Does Flexibilization Lead to Migration?
Universiteit Leiden
Master Thesis
Public Administration – Economics & Governance Track
10-6-2019
Boaz Kaarsemaker
S2290383
Supervision:
Dr. Alexandre Afonso
Prof. dr. Olaf van Vliet
Abstract:
This paper analyses the connection between Employment Protection Legislation (EPL) and migration inflows in OECD countries during a 15-year period. The paper argues that either lenient or strict EPL could lead to higher inflows of migration and starts from Piore’s (1979) assumption, that migrants will accept jobs of lower social status, income and opportunities of advancement than natives. Fixed effects models were estimated on cross-sectional time-series
data. Evidence was found for a significant negative effect of EPL on labour migration inflows, although no effect was found on total inflows.
Index Introduction ... 1 Theory ... 3 Lenient EPL ... 3 Strict ELP ... 5 Data ... 7 Migration inflows ... 7
Employment Protection Legislation ... 8
Control variables ... 10 Methods ... 11 Results ... 13 Conclusion ... 17 Discussion ... 18 Reference ... 20 Appendix ... 22
1
Introduction
“Immigrants are taking advantage of the system”. This is a belief which is catching on in Europe and is being voiced by right wing parties more and more. It is up to scholars to provide facts for the political debate on this issue. However, there is a big discussion going on in literature of political economy and migration research on the relation between welfare states, labour markets and migration (Burgoon, 2014; Freeman, 2004; Migali, 2016). An idea which is gaining traction in this debate is the welfare-magnet thesis proposed by Borjas (1999). This thesis states that higher levels of benefits will attract migrants to bigger welfare states. The evidence is divided however. There has been some support of the welfare-magnet thesis (Razin & Wahba, 2014), but Devitt (2011) points out that Nordic European states received less migrant workers than other Western European states in the decade 1997-2006, which goes against the welfare-magnet thesis. What could explain this finding is that bigger welfare states often go together with high taxes and wage floors. This makes those countries less attractive for economic immigrants (Afonso & Devitt, 2016). Devitt (2011) concludes that the variation of migration across European countries has not been adequately explained by the current literature.
In explaining the differences in labour migration flows across Europe, the literature can be categorized in two fields: the supply- and the demand side oriented research. The supply side, which is the more developed of the two, looks at motivations for immigrants to migrate. Recently research has started to examine at the demand side of labour migration flows as well. This line of research tries to explain migration flows by looking at “where migrants are ‘wanted’ by employers and states” (Devitt, 2011, p. 583). The assumption underlying this proposition is that economic migrants will tend to go where demand for them is higher. The reasoning behind this strand of research originated with Piore (1979), who claimed that migrants will accept jobs of lower social status, income and opportunities of advancement than natives. This is relevant because more liberal states will have more demand for flexible labour markets, since these states have larger low-skilled low-wage sectors (Devitt, 2011). This demand for flexible labour is better supplied by migrants than natives. Empirical test of this hypothesis have mostly been done by comparing different models of capitalist states (Devitt, 2011; Zimmermann, 2011). This paper will estimate the influence of specific labour market institutions, going beyond comparing countries, strengthening the empirical test of this hypothesis. The main indicator this paper will use is Employment
2 Protection Legislation (EPL) since this is one of the principle ways through which labour markets become more or less flexible. The research question of this paper is as follows:
Does flexibilization of the labour market lead to a higher inflow of migrants?
This paper will answer the research question by means of OLS regression where I control for certain variables expected to cause omitted variable bias. The only scholars who included EPL in their model predicting migration flows before are Cigagna and Sulis (2015). Their research was of a more explorative nature than this paper will be. This paper will focus specifically on the relation between EPL and migration inflows, whereas they were interested in the general drivers of migration. They found a strong negative effect of employment protection on migration flows. It should be mentioned that the model they used in their paper did not hold up under all robustness checks. So, the estimates provided by their model cannot be relied upon completely. Furthermore, they looked at total migration flows only, whereas this paper will differentiate between total and labour migration flows. The theoretical reasoning of this paper is mostly economic. Labour migration can be expected to be more economically motivated than asylum seeking, forced migration or family reunification (McGovern, 2007). Therefore, I will focus on labour migration. Another way this paper will try to improve upon the work by Cigagna and Sulis (2015) is by incorporating the Immigration Policies in Comparison (IMPIC) data on migration policies (Helbling et al. , 2016). Cigagna and Sulis (2015) used the immigration policy measure from the data of Ortega and Peri (2012). This measure only considers one specific type of legislation, the IMPIC dataset serves as a more valid control measure, since this indicator is made up of multiple immigration policy measures, including a measure of labour immigration policy. The data this paper will use is thus made up of OECD data on migration flows, GDP per Capita, unemployment and EPL merged with IMPIC data on migration policies. The data consists of 21 OECD countries and spans over 15 years form 1995 up until 2010.
In the following sections I will first describe the theoretical framework, after which the data section will follow, then I will discuss the empirical methods, followed by the results, finally drawing conclusions and reflecting on the research.
3
Theory
Theoretically the relationship between Employment Protection Legislation (EPL) and labour migration flows can be argued to go one of two ways: either strict (high) or lenient (low) EPL attracts higher labour migration flows. Both connections are based on the assumption that migrants are more willing to accept more flexible terms of employment, lower wages and worse employment conditions than natives (Piore, 1979). Essentially this comes down to the idea that migrants accept precarious jobs more often than natives. The first reasoning is that more lenient EPL will result in a more flexible labour market with more precarious employment which will attract migrants. The second reasoning starts from a dual labour market hypothesis and argues that strict EPL will lead to a rigid primary labour market, resulting in a high demand for flexible work in the secondary sector, which attracts migrants.
Lenient EPL
The literature explains how demand for migrants is shaped by different varieties of capitalism. I will argue for a more specific case and focus on the degree of flexibility in labour markets instead of entire state classifications. The reasoning by Devitt (2011), who builds on Piore’s (1979) work, is that more liberal states will have more demand for flexible labour, due to the fact that these states have larger low-skilled low-wage sectors. This is different for Nordic regimes which draw on high wage floors and higher skills causing the low-skilled sector to be smaller than in Liberal, Southern Statist or Conservative-Continental regimes. The demand for flexible labour is better supplied by migrants than natives, because migrants are thought to accept jobs of lower social status, income and opportunities of advancement than natives (Piore, 1979). Afonso and Devitt (2016) argue that migrants will accept precarious employment more due to their weaker social and political resources. Liberal, Southern Statist or Conservative-Continental regimes thus have a higher demand for migrants than Nordic regimes.
So far it is clear that employers and thus states from a liberal market model would have a higher demand for migrant workers than for instance Nordic states. It is not made explicit in the literature however, how this difference in demand results in differences in migration flows. The question I would like to raise is whether this demand creates its own supply? I argue that there are two ways in which the demand is expected to create the supply. The direct hypothesis assumes that migrants are utility maximizing individuals. They will pick the country with the best prospects. Due to a higher demand their employment chances
4 and wages will be higher as well. In the literature it is expected that migrants migration choices are large influenced by two main sources: their expected income in the host country and the probability of employment (Cigagna & Sulis, 2015). Migration flows will thus move to those countries where there is a higher demand for them since their employment chances there are higher.
The indirect reasoning would be that migrants are influenced in their migration decisions by immigration policies. The demand of employers for migrants could result in companies pressuring governments to adopt more lenient immigration policies. These lenient immigration policies would then act as the decision factor for migrants to come to countries where demand for them is more prominent.
This research will, instead of comparing varieties of capitalism, focus on the way in which the labour market is organized. I will use Employment Protection Legislation (EPL) as a proxy for labour market flexibility. There is a problem of reverse causality when researching the relationship between institutions and the inflows of migration. When comparing countries, it is difficult to attribute an increase in migration flows to the institutional set up of that country because it could be that the institutions are established in reaction to migration inflows. I have chosen EPL as my focus because I believe this lowers the chance of reverse causality. Since it is only one institution it is better possible to isolate the effect of this institution of migration flows. The problem of reverse causality is however still prevalent in this design as well. The assumption I make is that EPL is exogenous to migration flows. I furthermore argue that EPL is one of the main drivers creating demand for migrants in liberal states. Grubb and Wells (1993) define Employment Protection Legislation (EPL) using three elements: 1) the restriction on dismissals of workers with a regular contract; 2) the restriction on the use of temporary employment contracts; 3) the restriction on working hours. The OECD, whose data this study draws on, measures EPL using the first two elements of this definition. Lenient EPL makes it easier to lay off workers and to use temporary contracts such that there will be more flexible low paid jobs. Since migrants accept precarious jobs more than natives do, migrants demand and thus immigration flows will be higher when EPL is lenient. Therefore, I expect the effect of EPL on migration inflows along the lines of the following hypothesis:
5
Figure 1. Causal scheme H1
In Figure 1. the causal steps of my first hypothesis are laid out. In this research I will only measure the first and last step: EPL and the inflows of migration. I will thus not be able to open the black box of causality completely, but that is not the aim of this research. The aim is to make generalizable claims about the relationship between EPL and migration flows. It would be interesting for future research to see what specific mechanisms are at play.
Strict ELP
The second theory argues that strict EPL could also lead to higher migration flows. It is concerned with the tension which was also the main concern of Polanyi’s (1957) Great
Transformation, namely how to reconcile the flexibility requirements of a market economy
with the protection requirements of society. Market economies need a certain degree of flexibility, but employees want some measure of protection. Migrants can provide a solution to this tension. When EPL becomes stricter, the level of protection in society against the logic of the market improves. However, the flexibility needed for a market economy is then not reached. This can be solved by taking in migrants who fall outside of the protection but do provide the flexibility. This creates what Piore (1979) described as the dual labour market. This dual labour market consists of a primary sector, which is characterised by secure, protected jobs with high wages and is populated by natives and a secondary sector, where migrants are more likely to work which is characterised by precarious employment, low wages and bad working conditions. The reasoning is that migrants are, because of their weaker social and political resources, willing to accept these flexible terms of employment, lower wages and worse employment conditions (Piore, 1979). This can also be understood along the lines of the ‘insider, outsider’ typology from Rueda (2006) or the ‘standard, non-standard’ typology from King & Rueda (2008). High EPL only protects the insiders with standard employment within a country. High EPL also creates a bigger role for flexible labour in the secondary labour market because this type of labour is expelled from the primary
6 market via strict EPL. This causes a bigger role for the secondary labour market which consists of outsiders. Migrants are more prone to be outsiders with non-standard jobs and will be drawn towards those countries where EPL is strict.
So, when the primary labour market is protected via strict EPL there will be a bigger role to play for the secondary labour market, which is the labour market where most migrants work. If the secondary labour market becomes more important in a country, this will create a pull factor for migrants.
H2: Stricter EPL will lead to an increase in migration inflows
Figure 2. Causal scheme H2
In Figure 2. the causal links of the reasoning behind the second hypothesis are laid out. In this research I will only measure EPL and the inflows of migration. I will thus not measure the size and importance of the primary or secondary labour market. As I mentioned before the aim of this research is to make generalizable claims about the relationship between EPL and the inflows of migration and not to lay bare the specific mechanisms at play.
7
Data
For this research I constructed my own dataset using existing sources, mainly the OECD. I ended up using time-series cross-section data on 21 countries over 15 years. In the following sections I will discuss the data source and operationalization for each variable considered in my models. The different data sources do not cover the same countries and years. This means my panel data is unbalanced. In the appendix an overview of the coverage of each variable as well as the descriptive statistics are included. This paper will only include complete cases in the analysis.
Migration inflows
Migrants are often hard to define and to compare across countries because different countries use different definitions of migrants. In one country citizenship can be determined by birth right in another by virtue of the parents. In this research I focus on labour migration which is even harder to define. Earlier research on this topic has only looked at total migration flows (e.g. Cigagna & Sulis, 2015). Since my theory uses mostly economic arguments and one would expect this reasoning to hold up mainly for economic migration and less for asylum or family reunification motives of migration, it is worthwhile to differentiate between labour migration and total migration inflows. This paper will use migration inflows data from the OECD Permanent Immigrant Inflows dataset (2019). This dataset covers the regulated movements of foreigners considered to be settling in the country as perceived by the destination country. The OECD has performed a standardisation process which they argue makes the data eligible for cross-country comparisons. The OECD Permanent Immigrant Inflows data differentiates between work related inflows, free movement and several other sources of immigration. This makes it able to isolate labour immigrants. In this paper I will consider both the total inflows and the work-related inflows, both are measured by numbers of permanent inflows. The work-related inflow data covers 25 countries and spans over 22 years, from 1995 to 2016. The total inflow data covers 27 countries and spans over the same 22 years, but has more complete coverage. To account for differences in population size I measured the inflows of labour migrants as a percentage of the population.
8
Figure 1. Inflows of Permanent Migrants
Note: This figure displays the inflows of migrants as percentage of the total population
In figure 1. we can see the development of work related and total inflows of permanent immigrants. Only the countries and years which are included in the final sample are shown here. We see that the data is not balanced and that there are differences in coverage between work related inflows and total inflows as well. What catches the eye is that Ireland, New Zealand and Switzerland have relatively high inflows of total migration. We see some peaks around the time of the enlargement of the European Union. Furthermore, we see some countries where there are little changes in labour migration over time.
Employment Protection Legislation
Grubb and Wells (1993) define Employment Protection Legislation (EPL) using three elements: 1) the restriction on dismissals of workers with a regular contract; 2) the restriction on the use of temporary employment contracts; 3) the restriction on working hours. The data
9 this paper will use is from the OECD. The OECD uses the first two indicators as Grubb and Wells define. But instead of the restriction on working hours, they include a measure for the additional costs of collective dismissals. The OECD has two main indicators: the strictness of EPL on regular employment with collective dismissals taken into account and a separate indicator of the strictness of EPL on temporary employment. Of these indicators the OECD has three different versions in which the individual components are aggregated in a different way. This paper is forced to use the first version of both indicators as this version is the only one which goes back to 1995. I added the two measures, to end up with one indicator giving a complete measure of the strictness of EPL in a country. The EPL variable on temporary contracts and the variable on regular contracts consist of a six-point scale. Because I took the sum of these two variables, the variable EPL ranges from zero to twelve, where zero corresponds with the most lenient EPL and twelve with most strict EPL.
Figure 2. Development of EPL over time
Notes: The graph displays the development of EPL over time in 21 OECD countries. The index varies between 0(lenient) and 12 (strict).
10 In figure 3. we can see the variation in EPL over time. We see that there is little variation over time, in some countries there is even none. This will make it more difficult to find an effect with EPL as explanatory variable. There are also quite substantial differences in EPL with Portugal having high levels and the United States low levels of EPL.
Control variables
In my model I will include some control variables to deal with omitted variable bias. The most important control variables are the immigration policy, GDP per capita, and the unemployment rate in a country. These measures are commonly included in models predicting migration flows (Cigagna & Sulis, 2015; Ortega & Peri, 2012).
Immigration Policy
Immigration policy is expected to potentially influence both immigration flows and EPL, which is why it must be controlled for. Immigration policy is defined as the strictness of immigration laws within a destination country. The data this paper will use to measure immigration policy is from the Immigration Policies In Comparison (IMPIC) project. Helbling et al. (2016) from the IMPIC project define immigration policy as “Government’s statements of what it intends to do or not do (including laws, regulations, decisions or orders) in regards to the selection, admission, settlement and deportation of foreign citizens residing in the country” (Helbling et al., 2016, p. 82). The IMPIC data is collected by consulting with legal experts. Experts were asked to give objective information instead of a subjective interpretation of restrictiveness to gain the most objective measure of immigration policy. The final measure this paper will use was constructed by averaging 80 items on four distinctive immigration policy fields. The measure for immigration policy ranges between 0 (open) and 1 (restrictive). The variable immigration policy is available for most of the countries in my dataset (See appendix for full overview).
Gross Domestic Product per Capita
Another variable which may cause omitted variable bias is Gross Domestic Product (GDP) per capita. Some papers have included GDP growth instead of GDP per capita. This paper will use GDP per capita as measured by the OECD because I like to include the same types of variables in my econometric model. Since I do not take the changes in migration flows or EPL I will also not consider the change of GDP but take the absolute value. In the robustness
11 checks I will however specify a first-difference model in which the difference in GDP is included which can be interpreted as GDP growth.
Unemployment
The final variable which this paper will control for is the unemployment rate in a country. This paper opts to use the Harmonized Unemployment Rate (HUR) as measured by the OECD, because this indicator is better internationally comparable than estimates based on national definitions of unemployment. The HUR is defined as the people of working age who are without work, available for work and have taken specific steps to find work (OECD, 2019). This indicator is expressed in unemployed people as a percentage of the labour force. The labour force is seen by the OECD as the total number of unemployed people as well as the people in civilian employment (OECD, 2019).
Methods
To estimate the effect of EPL on migration flows I will perform Ordinary Least Squares (OLS) estimations on my time-series cross-sectional data. As mentioned in the data section, I will include immigration policy, GDP per capita and the unemployment rate as control variables in my OLS estimations. Because I am using panel data I can control for unobservable variables across countries. In my case, I can expect for instance cultural factors to be of influence on both EPL and migration flows. I can account for individual heterogeneity between countries by including country dummies in my model. This would mean estimating a fixed effects model. I performed an F-test1 to estimate the joint significance of the fixed effects intercepts. From this test I can conclude that it is better to use a fixed effects model than an OLS model. I furthermore conducted a Breusch-Pagan Lagrangian multiplier test2 which also confirmed that there are panel effects present.
When using a fixed effects model, one analyses the impact of variables within a country over time. In a fixed effects model, it is assumed that the time-invariant characteristics you control for are unique to the country and should not be correlated with other country characteristics (Allison, 2009). The error terms of the countries should not be correlated. If this is the case one could use a random effects model in which it is assumed that
1 F = 20.515 , p<.001 2 Chi2 = 254.69, p<.001
12 the variation across countries is random and uncorrelated with the predictor or independent variables (Allison, 2009). A test which could help determine whether to use a random or fixed effects model is the Hausman test. The Hausman test reported significant results3, which means that fixed effects are preferred over random effects. This in line with my expectations, since I expect cultural factors to be of influence on both EPL and migration flows. As a robustness check I will however also conduct an OLS estimation using a random effects model. I furthermore checked to see whether it was needed to include time fixed effects in my model as well. Both an F test4 and a Lagrange Multiplier test5 led me to conclude that it was not needed to perform a two-way fixed effects model by including time dummies in my model as well. I will do so as an over specification in the robustness checks. I will not include any lagged variables in my analysis as I have no theoretical grounds to do so and other scholars have not done so either (Ciganga & Sulis, 2015).
The main equation for my fixed effects model will be: [Eq1]
Inflows_workct = αc + β1EPLct + β2Immigration_Policyct + β3GDP_per_Capitact +
β4Unemploymentct + γ1C1 +…+ γnCn+ ct
Where
- αc (c=1….n) is the unknown intercept for each country (n country-specific intercepts).
- inflows_workct is the dependent variable (DV) where c = country and t = time.
- Xct represents one independent variable (IV),
- βx is the coefficient of the IV,
- Cn is country n. There are n-1 country specific dummies included in the model
- γn is the regressor for that country dummy
- ct is the error term
3 Chi2 = 66.566, p <.001
4 F = 1.371, p = .168 5 Chi2 = .086, p = .0769
13 As an additional specification of the model I will also consider total inflows as the dependent variable. Which results in the following model
[Eq2]
Inflows_totalct = αc + β1EPLct + β2Immigration_Policyct + β3GDP_per_Capitact +
β4Unemploymentct + γ1C1 +…+ γnCn+ ct
In order to report unbiased estimates, I performed a variety of tests to see which transformations were needed. I preformed both the Breusch-Pagan LM test of independence6 and the Pasaran CD7 test to check for cross-sectional dependence. The results are
contradictory but because of the nature of my dataset it is reasonable to expect issues of cross-sectional dependence. Furthermore, the Breusch-Godfrey otherwise known as the Wooldridge test for serial correlation8 pointed out that there is a problem of serial correlation in my data.
From the Augmented Dickey-Fuller test9 I learned that there is no unit root present and that
my data appears to be stationary. Lastly, the Breusch-Pagan test10 informed me that there is a problem with heteroscedasticity in my data.
In order to control for both heteroscedasticity and cross-sectional dependence I will use panel-corrected standard errors (Beck & Katz, 1995). To control for serial correlation, I will perform a Prais-Winston transformation. This should help me to achieve less biased estimates and standard errors. As a robustness check I will perform some over-specifications of my model. I will include a first differences model to control for the possible non-stationary nature of my data, a random model to control for the possibility that the error term is correlated between countries and lastly, I will include a two-way fixed effects model whereby I control for time effects. A simple OLS model will also be included.
Results
Figure 3 displays the predicted values of labour inflows based on EPL. This graph points towards a negative relation between EPL and labour inflows. It seems that the effect is strong 6 Chi2=432.58, p<.001 7 z = 1.941, p = 0.0523 8 Chi2=21.861, p <.001 9 Dickey-Fuller = -4.281, p < .001 10 BP = 852.67, p<.001
14 in some countries such as Portugal, Japan, Mexico and Ireland. In others it becomes apparent that there is no change in EPL over time which results in a singular predicted value, see for instance Italy or the United Kingdom.
Figure 3. Predicted labour inflows
Notes: the graph depicts the predicted values (Y-hat) of labour inflows based on EPL, the grey lines depict the 95% confidence interval
The results of the regressions are presented in Table 2. Model 1 indicates that EPL is negatively and significantly related to work related inflows of migrants. In years where there EPL is one step higher on the scale, there is a .078 percentage point lower inflow of work-related migrants as a percentage of the population. This effect remains negative and significant when the control variables are added in Model 2. It rises to an effect of .099. Which means that when EPL increases by one step, while the unemployment rate and immigration policy are held constant, there is almost a .1 percentage point decrease of labour inflows as a percentage of the population. Model 2 furthermore shows that GDP per Capita
15 has a significant and positive influence on labour migration inflows. The Unemployment rate is negatively and significantly related to migration inflows. There does not seem to be an effect of immigration policy on migration inflows in this specification.
Table 1. Main results
Model 1 Work Inflows Model 2 Work Inflows Model 3 Total Inflows Model 4 Total Inflows EPL -.078*** (.018) -.099*** (.022) -.084 (.062) -.069 (.595) GDP per Capita .000001** (.000) .00001*** (.000) Unemployment Rate -.013** (.006) -.040** (.014) Immigration Policy .261 (.499) -.456 (1.126) Constant .379*** (.044) .446** (.148) 1.037*** (.158) .814* (.418) Rho .470 .507 .651 .593 N × T 193 193 221 221 Adjusted R2 .517 .539 .515 .650
Notes: Unstandardized coefficients; panel-corrected standard errors in parentheses. Country
dummies where included in all models.; *p<0.1;**p<.0.05;*** p<.001
Model 3 and 4 have as dependent variable the total inflows of migrants as percentage of the population. From Table 2. there is no indication that there is a significant effect of EPL on total inflows of migrants. The direction is still the same, although not significant. An increase in EPL corresponds to a decrease in migration inflows. From Model 4 it becomes apparent that GDP per Capita and the unemployment rate have a significant impact on the total migration inflows. GDP per Capita is positively related to total migration inflows and unemployment is negatively related to total inflows. Immigration policy is not significantly related to total inflows.
In table 2. three over-specifications of the labour migration model are presented as robustness checks. Model 1. is a simple OLS model, Model 2 is a random effects model, Model 3 is a
16 two-way fixed effects model and in Model 4. the first differences of all variables are included in the one-way fixed effects model. The software did not allow me to use panel corrected standard errors in the random effects model. As a second-best option I chose to report robust standard errors, this way heteroskedasticity is still controlled for.
Table 2. Robustness Checks
Model 1 (OLS)
Work Inflows
Model 2 (RE) Work Inflows
Model 3 (2-way FE) Work Inflows
Model 4 (First Dif) Work Inflows EPL -.013** (.004) -.025* (.015) -.104*** (.022) -.111*** (.030) GDP per Capita -.000003** (.000) .0000001 (.000) .000001** (.000) .0000001 (.000) Unemployment Rate .003 (.003) -.007 (.005) -.015*** (.002) -.002 (.008) Immigration Policy .498*** (.102) .433 (.278) .420** (.160) -.297 (.520) Constant .006* (.064) .073 (.073) .450*** (.094) .015* (.008) Rho .706 .363 -.092 N × T 193 193 193 172 Adjusted R2 .182 .094 .628 .118
Notes: Unstandardized coefficients; for Models 1, 3 and 4, the panel-corrected standard errors are
in parentheses.; For Model 2 robust standard errors are shown; Year dummies were included in model 3. ; *p<0.1;**p<.0.05; *** p<.001.
From Table 4 we can make up that EPL is negatively and significantly related to labour migration flows across all robustness checks, although not at the conventional level11 in the random effects model. The effect becomes smaller in both the simple OLS and the random effects model, here the effect is -.025 whereas in the base model it was -.099. In the 2-way fixed effects model and the first-difference model the effect becomes marginally larger. In the simple OLS specification GDP per capita is a negative and significant predictor. GDP per capita seizes to be a significant predictor in the random effects model and the first-difference
17 specification. In the 2-way fixed effects model GDP per capita is significant and positively related with labour migration inflows. The unemployment rate is negatively and significantly related to labour inflows in the 2-way fixed effects model but is not significant under the OLS, random effects and first-difference specification. Immigration policy is positively and significantly related to labour migration inflows in the OLS and the 2-way fixed effects model. This could be a case of reverse causality. In countries where there is a higher inflow of migrants, governments might want to counter this by imposing strict immigration policies. However, immigration policy is not a significant predictor in the random effects not the first-difference model.
Conclusion
In this study I set out to discover whether flexibilization of the labour market leads to a higher inflow of migrants. Flexibilization of the labour market was operationalised as employment protection legislation (EPL). I hypothesized both ways arguing either lenient or strict EPL would lead to higher inflows of migration. Lenient EPL could lead to more possibilities for precarious employment in a country. Migrants take up this kind of work more often (Piore, 1979) which would mean a higher demand for migrant workers resulting in higher inflow rates. Stricter EPL could cause a rigid primary sector, which would mean more room for flexible work in the secondary sector. Since migrants are more prone to work in the secondary sector this could create a pull factor. In order to see which mechanism was at play, I fitted a fixed effects OLS model on panel data constructed using OECD and IMPIC data. I found evidence supporting my first hypothesis, namely that there is a significant and negative effect of EPL on work related migration inflows. There is however no evidence that there is an effect of EPL on total migration inflows. These findings hold up under most specifications of my model. Only in the random effects specification, does the significance of the effect of EPL on labour migration drop below the conventional levels.
These findings support the ideas of Afonso and Devitt (2016) and are partly consistent with the findings of Cigagna and Sulis (2015). My results are what you would there is indeed a higher demand for migrants in welfare states with a larger low-skilled sector caused by EPL as expected by Afonso and Devitt (2016). Cigagna and Sulis (2015) found an effect on total migration flows whereas in this paper I did not find this effect. This could be because they analysed bilateral migration flows which makes their number of observations a lot higher, so it is easier to find an effect. Or maybe the fact that I used the IMPIC immigration policy control variable (Bjerre et al., 2015) instead of the one constructed by Ortega and Peri (2012)
18 made the effect disappear. It could be that this effect of EPL on migration flows is concentrated only within labour migration as my research point out. It could also be that if my N would have been larger I would also have found an effect on total migration inflows. From my research we can however be more certain that there is an effect on labour migration inflows than there is on total migration inflows. That I did not find an effect on total inflows of migrants is also in line with my expectations. Total inflows also include asylum seekers, family reunifications and other non-economical motivations. These people are taking the labour market conditions of the destination country into account to a lesser degree. They are more motivated by the fact that their family lives in that particular country in the case of family reunification or that it is save and close to the country of origin in the case of asylum motives (McGovern, 2007). My findings are furthermore not in line with the dual labour market hypothesis as proposed by Piore (1979) it seems that the divide between insiders and outsiders (Rueda, 2006) is not strong enough that it causes a bigger inflow of migrants when EPL is stricter.
The findings of this research point out that bigger welfare states might not be the ones which attract the most migrants if these bigger welfare states also have less flexible labour markets. So the reality is more complex than portrait by the welfare state magnet hypothesis of Borjas (1999). My findings indicate that more flexible labour markets attract more labour migrants. Countries should take this into consideration when coming up with new labour market policies. When designing the labour market, a country also designs its own profile for migrants who seem to find flexible labour markets attractive. It depends on the policy aims of the country how this should be taken into account. When a country wants to lower the inflows of migrants it could for instance adopt more lenient employment protection legislation. The lesson governments can draw from this research is that labour market policy influences immigration. So, when coming up with labour market policy governments should consider the negative effects of EPL on labour migration inflows.
Discussion
Many choices were made during this study. Often, I have chosen the lesser of two evils, as such it would be good to point out some of the resulting limitations.
One limitation of this research is that flexible labour markets are solely operationalized as Employment Protection Legislation (EPL). EPL is of course only one of the institutions that make up the flexibility of the labour market. Unions and collective
19 bargaining power for instance also play a role. In future research, it would be interesting to include these additional institutions as well. Another issue with EPL is that it measures the intent but not the implementation of policy. It could for instance be that two labour markets have the same level of EPL on my scale, but the implementation is so different that the two cannot be seen as the same. What furthermore could be problematic with my operationalization of EPL is that I simply added the two measures of EPL, temporary and regular contracts and regarded them as equal. While this does not necessarily have to be the case in a country. It could be that there are 60 percent regular contracts and 40 percent temporary contracts, while I regard them as 50/50. This way I possibly do not give proper weight to temporary and regular EPL. In future research the ratio of regular versus temporary contracts could be taken into consideration.
Another limitation of this research could be the operationalization of migration flows. I took the permanent inflows of migrants as a proxy of migration inflows. Only migrants who are considered to be settling in the country from the perspective of the destination country are included. This might however not be fully accurate. Preferably, this research would have also taken the undocumented labour market into account. In order to accurately test my second hypothesis, it is needed to consider the undocumented labour market and illegal immigration. It could be that heightened EPL creates a bigger undocumented labour market in which migrants are employed. These migrants are most likely not included in the permanent migration inflows data of the OECD. This is a limitation of this research and if data became available on these migration flows this would strengthen the analysis.
The focus of this study has been on the macro relation between EPL and migration inflows. This research cannot draw any hard conclusions on which mechanism is at play however. To do that, more research is needed which should open up the black box between EPL and labour migration inflows. We now have indications that there is a negative effect, but future research should point out what the underlying mechanism looks like. Is it a cost-benefit analysis of migrants, is it the lobby of employers or is there something else at play?
20
Reference
Afonso, A., & Devitt, C. (2016). Comparative Political Economy and International Migration: EBSCOhost. Socio-Economic Review, 14(3), 591–613. Retrieved from
http://web.a.ebscohost.com/ehost/pdfviewer/pdfviewer?vid=1&sid=ad22a6a1-e48a-4a25-8340-ed8c70ae607f%40sessionmgr4006
Allison, P. D. (2009).Quantitative Applications in the Social Sciences: Fixed effects
regression models. Thousand Oaks, CA: SAGE Publications, Inc. doi:
10.4135/9781412993869
Beck, N., & Katz, J. N. (1995). What to do (and not to do) with Time-Series Cross-Section Data. The American Political Science Review, 89(3), 634–647.
Borjas, G. J. (1999). Immigration and Welfare Magnets. Journal of Labor Economics, 17(4), 607–637.
Burgoon, B. (2014). Immigration, Integration, and Support for Redistribution in Europe.
World Politics, 66(03), 365–405. https://doi.org/10.1017/s0043887114000100
Cigagna, C., & Sulis, G. (2015). On the potential interaction between labour market
institutions and immigration policies. International Journal of Manpower, 36(4), 441– 468. https://doi.org/10.1108/IJM-11-2013-0259
Devitt, C. (2011). Varieties of capitalism, variation in labour immigration. Journal of Ethnic
and Migration Studies, 37(4), 579–596. https://doi.org/10.1080/1369183X.2011.545273
Freeman, G. P. (2004). Western Democrucies ’. 38(3), 945–969.
Grubb, D., & Wells, W. (1993). Employment regulation and patterns of work in EC countries.
OECD Economic Studies, 21, 7–58.
Helbling, M., Bjerre, L., Römer, F., & Zobel, M. (2016). Measuring immigration policies: the impic database. European Consorium for Political Research, 16(1), 79–98.
https://doi.org/10.1057/s41304-016-0023-5
King, D., & Rueda, D. (2008). Cheap Labor: The New Politics of “Bread and Roses” in Industrial Democracies. Perspectives on Politics, 6(2), 279–297.
https://doi.org/10.1017/s1537592708080614
McGovern, P. (2007). Immigration, labour markets and employment relations: Problems and prospects. British Journal of Industrial Relations, 45(2), 217–235.
https://doi.org/10.1111/j.1467-8543.2007.00612.x
Migali, S. (2016). Essays on International Migration : Institutions , Skill Recognition , and
21 OECD (2019), Permanent immigrant inflows (indicator). doi: 10.1787/304546b6-en
(Accessed on 21 May 2019).
OECD (2019), GDP per capita (indicator). (Accessed on 21 May 2019)
OECD (2019), Harmonised unemployment rate (HUR) (indicator). doi: 10.1787/52570002-en (Accessed on 21 May 2019)
Ortega, F., & Peri, G. (2012). The Role of Income and Immigration Policies in Attracting International Migrants. IZA Discussion Paper No. 6655, 95616(June). Retrieved from http://ideas.repec.org/p/cda/wpaper/12-14.html
Piore, M. J. (1979). Birds of Passage: Migrant Labor and Industiral Societies. https://doi.org/10.12816/0049453
Polanyi, K. (1957). The great transformation. Boston: Beacon Press.
Razin, A., & Wahba, J. (2014). Welfare Magnet Hypothesis, Fiscal Burden, and Immigration Skill Selectivity. The Scandinavian Journal of Economics, 117(2), 369–402.
Rueda, D. (2006). Social Democracy and Active Labour-Market Policies : Insiders , Outsiders and the Politics of Employment Promotion Stable URL :
http://www.jstor.org/stable/4092255 Linked references are available on JSTOR for this article : Employment Promotion Policies : British Journal of Political Science, 36(3), 385–406.
Zimmermann, K. F. (2011). Tackling the European Migration Problem. Journal of Economic
22
Appendix
Table 3. Coverage of data sources
EPL IMPIC GDP Unemployment Work Inflows Total Inflows Australia 1985-2013 1980-2010 1980-2016 1980-2017 2003-2010 2003-2010 Austria 1985-2013 1980-2010 1980-2016 1993-2017 2003-2010 2003-2010 Belgium 1985-2013 1980-2010 1980-2016 1983-2017 2005-2010 2005-2010 Canada 1985-2013 1980-2010 1980-2016 1980-2017 1995-2010 1995-2010 Chile 2008-2013 1980-2010 1980-2016 1986-2017 N.A. N.A. Denmark 1985-2013 1980-2010 1980-2016 1983-2017 2003-2010 2003-2010 Estonia 2008-2013 1980-2010 1993-2016 1997-2017 N.A. N.A. Finland 1985-2013 1980-2010 1980-2016 1988-2017 2006-2010 1995-2010 France 1985-2013 1980-2010 1980-2016 1983-2017 2001-2010 2001-2010 Germany 1985-2013 1980-2010 1980-2016 1991-2017 1999-2010 1999-2010 Greece 1985-2013 1980-2010 1980-2016 1999-2017 N.A. N.A. Hungary 1990-2013 1980-2010 1991-2016 1996-2017 N.A. N.A. Ireland 1985-2013 1980-2010 1980-2016 1983-2017 2002-2010 2002-2010 Israel 2008-2013 1980-2010 1980-2016 1995-2017 N.A. N.A. Italy 1985-2013 1980-2010 1980-2016 1983-2017 2003-2010 2003-2010 Japan 1985-2013 1980-2010 1980-2016 1980-2017 1995-2010 1995-2010 Latvia 2012-2013 N.A. 1995-2016 1999-2017 N.A. N.A. Luxembourg 2008-2013 1980-2010 1980-2016 1983-2017 N.A. N.A. Mexico 1990-2013 1980-2010 1980-2016 1987-2017 2010 1997-2010 Netherlands 1985-2013 1980-2010 1980-2016 1983-2017 1995-2010 1995-2010 New Zealand 1990-2013 1980-2010 1980-2016 1980-2017 2001-2010 2001-2010 Norway 1985-2013 1980-2010 1980-2016 1989-2017 2003-2010 2003-2010 Poland 1990-2013 1980-2010 1980-2016 1997-2017 N.A. N.A. Portugal 1985-2013 1980-2010 1980-2016 1983-2017 1999-2010 1999-2010 Slovak Republic 1993-2013 1980-2010 1992-2016 1998-2017 N.A. N.A. Slovenia 2008-2014 N.A. 1995-2016 1996-2017 N.A. N.A. Spain 1985-2013 1980-2010 1980-2016 1987-2017 2007-2010 2007-2010 Sweden 1985-2013 1980-2010 1980-2016 1983-2017 2000-2010 2000-2010 Switzerland 1985-2013 1980-2010 1980-2016 2010-2017 2004-2010 2002-2010 United Kingdom 1985-2014 1980-2010 1980-2016 1983-2017 2003-2010 2003-2010 United States 1985-2013 1980-2010 1980-2016 1980-2017 1996-2010 1996-2010
23
Table 4. Descriptive Statistics
N Mean Standard Deviation Min Max Inflows_work 193 .092 .106 .002 .664 Inflows_total 221 .542 .407 .004 2.030 EPL 221 3.578 7.789 .507 7.396 GDPcap 221 33159.250 9470.214 8817.659 61759.990 Unemployment 221 6.426 2.657 2.505 19.875 Immigration Policy 221 .405 .071 .303 .704
