Econometric Analysis of the Impact of Immigration on the German Economy

Abstract

For long, the issue of immigrants coming into high-income countries has kept governments and natives concerned and researchers interested. This paper contributes to the existing vast discussion on the impact of immigration on the receiving country’s macroeconomic indicators by examining the effect that immigrants have had on Germany’s household consumption expenditure and unemployment rate, while, also analyzing if the 2015 European refugee crisis has had any significant effect yet. The study estimates two models over the period 1983-2017 using Auto-regressive Distributed Lag (ARDL) Bounds testing approach. The econometric analysis of this paper revealed that immigrants can be associated with negative effects on household consumption expenditure, while, they remain insignificant to any change in the unemployment rate in both, short and long run. Further, the 2015 refugee crisis appeared to have a positive impact on household consumption expenditure but had no significance on unemployment. The findings were consistent with the existing papers and theoretical background for the most part.

1. Introduction

In the last few decades, the world has witnessed various changes in the socio-economic and demographic trends. The most significant of such changes is the sharp rise in foreign-born population, particularly in high-income countries. The constantly increasing proportion of foreign-born population in many high-income countries is an issue that concerns not only the government but also the natives of the receiving countries. UN (2017) estimates that between 1980 and 2017, the foreign-born population in the world increased by 2.5 times to almost 258 million, with more than three-fifth of these living in high-income countries (UN, 2017). The share of immigrants in the population of OECD countries has risen from 9% in 1995 to almost 13% in 2006 (Longhi, et al., 2010, p.820).

The transnational migration from low and middle-income countries to high-income countries has many motives, such as, family reunification, desire for a better standard of living or even escaping a war struck country (Longhi, et al., 2010). The consequences for the receiving country range from demographic and social to economic and political changes. The native population is particularly unhappy about the rise in immigrants and have thus given rise to political debates and demand for policy reforms. In the last decade, only the political parties which supported right-wing ideologies and promised to restrict immigration in the country attracted the most support from the natives. This rise of right-wing populist political parties has not only promoted the rhetoric of anti-immigration but, also as a consequence, has normalized the idea of closed borders among the natives. Increasing proportion of natives now believe that restricting foreign-born immigration will drastically reduce crime rate and spur economic growth (Kubrin, 2013).

The rhetoric of anti-immigration arises primarily from two factors; economic and cultural (Tabellini, 2018). Natives believe that more immigrants lead to an increased competition in the labor market and therefore reduce employment opportunities for them, and at the same time, natives also consider immigrants a threat to their cultural and social cohesion (Tabellini, 2018). The consequences of this rising concern of the natives can be seen through the support and popularity for the ‘Border Wall’ in the US, the ‘Brexit’ movement in the UK and the rise of AfD in Germany[1].

Amongst the OECD countries, Germany has arguably had one of the most open borders for refugees and asylum seekers over the years[2]. It receives the second highest number of asylum applications, after the US, with almost 198,000 applicants in 2017 (OECD, 2018a). According to Eurostat (2019), Germany has the highest numbers of foreign-born[3] people among the EU-28 states with almost 14 million people in 2018 (Eurostat, 2019). Given the large number of immigrants it is not surprising that Germany is also a home for many anti-immigration movements, most notably, Patriotic Europeans Against the Islamisation of the Occident (PEGIDA)[4]. According to a survey conducted by Pew Research Center (2018), more than 58% of the people surveyed in Germany believed that the country should let fewer or no immigrants in, in contrast, only 29% of the Americans surveyed think the same.

For decades, the issue of immigration has been at the center of all political and scholarly debates in Germany. Many right-wing groups have taken to the streets demanding stricter policies on immigration and are simultaneously known to have committed hate crimes against refugees and people of different races (Benček & Strasheim, 2016). Their fear of immigrants range from increased job losses, due to availability of cheaper labor, to increased crime and cultural disharmony. This paper attempts to examine this very fear of the natives and analyse if immigration has any effect on Germany’s household consumption expenditure and unemployment rate, at the same time, the paper will also analyse if the 2015 Refugee Crisis has had any effect on the dependent variables so far. The importance of this study, both, to the government and the natives, cannot be unstated, with the former requiring it to formulate better employment and immigrant policies, and latter, to understand the reality better. This paper argues that there might be short term negative effects of immigration, but in the long-term, immigration will have no effect on the two dependent variables.

The paper is divided in 9 sections. The second section reviews the existing literature along with the theoretical issues involved in this study. Section 3 and 4 provide historical and theoretical background, with data sources, methodology and model specification in the fifth section. Sixth section analyses the data used and provides the result of all misspecification tests performed. Model estimation done using the Auto-regressive Distributed Lag (ARDL) Bounds testing approach is given in section 7. Finally, the paper ends with results discussion and conclusion in section 8 and 9 respectively.

2. Literature Review 

Until the late 1990s there were very little empirical studies on the impact of immigration, Bauer & Zimmermann (1995, p.95) went on to call it a ‘black hole’ in economics. They believed that this was primarily because of the lack of desired time series data and the limited interest of researchers on the topic. However, since then, the topic has gained a lot of traction. Presently, there are hundreds of studies examining the impact of immigration on the receiving country’s labor market and they vary across methodology and the kind of data used.

The scholarly and political discussions on the impact of immigration are based on three major aspects; immigrants’ performance in the receiving country’s economy, their effect on the natives’ unemployment, and finally, the right immigration policy that benefits receiving country the most (Borjas, 1994)[5]. Borjas (1994) believes that the perception that all immigrants have an adverse effect on the native labor market is incorrectly formed. Immigrants have varied levels of productivity and skills, a highly productive immigrant, one who can easily adapt to local conditions, can have a significant positive effect for the natives, while, a low skilled immigrant who can’t adapt will only increase the costs of income maintenance programs (Borjas, 1994). Altonji and Card (1989) also found that the effect of immigration depends largely on their skill composition, and that there exists a moderate correlation between the less skilled natives and the immigrants, in particular, an increase in immigration increases the fraction of immigrants in low-skilled labor force (Altonji & Card, 1989). 

The topic is primarily examined using cross-country data (panel data) or time series data, with most of the papers using the former (Shan, et al., 1999). Damette & Fromentin (2013) used panel cointegration approach to study the effect of migration on the labor markets of OECD countries and found no evidence of a significant effect in the long run. On the other hand, Angrist and Kugler (2003) also studied a panel of 18 OECD countries over 1983 to 1999 and found a small but negative effect of immigrants on natives’ employment. Use of time series analysis for this topic is fairly recent, however, more and more studyies are switching to time series analysis. Shan, et al. (1999) used Granger causality testing procedure to study Australia and New Zealand, and concluded that there is no causal relationship. Ortega and Peri (2009) derived a pseudo-gravity model and found no correlation between immigration and per capita income in 22 OECD countries. In a similar study on OECD countries, Boubtane, et al. (2013) found that immigrants can actually be linked with higher GDP and economic prosperity in the host country.

Weiske (2019) used data from the Current Population Survey (CPS) to analyse the macroeconomic effects of immigration in the US and found that immigration is insignificant to the US business cycles, in addition, he pointed out various problems associated with studying the effects of immigration; (1) immigration has to be treated as an endogenous variable since the decision to migrate depends on the conditions of both, the home and the destination country[6] leading to biasedness in estimation; (2) natives can switch to other sectors or move to other regions which might not be affected by immigrants; (3) change in wages in the short run is highly dependent on the rate of capital adjustment following an immigration shock.

Fertig & Schurer (2007) discuss the importance of heterogeneity, assimilation of immigrants and the quantitative role played by attrition bias while exploring the impact of immigration in Germany. They found a fair degree of heterogeneity, with regard to assimilation, in the probabilities for both annual income and unemployment. According to Buhr & Weber (1996), much of the increase in government spending on social security can be attributed to migrants.  This gives rise to another major concern of the natives, i.e., high utilization of social security benefits by the migrants, which increases the burden on the state and therefore on the taxes paid by the natives (Bauer, et al., 2005). Bonin (2005) uses skill group approach to analyse the impact of immigration on the labor market for native Germans, and found that a 10% increase in the immigrants’ proportion in total workforce would normally reduce the natives’ wage by less than 1%. In addition, he points out that the effect slightly worsens as a result of low-skilled and old aged immigration, which interestingly is in sharp contrast to the results of similar studies on the US labor market (Bonin, 2005). Natives fear that more immigrants would translate to job losses and lower wages, however, Schmidt (1993) showed that the immigrants in Germany are far less likely to get a white-collar job or become civil servants as compared to the native Germans, which suggests that natives do not run a risk of losing high-skilled or government jobs as a result of an immigration shock. Bentolila, et al. (2008) analyzed how immigration affects the New Keynesian Phillips Curve and found that had it not been for a huge influx of immigrants in Spain, the inflation would have risen by 2.5% after the fall in unemployment.

OECD (2018a) estimates that the employment rate of immigrants in OECD countries was 67% in 2017 and the gap between the unemployment rate of the natives and immigrants was about 3%. Winkelmann & Zimmermann (1993) made one of first attempts to examine the impact on employment and unemployment in Germany. They studied a period of 10 years and based their analysis on 1,830 males, which included 586 foreigners. They found that unemployment would increase substantially as a result of immigration, however, Mühleisena & Zimmermann (1994) found no evidence to support this. According to Bauer, et al. (2005) there isn’t much evidence to support the notion that immigration has any adverse effect on wages and unemployment in Germany, which is consistent with studies on other European countries or the US. They however conclude that the effect largely depends on the competitiveness of the labor markets, as, a less competitive market will have higher risks for natives as compared to a highly competitive market (Bauer, et al., 2005). 

The existing literature is quite vast and comprehensive, it covers almost all aspects of a labor market and how immigration can have an impact. Most of the papers have a similar conclusion, i.e., low or insignificant effect of an immigration shock on the local labor market, with a few exceptions. A major takeaway is that, the effect on labor market depends a lot on factors like, age, sex, skill and education level of the immigrants, and also on the macroeconomic and the labor market conditions of the receiving country.

Although the impact of immigrants on an economy is a well-researched topic, this study attempts to contribute to the existing literature by analyzing the short and long run effects of immigration on unemployment rate and household consumption expenditure for the period 1983 to 2017, thereby, also including the effect of the 2015 Refugee Crisis, which the existing papers lack. Further, by also examining the impact on household consumption expenditure, the study is including an important aspect of the economy, which is often left out, the households. In addition to this, the study will also attempt to provide a more comprehensive analysis by focusing entirely on the German economy, as most existing papers instead aim to provide a comparison of the impact on various OECD economies than a single country analysis.

3. Historical Background

3.1 Immigration in West and East Germany

The Berlin Wall, manifested in 1961, stood as a divide between East and West Germany[7] for over 28 years. It was not only a physical disturbance but also a cultural and most importantly an ideological divide (Leventhal, 2010). Since its inception, Germany faced a many-sided migration and labor market experience with low supply of skilled labor and yet high and rising unemployment (Bauer, et al., 2005). Until the 1973, it recruited low skilled labor from southern and eastern Europe[8], most of whom were later given German citizenships (Fertig & Schurer, 2007). By the 1980s, the political changes allowed for a fresh period of immigration in Germany with a high inflow of asylum seekers along with east-west migration (Bauer, et al., 2005). Foreign-born people’s share had risen to almost 8% of the total population and an equivalent share of the labor force, by 1989 (Bauer, et al., 2005).

3.2 Immigration in Reunified Germany

Figure 1 shows that the number of foreign-born persons coming into Germany each year from 1980 to 2018 has an upward trend with two major events of sharp increase, 1991 and 2014. After the fall of Berlin Wall and the Reunification of Germany, the country witnessed a spike in internal migration and an inflow of almost 400,000 refugees and asylum seekers (Bauer, et al., 2005). This was followed by a period of restrictive immigration policies put in place by the Federal Government, and as a consequence, there was a dramatic decline in immigration of foreign-born persons each year up to 2005. After years of political impasse, a new immigration law was passed in 2004 allowing highly skilled workers and entrepreneurs who promised to generate local employment, to immigrate (Bauer, et al., 2005). Germany’s stance on immigrants and in particular asylum seekers relaxed when Angela Merkel assumed office in 2005 and the country started taking more asylum applications. In the first few months of 2015, close to 500,000 refugees, most of them fleeing war-struck Syria, arrived on European shores, giving rise to the European refugee crisis of 2015 (Holmes & Castañeda, 2016). Germany with its relatively relaxed immigrant policy let in most of the refugees which spiked the number of foreign-born persons coming into Germany between 2015 and 2017 (see figure 1).

4. Theoretical Background

Multiple theories can help explain the relationship between foreign-born migrants and receiving country’s macroeconomic indicators. According to the Neoclassical approach, the primary reason for migration between two regions is the difference in their demand and supply of labor (see Ravenstein, 1889). Regions where the labor demand exceeds supply tend to have a higher equilibrium wage and regions with excess supply tend to have lower, as a result, labor migrates from a region with excess supply to a region with excess demand. In this process, the wage in the previously excess demand region fall until the difference in wages is equal to the cost of migration (Hicks, 1932, as cited in Bauer and Zimmermann, 1995). Harris & Todaro (1970, as cited in Bauer and Zimmermann, 1995) extended this model by dropping the assumption of full employment, now, a migrant base his/her decision on the probability of employment and the expected earnings.  

Figure 3: Push Migration with Fixed Wages
Source: Bauer & Zimmermann (1995, p.100)

Figure 2: Push and Pull Migration
Source: Bauer & Zimmermann (1995, p.100)

The theory of ‘push and pull’ factors of migration provide simple dynamics of the impact of immigration on the receiving country. From the figure 2, if the demand for labor increases in the host country, the price for labor starts rising, in an effort to control inflation the host country might choose to allow immigration, as a result, the supply curve shifts downward, increasing output and decreasing price. This shift of equilibrium point from A to B is known as pull migration (Bauer & Zimmermann, 1995). On the other hand, if the supply curve shifts downward, as a result of an immigration shock, with no change in demand, the equilibrium will move from A to C. This movement is known as push migration[9] (Bauer & Zimmermann, 1995). Conversely, if the real wages are fixed at W, EF will represent the level of unemployment in the host country (see figure 3). An immigration shock in this case will result in greater unemployment and prices[10] with constant output (Bauer & Zimmermann, 1995). Hence, a push migration, in such a scenario, would cause stagflation (Bauer & Zimmermann, 1995). Germany experiences both, push and pull factors of migration, and as Heckel, et al. (2008) point out, there is some degree of wage stickiness present across Europe, as a result, the second scenario, given in figure 3, can be experienced to some extent.  

High-income countries, like Germany, tend to have a high demand for low-skilled labor with insufficient supply, and although, it might be fair to assume that immigration will lower the average wage, Longhi, et al. (2010) believe that, the lower wages will not only help firms expand production but also allow them adopt labor intensive techniques which weren’t possible earlier, this effect can be seen in figure 2, which shows an increase in output and fall in wages. The increased employment, production and consumer demand associated with bigger population will yield larger tax revenue and as a consequence, greater public services (Longhi, et al., 2010). Cheaper immigrant labor can also translate to higher productivity levels (or GDP per hour worked), which will induce more investment until the rate of return to capital is restored to national average in the long run (Longhi, et al., 2010). Nickell (2010) believes that natives shouldn’t be affected by immigrants in the long-run as the economy will be bigger and can hence accommodate more workers. 

To predict the effect that immigrants have on unemployment in the receiving country is rather tricky. As Jean & Jiménez (2011) explain, immigrants increase both labor supply and demand, but, not necessarily simultaneously. The occurrence of these increases depends on multiple factors, which include, rate of immigrant assimilation and labor market policies put in place by the host country’s government (Jean & Jiménez, 2011). For instance, in the presence of high level of dualism in labor market protectionism, the employment gap between foreign-born and natives reduces, and the wage gap rises, at the same, high tax wedges cause employment rate of immigrants to fall (Causa & Jean, 2007, as cited in Jean & Jiménez, 2011). In addition, the impact of immigration on unemployment will be lower if a change in employment policy increases how wages respond to unemployment (Jean & Jiménez, 2011). According to Nickell (2010), in the short run, unemployment rises after an immigration shock and it stays high if the employment policies are restrictive, however, in the long run, immigrations helps to increase the flexibility of local labor market and reduces the equilibrium rate of unemployment (Nickell, 2010). In short, an economy should witness negative effect on average wages and a positive effect on employment, output and productivity, however, these effects are expected to die out in the long run.

5.    Data and Methodology

5.1 Variables and Data Sources

This study uses an original dataset compiled using 35 annual observations of time series data of 7 variables over the period 1983-2017. Data for 6 of the 7 variables is obtained from the enormous database of Organization for Economic Co-operation and Development (OECD) while the data for immigration in Germany is obtained from the database of Federal Statistical Office, Germany called Genesis Destatis[11]. OECD data is known for its robust and routinely reported statistics which is widely accepted and used by governments, financial markets and businesses, suggesting its fair degree of reliability and comparability (Reinhardt, et al., 2002). Since each country defines its immigrants differently (see Bauer & Zimmermann, 1995), Genesis Destatis was used to obtain the time series on immigrants. Zühlke, et al. (2004) presents evidence for the robustness of data obtained from Federal Statistical Office and its various research centers.

For empirical analysis, this study uses log differenced form of household consumption expenditure (HHEXP) to represent Germany’s annual percentage change in consumption. Unemployment rate (UNEMPR) and employment rate (EMPR)[12] are included to account for the absorption role that a labor market plays in an economy (Hausmann & Rigobon, 2003), while, a log differenced form of unemployment rate (UNEMPC) is included to examine the rate of change of unemployment. As suggested by Bentolila, et al. (2008), the study includes inflation rate (INFL) to analyse if immigration has any effect on the Phillips curve. GDP per hour worked is used as a proxy for productivity level (PROD). CRISIS, a dummy variable, that takes the value 0 for years 1983-2014 and the value 1 for 2015-2017, is included to account for the European refugee crisis of 2015[13]. Finally, IMMGTN is the number of foreign-born persons coming into Germany each year from Asia, America, Africa and Australia and Oceania and also people who are ‘Stateless, unknown, uncertain, not specified’ (Federal Statistical Office, 2019). Migrants from European countries were not included because of the freedom of movement and work within the EU countries, as inferred from Nickell (2010).

5.2 Problems

Number of immigrants is a challenging statistic for a number of reasons; poor monitoring of registration and deregistration of new and outgoing migrants, large number of undocumented and unrecorded migrants. This leads to different sources telling largely different figures of immigration (Bauer & Zimmermann, 1995). Further, the fall of Berlin Wall and the German Reunification in 1990 not only limits the availability and consistency of the macroeconomic statistics of Germany but also raise questions on its integrity. Before 1990, West and East Germany were two separate countries with two contrasting political systems, as a result, a simple aggregation of the data doesn’t give the true picture of the German economy, henceforth, the results should be interpreted with caution.

5.3 Methodology

Although most papers employ panel data analysis, this study uses time series analysis and employs the Auto-regressive Distributed Lag (ARDL) Bounds testing approach, developed by Pesaran and Shin (1999) and Pesaran et al. (2001). The general ARDL (p,q) model is given by,  

Yt=γ0i+∑i=1p δiYt-i+∑i=0q βi'Xt-i+εit

Here, the dependent variable (Y) is represented as a function of its on lagged variables, current and lagged values of explanatory variables (X). This procedure is adopted for four broad reasons; (1) its simplicity, when compared to multivariate cointegration techniques like Johansen cointegration, Bounds test allows cointegration to be estimated using OLS; (2) Bounds test doesn’t require variables to be tested for stationarity beforehand, unlike VAR technique, variables can be used in an ARDL model irrespective of their order of integration, hence, regressors can be stationary I(0), integrated of order one I(1) or even mutually cointegrated, however, it is necessary to ensure that none of the variables are integrated of order two I(2); (3) this technique gives more robust results in finite or small samples; (4) ARDL uses one of the information criterion, e.g. Akaike info criterion (AIC), to choose the appropriate lag length for each regressor, as opposed to VAR, where all regressors have the same lags (Frimpong & Oteng-Abayie, 2006).

Figure 4: ARDL Bounds testing procedure

The ARDL Bounds testing procedure, developed by Pesaran, et al. (2001), is used to estimate the short and long-run impact of immigration on unemployment rate and household consumption expenditure for the period 1983 to 2017. Broadly, the procedure has four steps (see the figure 4 above), first, perform unit root tests to ensure that none of variables used are integrated of order two I(2), second, specify the correct ARDL model, tested for all the misspecification tests, third, perform a bounds test for cointegration and if cointegration among the regressors is found, the final step is to use error correction and estimating a long run model, else, estimating only the short-run model. This study uses EViews (version 10) to perform all data analysis and takes help from Startz (2015).

5.4 Model Specification 

5.4.1  Model 1

HHEXPt=κ0+∑i=1h κiHHEXPt-i+∑i=0j πiIMMGTNt-i+∑i=0f ψiPRODt-i+∑i=0n ϕiUNEMPRt-i+CRISIS+ε2t

The purpose of this model is to examine the relationship between the annual change in household consumption expenditure (HHEXP), number of immigrants (IMMGTN), GDP per hour worked (PROD) and the rate of unemployment (UNEMPR). CRISIS is again the time dummy.

5.4.2  Model 2

UNEMPCt=β0+∑i=1p βiUNEMPCt-i+∑i=0q γiIMMGTNt-i+∑i=0r αiINFLt-i+∑i=0v ρiEMPRt-i+CRISIS+ε1t

This model is based on the traditional Phillips curve according to which unemployment and inflation have an inverse relationship (Bentolila, et al., 2008). Similar to Bentolila, et al. (2008) this study will also analyse how immigration affects the Phillips curve of Germany by estimating the same model without IMMGTN and CRISIS[14]. The primary model examines the correlation between the percentage change in unemployment rate (UNEMPC), number of immigrants (IMMGTN), inflation (INFL), employment rate (EMPR) and the time dummy (CRISIS).

6. Data Analysis

6.1  Preliminary Analysis

 The a-priori analysis of the data involved in this study uses descriptive statistics given in table 1. The table shows a summary of mean, median, maximum, minimum, standard deviation, skewness, kurtosis, Jarque-Bera and probability. From this table, it can be seen that all thevariableshave a standard deviation lower than 1 (except INFL), and the maximum and minimum values are near the mean, except for IMMGTN, which has a few outlier values from the 2015 refugee crisis (see series and box plots in appendix 2-figure2.c). Further, probability suggests that all the variables, except INFL[15], are normally distributed. Hence, it can be concluded that there is a fair degree of stability in the variables used.

6.2  Stationarity Tests

It is important to ensure that all the variables used in an ARDL model are either stationary I(0), integrated of order one I(1) or mutually cointegrated. If any variable used is integrated of order two I(2), the F-stat for bounds test, as provided by Pesaran et al. (2001), cannot be considered valid and the regression might give spurious results (Frimpong & Oteng-Abayie, 2006). Hence, the following stationarity tests are performed.

6.2.1  Graphical Analysis

A quick inspection of all the series plots (see appendix 2) show that EMPR and PROD have a clear upward trend, and the trend line in figure 1 suggests that IMMGTN also has a small but upward trend. While the plot of INFL seems to suggest no trend, the plots for UNEMPC and UNEMPR show a downward trend. Hence, for all the variables (except INFL), it is difficult to conclude stationarity, as a consequence, unit root tests are performed.

 6.2.2  Unit Root Test

This study employs both, the conventional Augmented Dickey-Fuller test (ADF) and the Kwiatkowski-Phillips-Schmidt-Shin test (KPSS) to test for the unit root processes in the series. ADF test is performed using two lag selection criteria; Akaike Info Criterion and Schwarz Info Criterion with maximum lags being 9. The bandwidth used for the KPSS test is given by Newey-West Bandwidth.

The null and alternative hypothesis for the Augmented Dickey-Fuller test is given by,

H0: the process has a unit root

H1: the process doesn’t have a unit root

And the null and alternative hypothesis for Kwiatkowski-Phillips-Schmidt-Shin test is,

H0: process is trend stationary

H1: it is a unit root process

Because of the different hypothesis, rejecting the null in ADF test and failing to reject the null in KPSS test, suggests stationarity. Table 2 below provi