
0.102
0.022
Im_year_unknown
Dummy (1 if date of immigration missing)18
Local U
Local unemployment rate
Conc_immi
Concentration of immigrants (all immigrants)
0.059
0.035
EC-12
0.272
0.445
Former Yugoslavia
Country indicator (European Community prior to the expansion in 1997) Country indicator
0.056
0.230
ODC
Country indicator (Other Developed Countries)
0.110
0.313
Turkey
Country indicator
0.092
0.289
Pakistan
Country indicator
0.06
0.229
Vietnam
Country indicator
0.018
0.133
Iran
Country indicator
0.041
0.197
No-state
Country indicator (No citizenship)
0.025
0.155
OLDC
Country indicator (Other Less Developed Countries)
0.177
0.382
18
0.103
Predominantly immigrants who entered Denmark before 1970.
25
0.024
Education
Length of education in Denmark
0.111
0.037
0.049
0.060
Experience
Labor market experience in Denmark in years
0.164
0.077
0.09
0.078
Years_U_in DK
Years spent unemployed in Denmark
0.036
0.045
Single
Dummy (1 if single)
0.274
0.446
Age_migration
Age when immigrated to Denmark
0.342
0.092
Refugee
Dummy (1 if considered a refugee)
0.184
0.388
DK citizen
Dummy (1 if hold Danish citizenship)
0.369
0.482
DK partner
Dummy (1 if Danish partner)
0.353
0.478
Child 0-2
Dummy (1 if children aged 0-2 years)
0.067
0.250
0.127
0.333
Child 3-9
Dummy (1 if children aged 3-9 years)
0.231
0.421
0.308
0.462
Child 10-17
Dummy (1 if children aged 10-17 years)
0.417
0.493
0.39
0.488
Big city
Dummy (1 if lives in a big city)
0.306
0.461
0.570
0.495
Prop. Owner
Dummy (1 if property owner)
0.684
0.465
0.332
0.471
0.237
0.425
Indicators for the individual being entitled to Public Self-employment Support (PSS) PSS11
=1 if entitled to PPS and it is the first period of 0.035 0.184 0.066 0.249 the entitlement =0 otherwise PSS12 =1 if entitled to PPS apart from the first period 0.045 0.208 0.130 0.336 of the entitlement (i.e. PSS11=0) =0 otherwise Indicators for the individual NOT being entitled (including expiration within the current year) to Public Self-employment Support (PSS) PSS21 =1 if entitlement to PPS expires within the 0.059 0.236 0.108 0.310 current year (conditional on being entitled in the current year, i.e. PSS11=1 or PSS12 =1). =0 otherwise PSS22 =1 if entitlement to PPS expired more than 1 0.901 0.299 0.782 0.413 year ago (including never existed). =0 otherwise Indicators for the individual being entitled to Unemployment Insurance Benefits (UIB) UIB11
=1 if entitled to UIB and it is the first period of 0.034 0.182 0.042 the entitlement =0 otherwise UIB12 =1 if entitled to UIB apart from the first period 0.641 0.480 0.529 of the entitlement (i.e. UIB11=0) =0 otherwise Indicators for the individual NOT being entitled (including expiration within the current year) to Unemployment Insurance Benefits (UIB) UIB21 =1 if entitlement to UIB expires within the 0.023 0.15 0.03 current year (conditional on being entitled in the current year, i.e. UIB11=1 or UIB12 =1). =0 otherwise UIB22 =1 if entitlement to UIB has expired within the 0.054 0.226 0.072 last 3 years apart from the current year (i.e. UIB21=0). =0 otherwise
26
0.202
0.499
0.180
0.259
UIB23
=1 if entitlement to PPS expired more than 3 year ago (including never existed). =0 otherwise
0.271
0.445
0.357
0.479
Extended Explanations Regarding Some Explanatory Variables
Entitlement to unemployment insurance benefits (UIB) requires membership of an unemployment insurance fund for more than one year and at least 26 weeks of employment 19
within the last three years. Until 1993 the entitlement expired after three years, whereas after 1993 special circumstances (e.g. participation in a re-employment program) might justify UIB for up to seven years. On the basis of these rules and using information from the unemployment registers, variables concerning eligibility are constructed. The first set of variables (‘UIB11’ and ‘UIB12’) describes whether the individual is entitled to UIB. The first variable (‘UIB11‘) describes whether the current year is the first year of entitlement, while the second variable (‘UIB12‘) describes whether the individual has been entitled for more than 1 year. The variables ‘UIB11’ and ‘UIB12’ are mutually exclusive. The second set of variables (‘UIB21’, ‘UIB22’ and ‘UIB23’) describes whether entitlement to UIB has expired or never existed. The first variable (‘UIB21‘) describes whether the entitlement expires within the current year (conditional on being entitled within the current year), while the second variable (‘UIB22‘) describes whether it expired within the last 3 years excluding the current year. The last variable (‘UIB23‘) describes whether the entitlement expired more than 3 years ago or never prevailed. The variables ‘UIB21’, ‘UIB22’ and ‘UIB23’ are mutually exclusive. During the period of consideration, different rules for public self-employment support (PSS) prevailed. Entitlement presumes UIB entitlement plus at least five months of unemployment within the last eight months. The PSS expires after approximately three years or if the labor market status changes. Along with the rules for entitlement, the rules for expiration have changed during our sample period. The first set of variables (‘PSS11’ and ‘PSS12’) describes whether the individuals are entitled to PSS. The first variable (‘PSS11‘) describes whether the individual is entitled to PPS and whether the current year is the first year of the entitlement, while the second variable (‘PSS12‘) describes whether the individual is entitled to PPS and has been entitled for more than 1 year. The variables ‘PSS11’ and ‘PSS12’ are mutually exclusive. The second set of variables (‘PSS21’ and ‘PSS22’) describes whether entitlement to PSS has expired. The first variable (‘PSS21‘) describes whether the
19
Different rules apply to e.g. students and individuals on leave.
27
entitlement expires within the current year (conditional on being entitled within the current year), while the second variable (‘PSS22‘) describes whether the entitlement expired more than one year ago or never existed. The variables ‘PSS21’ and ‘PSS22’ are mutually exclusive.
28
Appendix B Table B1 Results from Estimation of Multinomial Models for Natives Transitions out of NE to SE WE
γ1 γ2 γ3 γ4 γ4+ Local U Education Experience Single Child 0-2 Child 3-9 Child 10-17 Big city Prop. Owner PSS11 PSS12
-0.770 0.645 -1.029 0.661 -1.353** 0.681 -1.435** 0.688 -3.294** 0.682 -1.228 2.809 3.791** 1.582 -6.462** 1.001 -0.455** 0.166 0.342 0.225 0.170 0.165 0.394** 0.148 0.175 0.153 0.891** 0.155 0.344** 0.167 0.400** 0.176
-1.235** 0.143 -1.031** 0.145 -1.585** 0.157 -1.926** 0.170 -2.912** 0.158 1.843* 0.976 2.498** 0.573 0.672* 0.364 -0.040 0.058 0.277** 0.102 0.166** 0.068 0.202** 0.058 -0.005 0.053 0.119** 0.055
PSS21 PSS22
Transitions out of SE to WE NE -3.215** 0.463 -3.637** 0.465 -3.782** 0.466 -3.862** 0.465 -4.370** 0.459 4.956** 1.570 1.755* 0.990 6.535** 0.627 0.050 0.101 0.141 0.130 0.140* 0.080 0.042 0.072 0.061 0.088 -0.225** 0.095
-1.312** 0.621 -1.591** 0.634 -1.823** 0.641 -1.835** 0.647 -2.397** 0.637 5.772** 1.976 -6.265** 1.055 1.631** 0.696 0.408** 0.116 0.104 0.175 -0.222* 0.121 -0.092 0.103 0.169 0.105 -0.678** 0.107
2.080** 0.363 1.546** 0.337
0.387 0.251 -0.903** 0.208 -0.606** 0.213 -0.574** 0.097
UIB11 UIB12 -0.695** 0.309 -0.184 0.189 -0.298* 0.177
UIB21 UIB22 UIB23
-1.067** 0.112 -0.854** 0.067 -1.024** 0.058
Transitions out of WE to SE NE 0.237 0.303 -0.370 0.319 -0.495 0.331 -0.627* 0.340 -0.822** 0.323 -1.104 1.551 -0.083 1.066 -9.468** 0.709 -0.138 0.106 -0.044 0.125 0.057 0.082 0.119 0.075 -0.133 0.090 0.569** 0.093
0.193** 0.090 -0.263** 0.051
Unobserved heterogeneity†: µ1
-2.759**
-1.787**
29
-0.764** 0.142 -1.345** 0.145 -1.694** 0.150 -1.703** 0.156 -1.838** 0.142 8.972** 0.949 -8.888** 0.597 -8.680** 0.415 0.314** 0.054 -0.314** 0.092 -0.419** 0.061 -0.383** 0.051 -0.131** 0.052 -0.682** 0.052
-3.249**
2
µ
π12 π10 π02 π00 N L(full) L(const) Pseudo R2 **
0.331 0 0 0.912** 0.059 0.088 0.059 0 0 0 0
0.227 0.238 0.638 0.802** 0.085 0.198** 0.085 0 0 0 0
17033 -7669.1 -8970.5 0.145
16406 -6274.3 -6719.3 0.066 *
0.140 2.366** 0.089 0.145** 0.019 0.773** 0.027 0.001 0.003 0.081** 0.020 88239 -17426.8 -18867.6 0.076
Note: indicates significance at the 5% level, and indicates significance at the 10% level. L(const) is the likelihood value from an estimation including a constant term only and a correction for unobserved heterogeneity. The Pseudo R2 is calculated as: Pseudo R2=1-(L(full)/L(const)). Numbers below the estimates of the coefficients are their standard errors. † In some estimations, the full parametrization of the unobserved heterogeneity could not be identified. In that case, some of the parameters π and µ are restricted to 0 to achieve identification.
30
TABLE B2
Results from Estimation of Multinomial Models for Immigrants Transitions out of NE to SE WE
γ1 γ2 γ3 γ4 γ4+ Im_year_unkn. Local U Conc_immi EC-12 Ex-Yugoslavia ODC Turkey Pakistan Vietnam Iran No-state OLDC Education Experience Years_U_in DK Single Age_migration Refugee DK citizen
-1.806** 0.358 -1.991** 0.364 -2.096** 0.371 -2.243** 0.376 -2.643** 0.370 0.552** 0.138 5.983** 1.504 -7.381** 1.286 0.024 0.133 -0.770** 0.245 0.174 0.133 0.000 0.157 0.596** 0.151 -0.009 0.227 0.618** 0.140 0.341** 0.162 0.059 0.125 3.271** 0.647 -6.302** 0.829 -3.854** 1.106 -0.376** 0.091 -5.557** 0.704 -0.040 0.118 0.130* 0.078
0.238 0.170 0.775** 0.170 0.379** 0.174 0.280 0.179 -0.252 0.177 0.171** 0.070 1.035 0.748 -9.192** 0.618 0.217** 0.059 -0.114 0.087 0.217** 0.069 0.103 0.070 0.241** 0.077 0.059 0.123 0.136 0.090 -0.394** 0.113 0.293** 0.059 1.658** 0.348 0.603* 0.364 -5.192** 0.551 -0.050 0.043 -3.628** 0.339 0.047 0.055 -0.066* 0.039
Transitions out of SE to WE NE -1.654** 0.572 -1.708** 0.569 -2.174** 0.573 -2.043** 0.583 -2.628** 0.575 -0.193 0.206 4.677** 2.006 -6.229** 1.824 -0.286** 0.136 -0.509* 0.294 -0.094 0.168 -0.703** 0.217 -0.164 0.205 -1.818** 0.574 -0.875** 0.320 -1.238** 0.509 -0.206 0.158 1.943** 0.877 5.441** 0.935 -4.312** 1.400 0.114 0.129 -1.598* 0.963 -0.037 0.182 -0.059 0.102
31
-1.512* 0.774 -1.477* 0.788 -1.628** 0.794 -1.965** 0.790 -2.186** 0.793 -0.234 0.156 5.173** 1.630 -8.710** 1.495 -0.259** 0.131 -0.904** 0.350 0.028 0.144 0.412** 0.170 0.499** 0.162 -0.254 0.292 -0.329 0.210 0.015 0.261 0.436** 0.138 -0.299 0.751 -0.976 0.807 -1.003 1.181 0.176* 0.094 1.305* 0.778 0.129 0.146 -0.008 0.082
Transitions out of WE to SE WE -0.295 0.458 -0.711 0.457 -0.888* 0.471 -0.663 0.480 -1.082** 0.464 0.622** 0.172 1.727 1.658 -7.183** 1.718 0.105 0.120 -0.377* 0.229 0.357** 0.155 0.361* 0.185 1.057** 0.210 0.211 0.421 0.850** 0.280 1.233** 0.442 0.159 0.143 0.083 0.819 -4.955** 0.789 1.289 1.405 0.026 0.112 -2.478** 0.789 -0.620** 0.167 0.002 0.093
-2.081** 0.211 -2.807** 0.210 -3.046** 0.209 -3.078** 0.210 -3.438** 0.204 0.525** 0.071 4.394** 0.755 -12.666** 0.692 0.144** 0.055 0.222** 0.081 0.071 0.068 1.000** 0.074 0.982** 0.083 0.458** 0.148 0.548** 0.121 0.715** 0.175 0.511** 0.060 -2.838** 0.359 -6.365** 0.371 4.392** 0.581 0.181** 0.044 2.381** 0.342 -0.336** 0.062 -0.164** 0.040
DK partner Child 0-2 Child 3-9 Child 10-17 Big city Prop. Owner PSS11 PSS12
0.007 0.091 0.025 0.087 0.127* 0.073 0.159** 0.072 0.298** 0.086 0.458** 0.090 0.730** 0.090 0.630** 0.090
0.238** 0.045 0.010 0.047 0.041 0.037 0.045 0.037 0.335** 0.042 -0.035 0.045
PSS21 PSS22
0.305** 0.109 -0.143 0.136 -0.085 0.097 -0.092 0.093 0.269** 0.122 0.138 0.098
-0.080 0.095 0.079 0.100 -0.292** 0.084 -0.238** 0.079 0.344** 0.099 -0.459** 0.090
1.535** 0.238 1.003** 0.218
1.075** 0.173 0.118 0.155 -0.182 0.131 -0.491** 0.077
UIB11 UIB12 UIB21 UIB22 UIB23
-0.204 0.159 0.052 0.111 -0.173* 0.095
-0.659** 0.070 -0.987** 0.051 -1.268** 0.039
Unobserved heterogeneity† 0 µ1 0 0 µ2 0 0 π12 0 0 π10 0 0 π02 0 0 π00 0 N L(full) L(const) Pseudo R2 **
-2.005** 0.204 -1.194** 0.293 0.513* 0.358 0.223 0.357 0.143 0.159 0.121 0.155
42696 -19514.9 -21782.0 0.104
13700 -6704.1 -7129.9 0.060 *
-0.056 0.101 0.057 0.111 0.057 0.086 0.013 0.082 0.353** 0.107 0.363** 0.088
-0.383** 0.044 0.000 0.049 -0.066* 0.038 -0.068* 0.036 0.517** 0.044 -0.546** 0.041
0.244** 0.058 0.017 0.037
-3.011** 0.158 1.405** 0.078 0.380** 0.067 0.529** 0.071 0.014 0.018 0.077** 0.028 62442 -23018.8 -25395.2 0.094
Note: indicates significance at the 5% level, and indicates significance at the 10% level. L(const) is the likelihood value from an estimation including a constant term only and a correction for unobserved heterogeneity. The Pseudo R2 is calculated as: Pseudo R2=1-(L(full)/L(const)). Numbers below the estimates of the coefficients are their standard errors. † In some estimations, the full parametrization of the unobserved heterogeneity could not be identified. In that case, some of the parameters π and µ are restricted to 0 to achieve identification.
32
References
Alba-Ramirez, A. (1994), “Self-employment in the Midst of Unemployment: The Case of Spain and the United States.” Applied Economics 26 (3): 189-204. Aldrich, H. and R. Waldinger (1990), “Ethnicity and Entrepreneurship.” Annual Review of Sociology 16(1): 111-135. Arulampalam, W. (2001), “Is Unemployment Really Scarring? Effects of Unemployment Experiences on Wages.” Economic Journal 111 (475): F585-606. Bager, T. and S. Rezaei (2001), “Immigrant Businesses in Denmark: Captured in Marginal Business Fields?” CESFO WP No. 2001/1, Centre for Small Business Studies, University of Southern Denmark. Blanchflower, D. G. (2000), “Self-employment in OECD Countries.” Labour Economics 7: 471-505. Blanchflower, D. G. and A. J. Oswald (1998), “What Makes an Entrepreneur?” Journal of Labor Economics 16 (1):26- 60. Borjas, G. (1986), “The self-employment Experience of Immigrants.” Journal of Human Resources 21(4): 486-506. Borjas, G. and S. G. Bronars (1989), “Consumer Discrimination and Self-Employment.” Journal of Political Economy 97(3): 581-605. Carrasco, R. (1999), “Transition to and from Self-Employment in Spain: An Empirical Analysis.” Oxford Bulletin of Economics and Statistics 61 (3): 315-341. Clark, K. and S. Drinkwater (2000), “Pushed Out or Pulled In ? Self-Employment among Ethnic Minorities in England and Wales.” Labour Economics 7:603-628. Clark, K. B. and L. H. Summers (1979), “Labor Market Dynamics and Unemployment: A Reconsideration.” Brookings Papers on Economic Activity 1:13-72. Constant, A. and M.-L. Schultz-Nielsen (2004), “Immigrant Self-Employment and Economic Performance.” In Tranæs and Zimmerman (eds.), Migrants, Work and the Welfare State. UP of Southern Denmark: Odense. Constant, A. and K. Zimmermann (2004), “Self-Employment Dynamics Across the Business Cycle: Migrants Versus Natives” IZA DP # 1386. Evans, D. S. and B. Jovanovic (1989), “An Estimated Model of Entrepreneurial Choice under Liquidity Constraints.” Journal of Political Economy 97(4) 774-806. Evans, D. S. and L. S. Leighton (1989), “Some Empirical Aspects of Entrepreneurship.” American Economic Review 79 (3): 519-535.
33
Fairlie, R. W. and B. D. Meyer (1996), “Ethnic and Racial Self-Employment Differences and Possible Explanations.” Journal of Human Resources 31:757-793. Flinn, C. and J. Heckman (1982), “New Methods for Analyzing Structural Models of Labor Force Dynamics.” Journal of Econometrics 18: 115-168. Goldsmith, A. H., J. R. Veum and W. Darity (1995), “Are Being Unemployed and Being out of the Labor Force Distinct States?: A Psychological Approach.” Journal of Economic Psychology 16 (2): 275-295. Hammarstedt, M. (2001), “Immigrant Self-employment in Sweden- It’s Variation and some Possible Determinants.” Entrepreneurship and Regional Development 13: 147-161. Hout, M. and H. S. Rosen (1999), “Self-employment, Family Background and Race.” NBER Working Paper #7344. Husted, L., H. S. Nielsen, M. Rosholm, and N. Smith (2001), “Employment and Wage Assimilation of Male First Generation Immigrants in Denmark.” International Journal of Manpower 22 (1/2) 39-68. Jones, S. R. G. and W. C. Riddell (1999), “The Measurement of Unemployment: An Empirical Approach.” Econometrica 67: 147-162. Light, I. (1984), “Disadvantaged Minorities in Self-employment.” International Journal of Comparative Sociology 20: 31-45. Lindh, T. and H. Ohlson (1996), ”Self-Employment and Windfall Gains: Evidence from the Swedish Lottery.” Economic Journal 106: 1515-1526. Machin, S. and A. Manning (1999), “The Causes and Consequences of Longterm Unemployment in Europe.” Ch. 47 in Ashenfelter, O. and D. Card (eds.) Handbook of Labor Economics. Elsevier: North Holland. Moore, C. S. and R. W. Mueller (2002), “The Transition from Paid to Self-Employment in Canada: The Importance of Push Factors.” Applied Economics 34: 791-801. Pfeiffer, F. and F. Reize (2000), “Business Start-Ups by the Unemployed - An Econometric Analysis Based on Firm Data.” Labour Economics 7: 629-663. Schultz-Nielsen, M. L. (2001),“The Integration of Non-Western Immigrants in a Scandinavian Labour Market: The Danish Experience.” The Rockwool Foundation Research Unit, Copenhagen. Smith, N. ed. (1998), Arbejde, Incitamenter og Ledighed (In Danish, Work, Incentives and Unemployment). Rockwool Fondens Forskningsenhed, Aarhus Universitetsforlag. Taylor, M. P. (1996), “Earnings, Independence or Unemployment: Why Become Selfemployed?” Oxford Bulletin of Economics and Statistics 58: 253-266.
34
Taylor, M. P. (2001), ”Self-employment and Windfall Gains in Britain: Evidence from Panel Data.” Economica 68: 539-566. Train, K. (2003), Discrete Choice Methods with Simulation. Cambridge University Press, UK. Winkelmann, L. and R. Winkelmann (1998), “Why Are the Unemployed So Unhappy? Evidence from Panel Data.” Economica 65: 1-15. Yuengert, A. M. (1995), ”Testing Hypotheses of Immigrant Self-Employment.” Journal of Human Resources 30(1):194-204.
35
