Canada: Impact of immigrants on native wages
University of Amsterdam Bachelor’s thesis: Economics
Tatyana van der Horst Student number: 10353852
Supervisor: Zheng Jindi February 2015 Abstract
Conventional economic theory states that an immigrant related increase in labor supply depresses native wages in a labor market. However, past literature has claimed otherwise and this paper aims to add to the immigrant wage relationship discussion by examining the impact of immigrants on native wages in the context of the Canadian labor market between the years 2006 and 2011. It uses the skill cell method first proposed by Borjas (2002) but extends it to include both occupation and education data sets as suggested by Barrett et al (2009). Considering both education and occupation skill cell groups would produce a more robust and comprehensive analysis since it is possible immigrants compete with natives not only within an education framework but also from an occupational perspective. A regression analysis was performed and statistically insignificant results were obtained. Overall, the findings suggest immigrants to Canada have negligible wage impact on natives.
Section 1: Introduction
Global migration is becoming increasingly prevalent, with many countries experiencing a surge in immigrant inflow. According to the United Nations, there were approximately 231 million international immigrants in 2013, a 50% increase from just 154 million in 1990. Immigrants, who migrate in search of better work prospects, or migrant workers, do so due to urban-‐rural wage differentials. This phenomenon can be explained by the Harris-‐Todaro (1970) model, which assumes workers look at expected income differentials when making migration decisions. Comparing in the same currency, this implies that the wage received in the migrant’s home country is lower than the wage received in the foreign country. As such, foreign labor movement generally flows from less developed rural regions to industrialized countries such as the United States, United Kingdom and Australia where the expected real wage tend to be higher (Harris & Todaro, 1970).
In many nations with a large immigrant share, immigrants are treated with resentment as they are perceived to bring about various issues such as heightened job competition, an increase in crime rates and the loss of national identity. Conventional economic theory suggests that an increase in labor supply induced by an inflow of immigrants would depress earnings of natives in the short run. However, it would also raise the return to capital, stimulate investment, and improve the economy and in the long run, cause wages to return to the initial wage rate. Early empirical evidence using various models under different settings, has for the larger part, rejected this notion (see Goldin, 1994, Lalonde and Topel, 1991, Altonji and Card, 1991, Pischke and Velling, 1994). Friedberg and Hunt (1995) conducted a comprehensive survey on available literature and concluded that the effect of immigrants on a host country’s wages is small and in most cases, often negligible.
However, a study undertaken by George Borjas in 2002 found that a 10 percent immigrant induced labor supply decreased the wages in the United States by approximately 3 to 4 percent. The study of Borjas sparked a renewed interest in the immigration wage relationship, with many of the subsequent studies contending that the impact of immigrants on domestic wages is
negligible (see Card, 2001, Bonin, 2005, Ottaviano and Peri, 2008, Aleksynska and Tritah, 2009). As such, there seems to be no real consensus within this field of labor economics.
In general, empirical research regarding the immigrant wage relationship can be categorized into two main methods, namely, the local labor market approach and the national level approach. The former approach takes on the spatial correlation model, which divides the labor market into localities and exploits the geographic variation in migrant density across these localities. A regression of domestic wages against immigrant density across localities is then conducted, in order to estimate the correlation between earnings of natives and immigrant worker concentration. Notable literature within this framework includes Card (1990), Lalonde and Topel (2009), Altonji and Card (2011). However, Borjas (2002) argued that the spatial correlation method grapples with endogeneity issues. He suggested the national labor market approach, which considers the labor market as a whole and classifies workers into various skill cell groups according to education and experience. Influenced by his work, Barrett et al (2009) went on to augment the skill cell model of Borjas, including not only educational skill cells, but also analyzed workers from an occupational group perspective.
This study uses the modified variant proposed by Barrett et al (2009) and aims to add to the field of knowledge in two ways. First, the skill cell approach introduced by Borjas (2002) is relatively recent and only a handful of studies using this methodology exist (see Bonin, 2005, Ottaviano and Peri, 2007 and Barrett et al, 2009). Moreover, available literature using the same model have come to different conclusions when employed in different labor markets – Borjas (2002) applied the model to the U.S. market and found that a 10 percent size increase in a particular skill cell caused by immigration inflow decreased the wages of natives in that same skill cell by about 3 to 4 percent. Correspondingly, Bonin (2005) applied the model to the German labor market but found an insignificant wage impact on natives instead. As such, this paper adopts the skill cell model in an attempt to add to the current stock of studies pertaining to the wage immigration relationship debate.
Second, empirical research pertaining to immigrant impact on Canadian domestic wage is sparse and no consensus has been reached even though both methodologies have been employed. Aydemir and Borjas (2006) approached the study from the national level perspective, using the skill cell model (only education-‐age) to analyze impact of immigrants on wages in Canada. They arrived at the same conclusion as Borjas (2002) and found that a 10 percent immigrant induced increase in labor supply depresses the wages in Canada by about 3 to 4 percent. Their findings deviate from the results obtained by Jiong (2010), who adopted the local labor market approach. Using the spatial correlation model, the author divided the Canadian labor force according to both a worker’s geographic area and skill type, using data obtained from the 1991, 1996 and 2001 Canadian Census. Jiong found no significant adverse wage impacts in both the OLS and IV regressions conducted in her study.
This paper applies the modified skill cell model proposed by Barrett et al (2009), which goes one step further from the study conducted by Aydemir and Borjas (2006), and looks at both educational level and occupational group of workers in the labor market. Considering both education and occupation skill cell groups would produce a more comprehensive analysis since it is possible that immigrants compete with natives not only within an education framework but also from an occupational perspective. By doing so, I hope to augment the available literature by adding to the dearth of empirical evidence relating to the skill cell model and research within Canada.
The rest of this paper is as follows. Section 2 examines past literature, in accordance to the two major approaches (i.e. local and national) that have been undertaken. Section 3 provides an overview of the history of immigration in Canada. Section 4 focuses on the skill cell model and specification of data obtained from the 2006 and 2011 Canadian Census files. Section 5, which forms the core of this paper, presents the empirical analysis, including both a descriptive and econometric review. Lastly, section 6 concludes with final remarks and limitations of the study.
Section 2: Literature review
Using conventional microeconomic theory, the impact of migrant workers on a host country’s wages can be explained by a simply labor supply and demand framework (Borjas, 2002). As illustrated in Figure 1 below, an influx of immigrants would cause an outward shift in the labor supply curve (!! !" !!),
and the equilibrium real wage of domestic workers would decrease (!! !" !!). Moreover, according to Borjas (2002), so long as the labor supply curve is upward loping, the immigrant induced increase in labor supply must also reduce the supply of native workers (from !! !" !!).
Figure 1: Immigrant induced labor supply
In addition to that, Solow growth model can also be used to augment the labor supply and demand framework in order to analyze the influx of immigrants on native wages (Dadush, 2014). According to Solow growth theory, assuming native workers and migrants are perfect substitutes, an inflow of immigrants would have the same effect as a one-‐time increase in the domestic labor force. In the short run, there are two adverse outcomes: first, wages of the natives are depressed since capital-‐labor ratio decreases, and second, some employed natives would choose not to work at the lower wage rate. However, the Solow growth model states that an increase in the labor force attributed from the influx of immigrants would raise the return to capital, stimulate investment, and improve the economy. In the long run, the economy returns back to its initial
capital labor ratio and initial wage rate. Therefore, the detrimental impacts of immigrants on native wages are only temporary.
Due to growing beliefs that immigrants have adverse impacts on a host country’s wages, academics have sought various ways to dispute this notion, namely, the spatial correlation model and the skill cell method. Earliest empirical studies first started with the spatial correlation model, also known as the area approach. This technique segments an entire labor market into various localities in order to exploit the relation between immigrant density across various geographic areas and domestic wages. The skill cell model, which is more recent and rebukes the spatial correlation model, examines immigrants and natives within the same skill cell, but on a national level. In short, the spatial correlation model espouses a local labor market perspective, whereas the skill cell method adopts a national labor market approach.
2.1 Local level – Spatial correlation
The spatial correlation method, or area approach, approaches the immigrant – domestic wage from a local labor market perspective. The underlying idea is that immigrants tend to cluster within specific geographic areas, leading to a variation in migrant stock density across different areas. For example, in 1990, 32.5 percent of the migrant population lived in only three localities: New York, Los Angeles and Miami (Borjas, 2002). Spatial correlation studies exploit this geographic feature by performing a regression analysis of migrant stock density against natives’ wages in order to obtain the “spatial correlation”, which is used to estimate the impact of immigrants on a host country’s wages.
However, spatial correlation analysis tends to grapple with two issues. First, the distribution of migrant workers may not be arbitrary, as assumed in the spatial correlation model. Instead, it is possible that migrants respond to wage levels, namely, a region that experiences a demand shock and has higher wages will attract more migrant workers (Friedberg and Hunt, 1995). If such is the case, then flourishing localities will have higher densities of migrant workers, and a spurious positive correlation between immigrants and wages will be
present. There are various avenues to circumvent this issue, one of which includes analyzing countries for which regions within a country are allocated immigrant quotas. For instance, Glitz (2006) capitalizes on Germany’s immigration policy, which allocates immigrants to particular labor markets and prohibits them from moving to areas with more attractive wage benefits. Another method would be to use instrument variables, such as in the case of Altonji and Card (1991), Hunt (1992) and Jaeger (2007). Second, internal migration may occur (Borjas, 2002) as the labor market adjusts after an influx of immigrants. To elaborate, natives may respond to the influx of immigrants by moving to other areas, and this relocation might affect both the immigrants’ destination as well as the native’s new destination, both of which are not taken into consideration in the model mentioned above.
2.2 National level – Skill cell approach
The skill cell model, proposed by Borjas (2002), is a more recent method and examines workers on a national level instead of analyzing the immigrant impact across local labor markets. The typical skill cell method segregates a domestic labor force into various skill cells – including both educational attainment and work experience – and estimates the effects of an inflow of immigrants who belong to the same skill cell. Work experience is computed using the age of a worker as a proxy since work experience data is unavailable. To illustrate this empirical approach, consider an example of a worker who belongs to the skill cell defined as age group “15 – 24” and an educational attainment of “no certificate, diploma or degree”. The model then estimates the change in the wage of the domestic worker in that particular skill cell across two or more time periods, according to the change of immigrant supply within the exact same skill cell (Age 15 – 24 and has no certificate, diploma or degree) has occurred.
It is an alternative from the spatial correlation method, which assumes a labor market only has two types of workers: low-‐skilled and high-‐skilled. Moreover, the spatial correlation model also assumes low-‐skilled immigrants and native workers are substitutionary in nature, which makes an unlikely case
since wage ramifications caused by changes in immigrant supply are unevenly spread across the labor market. That is to say, the entry of immigrants would affect some natives more than others i.e. an influx of migrant uneducated youths would have a greater impact on native uneducated youths (low-‐skilled) than elderly uneducated native workers (also low-‐skilled). As a result, assuming a labor market only has two types of workers (high and low skilled) is questionable. In contrast, the skill cell model, in which workers are segmented into various skill cells according to experience (using age as a proxy) and education level, provides a more exogenous variation of immigration on wage impact.
Contrary to previous human capital literature, which uses only education, the skill cell model proposed by Borjas (2002) exploits both educational attainment and work experience of workers. The reason for this is that if one were to divide workers based on only educational attainment, there would be insufficient variation to estimate the impact. For example, the Canadian decennial Census used in this study consists of only five schooling groups, namely (1) No certificate, diploma or degree (2) High school diploma or equivalent (3) Apprenticeship or trades certificate or diploma (4) College, CEGEP or other non-‐university certificate or diploma, and (5) University certificate, diploma or degree. As such, each Canadian decennial census produces only five wage and employment observations. Furthermore, human capital theory has highlighted that workers do not only attain skills from schools, but also before and after one enters the labor market (Becker, 1975). Therefore, skill cells are defined in terms of both educational level and work experience, in which age is used as a proxy for work experience since data is unavailable.
In his paper, Borjas (2002) used the U.S. Decennial Census as well as the Annual Demographic Supplement of the Current Population Surveys (CPS) to form a data set of the U.S. labor force for the years 1960-‐1990 and 1998-‐2001. His findings greatly deviate from previous studies, which contend that immigrants do not have significant adverse impacts on the earnings and employment of native workers. In numerical terms, he found that a 10 percent increase in the labor force caused by an influx of immigrants in a particular skill cell, depresses the earnings of native workers in the same skill by 3 to 4 percent.
Since its induction, several authors have attempted to investigate the impact of immigrants on domestic wages by applying the skill cell model to various labor markets, namely Bonin (2005) on the German labor market, Ottaviano and Peri (2007) in the U.S., as well as, Barrett et al (2009) who employed the model to the Irish economy. Not only do their findings add to the scarcity of available literature, each paper also demonstrates the validity and reliability of the model, particularly with regards to the applicability of the model across varying labor markets.
Influenced by Borjas (2002), Holger Bonin (2005) was one of the first authors to adopt the skill cell model, in which he did so for the context of the German labor market. His paper uses German register data obtained from the Regional File of the IAB Employment Subsample (IABS-‐R) and spans across the period from 1975 to 1997. In his results, a wage elasticity of -‐0.102 was found, which implies that a 10 percent immigrant related increase in a particular education – experience group decreases the wage of native workers in the same skill cell by about 1 percent. Moreover, German local youths and older workers with a work experience of more than 25 years experience larger adverse impacts than the rest of the labor force. Bonin concludes that immigrants have a negligible impact on earnings of German native workers and he attributed this result to the possibility that migrant and native workers, even when belonging in the same skill group, work in different segments of the labor market. As such, foreigners and German native workers are complementary instead of substitutes. This conclusion deviates from the one Borjas (2002) made, in which he asserted a 10 percent increase in immigrant reduces native wages by about 3 to 4 percent in the U.S. labor market. A critical insight made by Bonin (2005) in his research is that the skill cell approach does not seem to be able to achieve stable and reliable estimates when applied to a different context, such as in the case of Germany compared to the U.S labor market. As such, more studies need to be done in order to access the reliability and validity of this model.
Another commonly cited paper that has approached the immigration and domestic wage debate from the national perspective is the one written by Ottaviano and Peri (2010). The authors argue that a partial effect model, such as the skill cell model, does not capture the full impact a total effect model can
produce. In the case of immigration and domestic wages, a partial effect would be the direct impact of foreign workers on native wages within a particular skill cell, given a fixed supply in the other skill cells. It differs from the total effect model, which not only takes the direct impact into consideration, but also accounts for the indirect impact of immigrants on domestic wages in all other skill groups. As such, they propose a nested Constant Elasticity of Substitution (CES) model, in which data for the model is constructed from the 1960, 1970, 1980, 1990 and 2000 U.S. Census, 2006 American Community Survey (ACS) as well as data from the Current Population Survey (CPS). Moreover, physical capital adjustments are also taken into consideration when estimating the elasticity of substitution between foreign and native workers. They find that natives enjoy a greater benefit from an inflow of immigrants since wages increase in both the short and long run. Ottaviano and Peri (2010) made a few conclusions analogous to the ones made by Bonin (2005): First, wage impacts are unevenly distributed, such that less educated natives experience a small decrease in earnings. Furthermore, they also found that natives and foreign workers, even within the same skill group, are imperfect substitutes. Lastly, similar to Bonin (2005), they conclude that the skill cell method arrives at different outcomes according to how one classifies the various skill groups therefore results vary according to the context and setting of the model.
The last piece of literature to conclude this review is the study published by Barrett et al (2009), which uses the skill cell model to estimate the effect of foreign labor on earnings of Irish natives. Instead of using education-‐experience skill cells such as suggested by Borjas (2002), Barrett et al (2009) construct two separate skill cell groups: education-‐experience and occupation-‐experience, in which experience is computed using age as a proxy. They contend that this is a more appropriate measure since immigrants to Ireland have a higher educational level, yet they only manage to attain a similar occupational level as native Irish, and this tendency does not diminish even as the length of residency increases. Such a finding also resonates with the foreign worker employment situation in the UK (Drinkwater and Clark, 2008). Therefore, it is plausible foreign labor and natives tend to compete within occupation skill cells instead of educational skill groups, prompting the authors to divide the Irish labor market
into both educational attainment and occupational level in order to estimate the wage impact.
In the analysis of the education-‐experience dataset, they managed to replicate the findings of Borjas (2002) and found an adverse impact on the earnings of native Irish. However, when examining occupation-‐experience skill cells, they found a positive wage impact instead. They explain that this is possible since the skill cell method does not estimate wage impact of the average worker, but the average wage impact within a specific skill cell instead. Furthermore, the contradictory results (i.e. negative impact in education cells but positive in occupation cells) can be attributed to the prevalent substitutionary nature between native Irish and migrant workers in the education cells, whereas this substitution effect is lacking in the occupation cells. A crucial insight from this paper is that it is perhaps insufficient to approach the skill cell method from only an education perspective, such as in the study of Borjas, since natives and foreign workers can also compete based on an occupational angle.
To summarize, there are two favored methods to approach the immigrant domestic wage debate: First, the spatial correlation analysis, which divides the labor market into localities to exploit the geographic variation of immigrant density; and second, the national approach, in which the labor market is looked at as an entirety, and workers are placed into various skill cells according to their educational attainment and/or occupational level. As both models have produced inconsistent results, there exists no consensus on the wage impact induced by a surge in immigrant inflow.
Section 3: Immigrants in Canada
In 2013, approximately 20% of Canada’s national population was composed of immigrants, making it an interesting and worthwhile case for studying immigrant induced labor market outcomes. Prior to the 1960s, immigration policy in Canada was national-‐origin preference based. This meant certain nationalities, such as Indians and Pakistanis, faced an entry limit whilst preferential treatment was given to others, mainly the British, French and Americans. In 1967, the Points System was introduced, in which potential
immigrants were given points based on certain criteria; two examples include having arranged employment and having a proper education. Unlike previous immigration policies, the Points System did not impose quotas based on nationalities, which led to an influx of immigrants from Africa, Asia and the Middle East, most of which settling into urban cities such as Montreal, Toronto and Vancouver. The rapid and unexpected inflow of non-‐white immigrants in these cities was greeted with hostility and racism, prompting the initiation of the Immigration Act in 1976. This new act gave authority to provincial governments to set their own immigration laws as well as prohibited entry to individuals who risked becoming a burden on social services. Moreover, four classes were initiated and immigrants had to enter under one of the four classes: 1) Refugees, 2) Families, 3) Assisted relatives and 4) Independent immigrants, which included individuals who entered for work or schooling purposes. In 2001, the Immigration and Refugee Protection Act was passed, and this act boasted tightened immigration regulations in response to the 11 September 2001 attack in the United States.
Figure 2: Immigrants into Canada (1860 – 2009)
Source: http://www.cic.gc.ca/English/resources/statistics/facts2009/permanent/index.asp
A distinct feature of the Canadian immigration policy is its “tap off, tap on” feature, in which the inflow of immigrants depended on the current macroeconomic condition of Canada, specifically, more inflow during economic
booms and less during recessions. For example, after the World War 2, Canada experienced a shortage in labor, rendering the authorities to allow for more immigrants in order to support the labor crunch. As a result of this policy, Canada is characterized by fluctuating levels of immigrant stock, with figures ranging from just 9000 immigrants in 1940 to more than 250,000 in 2009. This “tap off, tap on” immigration policy was forgone in 1990, and since then, Canada has seen a relatively high and stable level of immigrant inflow (refer to Figure 2 above).
Section 4: Model and data specification
This study uses the skill cell model proposed in the paper of Barrett et al (2009), which analyzes both education-‐experience as well as occupation-‐ experience skill cells, and extends it to the Canadian labor market. It is a modified version of the framework suggested by Borjas (2002), which considers only the education-‐experience dataset. In this framework, workers are segmented into various skill cell groups, classifying them according to both educational attainment and occupational group. A regression is then performed to estimate the impact of immigrants on the earnings of native workers:
!
!"#= α
!
!"#+ βW
!+ γX
!+ δ
!
!+
!
!"#Y
ijt represents the average hourly earnings of Canadians who have particulareducation level
i
, work experience levelj
, and are observed in periodt
;F
(thekey variable of interest) is the proportion of immigrants in the skill cell (ijt) with education
i
, work experiencej
and periodt
;W
i is the fixed effect vector relatedto educational level;
X
j is the fixed effect vector related to experience level; Zt isthe fixed effect vector related to the time period and
ε
ijt is the error term.The immigrant share,
F
ijt is the fraction of immigrants in a particular skill!
!"# =!
!"#!
!"# +!
!"#Where
I
ijt is the number of immigrants in skill cell ijt andM
ijt is the number ofnatives (defined as individuals who are born in Canada and having a Canadian citizenship) in the same skill cell group ijt.
As mentioned above, this study uses the modified variant of the skill cell model proposed by Barrett et al (2009) which constructs two sets of skill cell groups and includes both educational attainment and occupational level of a worker. The variables Educational Attainment and Occupational Level are controlled for using dummy variables.
For educational attainment, workers are classified into five levels of schooling: 1) No certificate, diploma or degree, 2) High school diploma or equivalent, 3) Apprenticeship or trades certificate or diploma, 4) College, CEGEP or other non-‐university certificate or diploma and 5) University certificate, diploma or degree. Labor market experience is determined by the number of years one has worked, however, since work experience data is unavailable, following Borjas (2002) work, age has been used as a proxy instead. Four experience groups are constructed based on the following age intervals: 1) 15-‐ 24, 2) 25-‐44, 3) 45-‐54 and 4) 55-‐64.
There are ten skill cell groups for occupation of workers: 1) Management occupations, 2) Business, finance and administrative occupations, 3) Natural and applied sciences and related occupations, 4) Health occupations, 5) Occupations in social science, education, government service and religion, 6) Occupations in art, culture, recreation and sport, 7) Sales and service occupations, 8) Trades, transport and equipment operators and related occupations, 9) Occupations unique to primary industry and lastly, 10) Occupations unique to processing, manufacturing and utilities. Similar to the education skill cell groups, age is used as a proxy for experience and three experience groups are constructed: 1) 15-‐24, 2) 25-‐54 and 3) 55 and older.
Data for education and occupational level of workers is extracted from the 2006 and 2011 Canadian Census of Population as well as the National Household
Survey (NHS). Both sets of data are produced by Statistics Canada, an agency under the Government of Canada. They are designed to provide statistical information about the economic and social structure of Canada – including a wide assortment of topics such as marital status, education, labor and types of dwelling – in order to help analyze and develop public policies and programs so as to improve the welfare of Canadian citizens.
Section 5: Empirical analysis
Figure 3 below shows the immigrant share in both 2006 and 2011 according to the various education-‐experience skill cell groups mentioned above. Comparing the immigrant share in each skill cell group across the years, it is evident that almost all groups experienced only minor changes ranging from -‐2% to 2% between the years 2006 and 2011. Not parallel to this common observation are two cell groups – First, Apprenticeship or trades certificate or diploma and age 55-‐64, and second, College, CEGEP or other non-‐university certificate or diploma and age 55-‐64. Both see a larger than average decrease in the proportion of immigrants, which are -‐5.15% and -‐3.07% respectively. Furthermore, the cell group University certificate, diploma or degree and age 45-‐54 is the only one in this dataset to see a large positive increase of 4.1% in the share of immigrants.
Across the twenty skill cell groups, two features can be observed. First, as illustrated in Figure 3 below, within each educational attainment set, the age group 15-‐24 has the smallest proportion of immigrants, regardless of education level. In contrast, the age interval 55-‐64 has the largest share of immigrants in each skill cell group. Second, the highest possible educational attainment, specifically the University certificate, diploma or degree skill cell set has the highest fraction of non-‐natives. This is in line with Canada’s immigration policy, which is altered periodically to suit the needs of its labor force (King, 2009). Since 1967, immigration in Canada has been based on a points system, in which each potential individual is examined based on the following six factors: age, education, work experience, arranged employment, fluency in Canada’s two official languages and one’s adaptability to society. The points system was designed to attract individuals with a high educational attainment and with almost half of the possible full points being allocated to education and proficiency of
language, it is expected that the inflow of immigrants for this particular group is larger than the others.
Figure 3: Immigrant share in education-‐experience skill cell
Figure 4 below shows the change in the proportion of immigrants according to the three age intervals and ten occupational groups between the years 2006 and 2011. An interesting feature is that the immigrant share tends to be lowest in each 15-‐24 age interval and largest in the older workers age group (55 and older). This phenomenon was also observed in the educational skill cell data set above. Furthermore, it is also apparent that the proportion of immigrants in each 55 and older age interval across the 10 occupation groups has experienced a sizeable decrease between the years 2006 and 2011, particularly -‐8.02% in Art, culture, recreation and sport as well as -‐7.51% in processing, manufacturing and utilities. In similar fashion, this phenomenon can be explained by the points system, in which a total of 21 points is allocated to work experience, increasing in the number of years one has worked. Therefore, since age is used as a proxy for work experience under the assumption that an older worker would have accumulated more years of work experience than a fresh graduate out of university. As such, older workers who belong to the age
0% 5% 10% 15% 20% 25% 30% 35% 40% 15 -‐2 4 25 -‐4 4 45 -‐5 4 55 -‐6 4 15 -‐2 4 25 -‐4 4 45 -‐5 4 55 -‐6 4 15 -‐2 4 25 -‐4 4 45 -‐5 4 55 -‐6 4 15 -‐2 4 25 -‐4 4 45 -‐5 4 55 -‐6 4 15 -‐2 4 25 -‐4 4 45 -‐5 4 55 -‐6 4 No certilicate, diploma or degree High school diploma or equivalent Apprenticeship or trades certilicate or diploma College, CEGEP or other non-‐
university certilicate or diploma University certilicate, diploma or degree 2006 2011
group 55 and older are able to attain a higher level of points than younger individuals in the age interval 15-‐24. The points system does not award points based on an individual’s occupation, hence it is difficult to account for the variation of immigrants across the different occupational groups. From Figure 4 below, it is evident that occupations related to natural and applied sciences as well as processing, manufacturing and utilities have a larger portion of immigrants compared to other occupational roles. One possible explanation for this observation is that Canada tends to tailor their immigration policy to meet the labor demands of an industry. According to a 2008 immigration paper published by Statistics Canada1, industries related to applied science skills such
as engineering, health and computer sciences face a labor shortage, thus benefitting immigrants with degrees related in these fields (Galarneau and Morissette, 2004).
Figure 4: Immigrant share in occupation-‐experience skill cell
1 http://www.statcan.gc.ca/pub/75-‐001-‐x/2008112/pdf/10766-‐eng.pdf 2 Canada’s national average weekly hours: http://www4.hrsdc.gc.ca/.3ndic.1t.4r@-‐
0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 15 -‐2 4 25 -‐5 4 55 and ol der 15 -‐2 4 25 -‐5 4 55 and ol der 15 -‐2 4 25 -‐5 4 55 and ol der 15 -‐2 4 25 -‐5 4 55 and ol der 15 -‐2 4 25 -‐5 4 55 and ol der 1 Management
occupations 2 Business, linance and administrative occupations 3 Natural and applied sciences and related occupations 4 Health
occupations 5 Occupations in social science, education, government service
and religion
Table 1 below summarizes the hourly wages of workers in each skill cell group. As hourly wage data is not available for workers according to educational attainment, two assumptions are made in order to derive the hourly wages from annual average salary. First, I assume that all workers in the Canadian labor force, regardless of age or educational level, work the same national average number of weekly hours worked – 36.9 hours in 2006 and 36.4 hours in 20112.
The second assumption I make is that there are 52 weeks in a year. Correspondingly, a total of 1918.8 hours are worked in 2006 and 1892.8 hours in 2011. In addition to the assumptions made, data for earnings in 2006 is inflated to 2011 prices using Statistics Canada’s Consumer Price Index3. According to
Table 1 below, all skill cells experience an increase in hourly wages, with a $1 to $2 increase in the majority of skill cells. One interesting feature is that the hourly wages in the skill cells of the highest educational attainment group (University certificate, diploma or degree) experienced an approximate 50% increase in hourly wages across the four age intervals between the years 2006 and 2011.
2 Canada’s national average weekly hours: http://www4.hrsdc.gc.ca/.3ndic.1t.4r@-‐
eng.jsp?iid=19
3 Canada’s Consumer Price Index Data: http://www.statcan.gc.ca/tables-‐tableaux/sum-‐
som/l01/cst01/econ46a-‐eng.htm 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 15 -‐2 4 25 -‐5 4 55 and ol der 15 -‐2 4 25 -‐5 4 55 and ol der 15 -‐2 4 25 -‐5 4 55 and ol der 15 -‐2 4 25 -‐5 4 55 and ol der 15 -‐2 4 25 -‐5 4 55 and ol der 6 Occupations in art, culture, recreation and sport
7 Sales and service
occupations 8 Trades, transport and equipment operators and related occupations
9 Occupations unique to primary
industry 10 Occupations unique to processing, manufacturing and utilities
Table 1: Hourly wage in education-‐experience skill cell
Education level Age Hourly wage 2006 2011 1. No certificate, diploma or degree 15-‐24 3.80 4.18
25-‐44 24.24 25.64 45-‐54 13.67 14.24 55-‐64 11.92 12.62
2. High school diploma or equivalent 15-‐24 6.23 6.25 25-‐44 30.05 31.87 45-‐54 17.98 18.69 55-‐64 15.60 16.50
3. Apprenticeship or trades certificate or
diploma 15-‐24 25-‐44 32.99 9.74 10.84 36.88
45-‐54 19.22 20.24 55-‐64 17.17 18.58
4. College, CEGEP or other non-‐
university certificate or diploma 15-‐24 25-‐44 35.55 8.33 37.98 8.61
45-‐54 21.29 22.61 55-‐64 18.51 20.11
5. University certificate, diploma or degree 15-‐24 8.01 16.12 25-‐44 44.65 87.03 45-‐54 31.04 58.42 55-‐64 28.26 54.29
For occupational skill cell groups, no assumptions were made to obtain hourly wage of workers since hourly wage data was available from Statistics Canada. Similar to education, to account for inflation, 2006 hourly wages are adjusted to 2011 prices using the Consumer Price Index4 released by Statistics
Canada. Almost all cells experienced only a minimal increase of $1 to $2 between the years 2006 and 2011, with the exception of three groups (Occupation group 6 & age interval 55 and older, Occupation group 9 & age interval 25-‐54 as well as 55 and older). Certain occupation groups (Group 1, 3, 4, 5, 6) see the age interval 55 and older attaining higher wages than those in the same occupation aged 25-‐ 54, which can also be explained by labor theory, in which experience is an attribute of a worker’s value, and thus his wages.
4 Canada’s Consumer Price Index Data: http://www.statcan.gc.ca/tables-‐tableaux/sum-‐
Table 2: Hourly wage in occupation-‐experience skill cell
Occupational level Age 2006 2011 1. Management occupations 15-‐24 14.98 16.27 25-‐54 33.92 35.25 55 and older 37.96 39.20
2. Business, finance and administrative occupations
15-‐24 12.79 13.92 25-‐54 21.43 22.76 55 and older 21.37 22.32
3. Natural and applied sciences and related occupations 15-‐24 17.42 19.26 25-‐54 30.95 32.64 55 and older 35.39 36.87 4. Health occupations 15-‐24 17.52 18.62 25-‐54 25.72 27.41 55 and older 26.41 28.49
5. Occupations in social science, education, government service and religion
15-‐24 15.39 16.33 25-‐54 28.22 30.10 55 and older 32.21 34.17
6. Occupations in art, culture, recreation and sport
15-‐24 13.13 14.53 25-‐54 24.42 25.92 55 and older 26.69 29.74
7. Sales and service occupations 15-‐24 9.98 11.09 25-‐54 16.61 17.89 55 and older 15.18 16.58
8. Trades, transport and equipment operators and related occupations
15-‐24 15.46 16.85 25-‐54 22.36 23.93 55 and older 21.69 22.98
9. Occupations unique to primary industry
15-‐24 13.63 14.70 25-‐54 19.56 22.64 55 and older 16.94 20.52
10. Occupations unique to processing,
manufacturing and utilities 15-‐24 25-‐54 13.78 19.40 14.71 20.11
55 and older 19.13 19.27
The following part consists of the econometric analysis of the skill cell model, in which an OLS regression method was applied to find the impact of immigrants on native wages in Canada. The regression uses a sizable variation of 40 skill cells in the education dataset and 60 skill cells in the occupational dataset to enable reliable results to be produced. Table 3 below reports the regression outcome obtained, where the dependent variable is the log of real
hourly wage in a specific skill cell group. Application of OLS regression finds a coefficient of 1.090 for immigrant share in education skill cell groups, which is also statistically insignificant. This implies that a change in the proportion of immigrants has no effect on native wages in a specific education skill cell group. The results obtain show negative coefficients in a decreasing manner for educational level, and this is anticipated since the reference dummy was taken to be University certificate, diploma or degree holders, which is the highest possible educational attainment of a worker. This can be explained by basic labor theory, in which education is considered to be an attribute to the value of a worker and hence his wage. Naturally, a worker with a university degree will command a higher compensation than one with only a high school certificate.
Table 3: Impact of immigrants on native wages using education skill cells
Variable Coefficient Standard Error P-‐level Immigrant share 1.090 0.666 0.112 Age (Ref: 16-‐24) Age 25-‐44 0.597 0.061 0.000 Age 45-‐54 0.361 0.067 0.000 Age 55-‐64 0.280 0.085 0.002
Educational level (Ref: University certificate, diploma or degree)
No certificate, diploma or degree -‐0.333 0.087 0.000 High school diploma or equivalent -‐0.206 0.085 0.021 Apprenticeship or trades certificate or diploma -‐0.070 0.112 0.540 College, CEGEP or other non-‐university
certificate or diploma
-‐0.101 0.092 0.278 Year (Ref: 2006) Year 2011 0.080 0.024 0.002 Constant 0.830 0.164 0.000 Observations 40 R-‐squared 0.951 F-‐statistic 64.608
Table 4 presents the results for the occupational skill cell groups. Similar to the regression results ran on the educational skill cell dataset, the coefficient for immigrant share (-‐0.126) is negative and statistically insignificant. This