The Assimilation of Hong Kong Immigrants in Canada

 
 
The Assimilation of Hong Kong Immigrants in
                                                Canada
                                           Jackie M.L. Chan∗



                                                   Abstract

         A flood of emigration from Hong Kong to Canada was observed prior to the handover from
      Britain back to China in 1997. I pool Canadian Census data from 1991 to 2006 to study the labor
      market assimilation of Hong Kong immigrants. The findings suggest that these immigrants faced a
      significant negative wage gap upon entry in comparison with Canadian-born individuals, and this
      was larger for cohorts arriving closer to 1997, suggesting a decline in cohort quality. The results
      also indicate that earnings convergence with the Canadian-born was slow for the majority of these
      immigrants.


      JEL Classification codes: N32, N35, J61
      Keywords: migration, assimilation, Hong Kong, Canada




   ∗ Address for correspondence: Department of Economics, Stanford University, 579 Serra Mall, Stanford, CA 94305,

jmlchan@stanford.edu. I thank Professor Ran Abramitzky, Art Tosborvorn, and an anonymous referee for their helpful
comments.


                                                        1
1.     Introduction
The labor market performance and assimilation of immigrants in their host nation over time is of interest
to scholars and policy-makers alike. Many factors may affect these outcomes, such as the immigrants’
country of origin. Throughout history, migrants have moved from one country to another in large
numbers for various economic or political reasons. The consequences of an influx of immigrants for the
host nation may be large as it often changes the composition of the labor market among other things.
This paper studies the economic assimilation of immigrants in Canada, with an emphasis on the localized
influx of migrants from Hong Kong. Using data from the 1991, 1996, 2001, and 2006 Canadian Censuses,
I examine the assimilation profiles of immigrants upon entering Canada.
     After more than a hundred years as a British crown colony, the city of Hong Kong reverted from
British to Chinese sovereignty in 1997. A flood of emigration was observed in the years leading up to this
historic event: total emigration from Hong Kong jumped from around 20,000 a year in 1987 to 60,000
by 1990 and stayed at this level for half a decade (Li, 2005).1 These people left for countries such as
Canada, Australia, the United States, and the United Kingdom. Canada was by far the most popular
destination: from 1990 to 1997, approximately 62% of migrants from Hong Kong landed in Canada.2
Figure 1 below shows the emigration trend from Hong Kong to Canada that emerged in the mid-80’s,
its acceleration in the early 90’s, and its subsequent collapse.3




                            Figure 1: Immigrants to Canada (Figure 1 from Li (2005))


     During the 80’s and 90’s, the majority of the migrants from Hong Kong to Canada arrived as
economic-class immigrants. In some years, the proportion of economic-class immigrants from Hong
Kong was over 80% (Li, 2005). The background of these migrants might lead one to expect that they
would have been relatively successful after moving overseas. Indeed, there is the view that they have
revitalized cities like Vancouver with their capital and investment, and perhaps more notably, driving up
the real-estate prices in the neighborhoods there (e.g., see Moretti (2012, p.241) and Li (1998, p.148)).
However, these stories may not necessarily be representative of the average immigrant, and the savings
and wealth they brought over may not be a good indicator of their success in the labor market. Indeed,
the evidence and analysis shown here present a direct challenge to this view and leads one to re-examine
the degree of success experienced by Hong Kong immigrants.
     New immigrants may be confronted with an earnings disadvantage in the labor market for various
reasons. For example, their credentials may not be recognized in the host nation or they may simply
be discriminated against. Typically, their wages grow over time as they become more assimilated and
their performance in the labor market improves. To quantify labor market assimilation, this paper


                                                       2
employs an empirical strategy based on the series of works by Borjas (1985, 1987, 1994). By regressing
log wage on years since migration and controls for socioeconomic characteristics, an assimilation profile
for immigrants can be constructed. Their performance is compared with individuals born in Canada,
which I refer to as “natives” for the purposes of this paper. An immigrant assimilation profile consists
of two components: the entry wage gap and the rate of earnings convergence; the latter term will be
used interchangeably with “rate of assimilation” throughout the paper. Thus, with these two numbers,
the central question to this study can be answered: How long did it take for an immigrant to achieve
earnings convergence with a native?
   The empirical analysis shows that immigrants experienced a large initial wage gap, and labor market
assimilation with the native-born, if any, was modest and slow. This is consistent with the findings
of previous studies on Canadian immigrants, which I review below. The estimates here suggest that a
Hong Kong migrant arriving in 1980 faced an earnings gap greater than 50% upon entry in comparison
with natives. The average Hong Kong migrant in fact fared better than both American and Chinese
migrants. (Unless otherwise stated, I use the term Chinese to refer to people born in China, and not in
the racial sense.) There is also a robust finding of declining cohort quality, a result not unique to this
study. Moreover, I find the annual rate of earnings convergence to natives for Hong Kong immigrants
to be around 3 to 4%. Therefore, assimilation for the majority of this group of migrants required a
minimum of 15 years, and oftentimes much longer.
   To examine more closely the issue of cohort quality, and specifically the evidence pertaining to the
case of Hong Kong immigrants in Canada, I construct four immigrant cohort groups separated by three
key historical events: the signing of the Sino-British Joint Declaration in 1984, the Tiananmen Square
incident in 1989, and the handover in 1997. Judging the quality of immigrants based on their entry
earnings disadvantage, the pre-1990 Hong Kong immigrant cohorts were of higher quality than post-
1990 cohorts. In fact, the entry gap that the 1990-97 Cohort faced is almost 30 percentage points larger
than that of the 1985-89 Cohort. The 1989 Tiananmen Square incident and impending handover in 1997
appears not only to have led to a surge of emigration from Hong Kong, but along with it, a precipitous
drop in the quality (or negative selection) of leavers. However, although the initial earnings gap of the
more recent cohorts is greater than that of their predecessors, their wage also grows faster. Cohort
quality in this dimension is thus higher for the more recent cohorts. Lastly, the key finding that there is
positive but slow labor market assimilation is confirmed with other labor market outcomes such as labor
force participation and number of weeks worked.
   Understanding immigrant assimilation has been an integral part of the migration literature, beginning
with seminal work by Chiswick (1978) and Borjas (1985) among others. The analysis may examine
immigrants in a host nation collectively, or the focus may be on a particular group of migrants from
a single country of origin. For example, Chiquiar and Hanson (2005) study the case of Mexicans in
the United States. Quantifying the degree or pace of assimilation has been challenging because of data
limitations. Abramitzky et al. (2014) address this issue by studying 19th to early 20th-Century European
migrants to the United States. Using modern econometric methods, they first compare the regression
results of single versus pooled cross-sections, and find a large decline in cohort quality over time. My
approach is thus very similar, and I shed light on the extent to which deterioration in the quality of
cohorts occurs in other contexts. This shared finding may be indicative of common forces at work.
For instance, it is consistent with the idea that newer immigrants benefit from the establishment of
communities and networks by their predecessors, and therefore their lower quality is less of a detriment
in their assimilation experience. Abramitzky et al. (2014) also analyze panel data, following the same
migrant over time, and conclude pooled cross-section results are biased from the negative selection of


                                                    3
return migrants.
   In the context of the history of international migration, this is an intriguing case study for several
reasons. First, there was an observable non-random selection of emigrants based on their economic
characteristics. By the 80’s and 90’s, the middle class of Hong Kong was larger and more affluent
under the prospering economy (Li, 2005). The typical Hong Kong adult was well educated and coming
from a British colony, was at least familiar with an official language of the host nation before landing.
Even then, the results here suggest that there was a considerable amount of heterogeneity in immigrant
quality across the years associated with main political events. Unlike wars or other events that may
begin or end abruptly, the handover in 1997 was something anticipated for more than a decade. Amidst
the atmosphere of fear and uncertainty, even following the Tiananmen Square incident in 1989, many
residents of Hong Kong still had time to consider the decision to move and respond to a political event
known well in advance. The selection of migrants appears to have been increasingly negative as the
return of Hong Kong to China approached.
   In addition, this relatively painless episode of cluster migration from Hong Kong to Canada is some-
what of an anomaly in the history of mass migration. For instance, it contrasts greatly with the localized
influx of migrants observed from East to West Germany following the fall of the Berlin Wall in 1989.
Other examples include the intake of refugees by the United States from Vietnam after the end of Viet-
nam War in 1975, or the emigration of Algerians to France at the outbreak of the Algerian War in
1954 (Castles and Miller, 1998). This experience is also different from other episodes of Asian migration
overseas, such as the movement of mostly unskilled labor from China and Japan to the United States in
mid to late 1800’s (Daniels, 1988). On the other hand, the evidence presented here, especially that which
challenges the “success view” of the Hong Kong migrants, is very relevant for current immigration policy
and migrants from mainland China, as the booming economy there has also generated an increasingly
prosperous middle class seeking residency in foreign nations.
   My paper also extends more specifically the line of research on the assimilation of immigrants in
Canada, reinforcing earlier findings with more recent data. This literature has typically found small or
even negative rates of assimilation for immigrants in the Canadian labor market, as well as substantial
deterioration in cohort quality over time (e.g., Baker and Benjamin (1994), Bloom et al. (1995)). Al-
though there is large variation across cohorts entering in different periods, the average number of years
to wage equality with natives is in the mid-20’s. More recently, Campolieti et al. (2013) study Canadian
Censuses from 1971 to 2006. Their results indicate that the more recent cohorts tend to have a faster rate
of assimilation along with a larger earnings penalty upon arrival, therefore supporting the hypothesis
of Grant (1999) that there may be a negative correlation between the initial wage gap and subsequent
earnings growth. My study confirms this result, even though the functional forms used in the regression
analysis are more flexible than those in Campolieti et al. (2013).
   Lastly, this study explicitly examines the heterogeneity of outcomes for immigrants from different
areas or countries of origin. Borjas (1987) studied this aspect of assimilation in detail and showed that
country of birth is an important variable in determining the rate of assimilation for immigrants in the
United States. In the Canadian context, evidence from Bloom et al. (1995) suggests that male immigrants
from Europe and the United States assimilated much more quickly with natives than those from Asia,
Africa, or Latin American. Using panel data, Li (2003b) finds that the collective group of immigrants
from China, Hong Kong, Taiwan, and Macau required the longest time to reach earnings parity with
the average Canadian out of all the broad areas of origin defined. The comparison between Hong Kong
and Chinese immigrants here demonstrates that there may be considerable heterogeneity even within
particular geographic regions of the world.


                                                    4
Table 1: Key Historical Events

     March 1981    British Parliament passes British Nationality (Hong Kong) Act
     Sept. 1982    Margaret Thatcher arrives in Beijing to discuss Hong Kong’s post-1997 future with Deng Xiaoping
     Dec. 1984     Signing of Sino-British Joint Declaration
     June 1989     More than one million people in Hong Kong protest against Tianamen Square incident
     April 1990    PRC government formally approves Basic Law
     Dec. 1990     Beginning of application period for British citizenship under British Nationality (Hong Kong) Act
     Oct. 1992     Governor Patten announces proposal for political reform
     July 1997     Hong Kong reverts to Chinese sovereignty
     Note: Excerpts from “Chronology of Key Events” in Carroll (2007).



     The paper proceeds as follows. Section 2 provides some brief historical background on this topic.
Section 3 describes the data used and reports descriptive statistics. Section 4 proposes the empirical
specifications and regression results are presented in Section 5. Section 6 summarizes and concludes.


2.      Historical background
Following the First Opium War and the Treaty of Nanjing in 1842, Hong Kong Island was ceded by
China to Britain and became the latter’s colony. The British later acquired the Kowloon Peninsula in
1860, and were leased the New Territories until 1997. As the expiration of the lease approached, the
transfer of power from British to Chinese rule created uncertainty for Hong Kong’s future. In 1982,
British Prime Minister Margaret Thatcher visited Beijing to discuss Hong Kong’s future with Deng
Xiaoping, paramount leader of the People’s Republic of China. After two years of negotiations, the Sino-
British Joint Declaration was signed in 1984. It was decided that Hong Kong was to become a Special
Administrative Region under Chinese sovereignty after 1997, an arrangement known as “one country,
two systems.” China’s mini-constitution for Hong Kong, the Basic Law, was approved in 1990.
     Hong Kong’s political transition, however, was not smooth. The Sino-British Joint Declaration had
not provided sufficient details on the structure and rules of the future Hong Kong government, and the
British and Chinese also had differing interpretations on what was agreed upon. Moreover, there was a
struggle between many Hong Kong citizens and Chinese officials over the level of democracy that should
be allowed. On June 4, 1989, the Chinese government crushed the pro-democracy public demonstrations
in Tiananmen Square and Beijing. This action added greater anxiety to an already tense environment
and made many worry about Hong Kong’s future under the rule of China. Many of the city’s residents
responded by leaving; Canada, Australia, and the United States were popular destinations. Those that
stayed saw the arrival of the last Governor, Chris Patten, in 1992 and his attempts at political reform.
His efforts of introducing a higher degree of democratization were not the most successful as they were
met with opposition from Chinese officials (Carroll, 2007, p.202). On July 1, 1997, Hong Kong reverted to
Chinese sovereignty, governed by newly appointed Chief Executive Tung Chee-hwa. The Asian financial
crisis began the next day on July 2, and Hong Kong’s economy would suffer tremendously for a prolonged
period (Kwok, 2008, p.72). As is apparent in Figure 1, mass emigration also came to a halt.
     While the handover was certainly a significant driving force of emigration out of Hong Kong, factors
including the economic conditions of Hong Kong and immigration policies of destination countries were
important elements of the migration decision as well (Skeldon, 1997, p.134). The British Nationality
(Hong Kong) Act of 1981 had denied the right of abode in Britain for citizens of Hong Kong, and in
doing so had provoked a sense of betrayal (Carroll, 2007, p.180). The British Nationality (Hong Kong)
Act of 1990 tried to rectify this by offering British nationality to 50,000 eligible people; however, in
part due to China’s refusal to recognize British citizenship of these people, there were noticeably fewer


                                                           5
Table 2: Immigrants Upon Arrival- Age, Marital Status, Schooling

                         Census year: 1991   Hong Kong    China    Other    Natives
                         Age                    25.7       39.9     28.6     32.5
                                               (14.2)     (17.6)   (16.9)   (21.2)
                         Married                0.62       0.71     0.57     0.52
                                               (0.49)     (0.45)   (0.49)   (0.50)
                         Years of school        13.4       11.7     13.4     12.1
                                                (2.9)      (4.5)    (3.6)    (3.3)
                         Observations            642        376     4492    670290
                         Census year: 1996   Hong Kong    China    Other    Natives
                         Age                    28.8       39.8     28.5     33.3
                                               (16.3)     (20.9)   (16.4)   (21.4)
                         Married                0.66       0.71     0.60     0.49
                                               (0.48)     (0.45)   (0.48)   (0.50)
                         Years of school        13.1       11.7     13.3     12.4
                                                (3.2)      (4.5)    (3.5)    (3.3)
                         Observations            576        589     5385    648457
                         Census year: 2001   Hong Kong    China    Other    Natives
                         Age                    31.6       30.6     27.2     34.5
                                               (17.7)     (16.0)   (16.4)   (21.6)
                         Married                0.54       0.81     0.66     0.46
                                               (0.50)     (0.39)   (0.47)   (0.50)
                         Years of school        13.4       15.1     14.1     12.8
                                                (3.0)      (3.2)    (3.5)    (3.2)
                         Observations             69       1152     6074    585752
                         Notes: Means for each immigrant sample and natives are re-
                         ported. Standard deviations in parentheses. The smallest
                         value of observations across variables is shown.


applicants into this scheme than had been expected (Carroll, 2007, p.193).
     Canada, viewed as a multicultural nation, accommodated many immigrants by contrast, especially
the tremendous influx of Hong Kong immigrants during this episode. In 1985, Canada altered its
policy towards admitting business immigrants by expanding its Business Immigration Program to include
entrepreneurs, self-employed persons, and investors (Li, 2003a, p.26). Although regulations became
stricter over the next 10 years, this was still a viable avenue of immigration for many belonging to Hong
Kong’s middle class. With so many businessmen and skilled professionals leaving Hong Kong, the “brain
drain” phenomenon appeared evident. The handover seemed to affect the economic class immigrants
more than the family class, as the rapid decline in immigration to Canada post-1997 was not observed
for the latter (Li, 2005). Ultimately, the decision of individuals and families to leave was complex, as
it involved political and economic motives. Salaff and Wong (1995) examine the factors that may have
influenced each social class in Hong Kong and how each might have responded.


3.     Data
The data are drawn from the Census Public Use Microdata Files (PUMF) on Individuals from the
1991, 1996, 2001, and 2006 Canadian Censuses.4 They are based on 3%, 2.8%, 2.7%, and 2.7% of the
population enumerated in the respective Censuses. The sample excludes the Atlantic provinces and
territories as the data cannot be coded to the same level of detail as those of the other provinces. Not
surprisingly, the Census has evolved over the span of 15 years. The variable coding on some of the same
questions asked varies across the Censuses, as categories may become more or less disaggregated over
time.5 Hence, the most aggregated level is taken to accommodate all of the Censuses. The 2006 Census
actually introduced the most changes; statistics such as age and years of schooling are unavailable as
numbers but are instead given with categories. In these cases, the median values within the categorical



                                                     6
Table 3: Immigrants Upon Arrival- Hours of Work and Wages

                            Census year: 1991    Hong Kong       China      Other     Natives
                            Weeks worked            28.6          31.6       29.8       41.7
                                                   (16.6)        (16.2)     (16.7)     (15.0)
                            Wage                   17691         14191      15547      30289
                                                  (24022)       (13969)    (18249)    (25564)
                            Observations            271           191        2025     334736
                            Census year: 1996    Hong Kong       China      Other     Natives
                            Weeks worked            31.4          32.3       32.9       41.7
                                                   (16.6)        (16.9)     (17.5)     (15.1)
                            Wage                   20219         13216      16767      31242
                                                  (24395)       (15629)    (21673)    (27228)
                            Observations            117           164        1408     307279
                            Census year: 2001    Hong Kong       China      Other     Natives
                            Weeks worked            36.7          31.4       31.9       43.3
                                                   (15.6)        (17.4)     (17.3)     (14.0)
                            Wage                   33378         14169      18805      34067
                                                  (42343)       (17866)    (24199)    (29706)
                            Observations             23           356        1726     294944
                            Notes: Means for each immigrant sample and natives are re-
                            ported. All samples are restricted to persons aged 15-65. Stan-
                            dard deviations in parentheses. The smallest value of observations
                            across variables is shown. Real wage is measured in 2005 CDN.
                            Weeks worked refers to the number of weeks in the year prior to
                            the Census worked (includes weeks of paid vacation or sick leave
                            with pay or paid absence on training courses; excludes housework
                            or other maintenance or repairs around the home and volunteer
                            work).


groups were used as approximations. I have excluded the 2006 Census as a robustness check for the
regressions in Section 5.
   Tables 2 and 3 provide summary statistics on a few characteristics of immigrants upon arrival.6 This
is the cohort that arrived one year before the Census year.7 The 2006 Census has been omitted because
there are much fewer observations for Hong Kong immigrants, so any inference based on them might be
misleading. One immediate trend observed in Table 2 is the rapid decrease of Hong Kong immigrants into
Canada in the intercensal period from 1996 to 2001, and the marked increase of emigration from China
to Canada. This matches the patterns in Figure 1. The number of immigrants from other countries
has also steadily increased. The age of the typical Hong Kong immigrant has risen over the years, from
mid-20’s in 1990 to early 30’s in 2000. On average, they have obtained slightly more years of schooling
than the Canadian-born, and this figure has stayed relatively constant over the years. As a result, the
level of education has fallen below that of the Chinese immigrant for the more recent cohorts.
   Table 3 presents descriptive statistics on two labor market outcomes. The sample contains only
individuals from the age of 15 to 65, and this restriction is applied for the remainder this paper (except
Table 4). Table 3 shows that that all groups of newly arrived immigrants tended to work 7 to 13 fewer
weeks in their first year than a typical native. Furthermore, their earnings in the year of arrival are on
average considerably lower; this is in part attributable to the fact that they are working less. Out of the
broad immigrant groups shown, Hong Kong immigrants have fared the best in this regard; the difference
is especially striking when comparing them to the Chinese.8
   Focusing specifically on Hong Kong residents of Canada, Table 4 tabulates the percentage living in
each major metropolitan area, and Table 5 presents statistics on labor force activity.9 Toronto and
Vancouver are the overwhelming favorite destinations to settle in, and now have extensive communities
established within each metropolis. The two cities in Alberta combine for close to 10% of the Hong Kong
immigrant population. Employment was high for those in the labor force. An unemployment rate of



                                                            7
Table 4: Areas of Residence                                 Table 5: Labor Force Activity

      Metropolitan Area   % of HK Immigrants                  Census year            1991    1996   2001    2006
      Toronto                    47.0                         Observations           4094    6164   5986    5637
      Vancouver                  36.4                         % Employed             63.7    53.8   59.3    64.5
      Calgary                    4.74                         % Unemployed           5.37    5.06   4.13    4.01
      Edmonton                   3.18                         % Not in labor force   30.9    41.1   36.6    31.5
      Montreal                   2.28                         Notes: The sample is restricted to persons aged 15-
      Ottawa - Gatineau          1.43                         65. Figures are based on the week prior to enumer-
      Source: 2001 Canadian Census.                           ation.


4-5% indicates that those who were searching for work could generally find it. About 30 to 40% of Hong
Kong immigrants were not in the labor force in each Census year.
     Figures 2 to 4 below give some preliminary evidence on the degree of assimilation for each immigrant
group; it is a first glance combining the cross-sections of the four Censuses. Each panel plots the mean
wage for a specific cohort of immigrants across the Census years. In each Panel (a), immigrants who
arrived between 1980 and 1989 are followed across the Census years, while Panel (b) captures the 1990-
1997 Cohort. The line for natives is identical in Panels (a) and (b), and is an average of all Canadian-born
individuals also aged 15-65. Figure 2 includes the entire sample; it shows that the 1980-89 Hong Kong
Cohort earned consistently more than the other immigrants who arrived around the same period of time.
The slopes of the lines across the immigrant groups are also similar. By contrast, the 1990-97 Hong Kong
Cohort faced a large earnings gap in comparison with the average native even by 2006. The advantage
over Chinese or Other immigrants is also absent, and in fact they had the worst performance out of all
the groups in 2001.
     Figures 3 and 4 examine subsets of the sample, restricting it to the two major provinces of British
Columbia and Ontario.10 Immigrants in British Columbia tend to fare worse than the whole of Canada,
but natives living there have the opposite experience. Hence the earnings gap is even greater in that
province. The wages of the 1980-89 Hong Kong immigrants are diminished greatly as well, making them
equal with the average of Other immigrants. The situation is more optimistic in Ontario, as Figure 4
illustrates. The levels of wages are higher on average. However, the contrast between the two cohort
groups of Panels (a) and (b) is still persistent.
    I have also explored the dimension of age, restricting the sample to a certain age group, and tracking
this group over the Census years. For instance, in the 1991 Census year, the mean wage for individuals
aged 20 to 40 is computed; then in 1996, the same statistic for those aged 25 to 45 is found, and so
on. The plots are similar to Figure 2, whether comparing across immigrant groups or across cohorts,
and thus I omit them.11 One notable difference is that the Hong Kong 1980-89 immigrant cohort on
average earned more than natives of identical age groups in 3 out of 4 Census years. While these graphs
can provide an informative picture of the level of earnings, conclusions on assimilation and earnings
convergence will require more rigorous econometric analysis, to which I now turn.


4.      Empirical specifications
I use several specifications to examine the assimilation of immigrants in Canada. I begin by estimating
cross-section regressions for each census year, which serves as a point of comparison to previous literature.
The regression equation is specified by

                             log wi = Xi φ + δAi + α0 Ii + α1 (Ii · Y SMi ) + εi ,                           (4.1)



                                                      8
45000                                                                              45000

                  40000                                                                              40000
   Wage (2005$)




                                                                                      Wage (2005$)
                  35000                                                                              35000

                  30000                                                                              30000

                  25000                                                                              25000

                  20000                                                                              20000

                  15000                                                                              15000
                       1991        1996                 2001               2006                          1991         1996                 2001               2006
                                          Census year                                                                        Census year
                              Hong Kong (80−89)                China (80−89)                                     Hong Kong (90−97)                China (90−97)
                              Other (80−89)                    Natives                                           Other (90−97)                    Natives


                                       (a)                                                                                (b)

Figure 2: Average wage of immigrants and natives (aged 15-65) over time. (a) 1980-89 immigrant cohort; (b) 1990-97
immigrant cohort.




                  45000                                                                              45000

                  40000                                                                              40000

                                                                                                     35000
   Wage (2005$)




                                                                                      Wage (2005$)




                  35000
                                                                                                     30000
                  30000
                                                                                                     25000
                  25000
                                                                                                     20000
                  20000
                                                                                                     15000
                  15000
                       1991        1996                 2001               2006                          1991         1996                 2001               2006
                                          Census year                                                                        Census year
                              Hong Kong (80−89)                China (80−89)                                     Hong Kong (90−97)                China (90−97)
                              Other (80−89)                    Natives                                           Other (90−97)                    Natives


                                       (a)                                                                                (b)

Figure 3: Average wage of immigrants and natives (aged 15-65) in British Columbia over time. (a) 1980-89 immigrant
cohort; (b) 1990-97 immigrant cohort.




                  45000                                                                              45000

                  40000                                                                              40000
   Wage (2005$)




                                                                                      Wage (2005$)




                  35000                                                                              35000

                  30000                                                                              30000

                  25000                                                                              25000

                  20000                                                                              20000

                  15000                                                                              15000
                       1991        1996                 2001               2006                           1991        1996                 2001               2006
                                          Census year                                                                        Census year
                              Hong Kong (80−89)                China (80−89)                                     Hong Kong (90−97)                China (90−97)
                              Other (80−89)                    Natives                                           Other (90−97)                    Natives


                                       (a)                                                                                (b)

Figure 4: Average wage of immigrants and natives (aged 15-65) in Ontario over time. (a) 1980-89 immigrant cohort; (b)
1990-97 immigrant cohort.




                                                                                  9
where wi is the wage of individual i, Xi are socioeconomic characteristics, Ai is age, Ii is an indicator
variable for the individual being an immigrant, and Y SMi is years since migration (0 if the individual
is a native). The variables included in Xi are a constant, years of schooling, and dummy variables for
being married, female, a member of the visible minority, and living in a metropolitan area. For flexibility,
squared terms are also included, so the following variations of equation (4.1) are estimated:

                log wi = Xi φ + δ1 Ai + δ2 A2i + α0 Ii + α1 (Ii · Y SMi ) + α2 (Ii · Y SMi2 ) + εi ,          (4.2a)
                log wi = Xi φ + δ1 Ai +    δ2 A2i   + α0,2 1[P OB2]i + α0,3 1[P OB3]i
                          + ... + α0,17 1[P OB17]i + α1 (Ii · Y SMi ) + α2 (Ii · Y SMi2 ) + εi ,              (4.2b)

where 1[P OB2] to 1[P OB17] are indicator variables for the individual’s place of birth. The excluded
group, P OB1, is the Canadian-born.12 In this second specification, country fixed effects are used instead
of the immigrant indicator to allow for differential impact due to the country of origin. It is assumed
that the individual stayed at their origin country until moving to Canada.
   As noted by Borjas (1985), this estimation strategy suffers from two confounding factors when inter-
preting the α coefficients: selective return migration and changes in cohort quality. To account for the
latter, the estimating equation proposed is:

     log wit = Xit φ + δ1 Ait + δ2 A2it + γIi + α1 (Ii · Y SMit ) + α2 (Ii · Y SMit2 ) + β1 Ii Cit + εit ,     (4.3)

where Cit is the calendar year of immigrant i’s arrival recorded at Census year t. This term is meant
to capture the cohort-specific effects. With a single cross-section, the model cannot be estimated due to
perfect collinearity: t = Y SMit +Cit . However, all coefficients are identified with multiple cross-sections.
For the variables not interacted with the immigrant dummy, i.e., Xit and age, common coefficients have
been imposed in the above specifications. While this does not have to be the case, in unreported results,
I have allowed for different slopes in, for instance, age between natives and immigrants and found the
coefficients to be not too different. Even the binary distinction between natives and immigrants is
somewhat ad-hoc, as further disaggregation of immigrants is possible. However, since these variables are
not the focus of this study, I maintain the assumption of common slope coefficients for them throughout.
   There is yet another effect that is unaccounted for in equation (4.3) though, which is the period effect.
This can be resolved by adding dummy variables to capture changes in wages specific to the Census years.
As discussed in Borjas (1994), the restriction that the period effect is identical for immigrants and natives
alike is imposed as one method of solving the identification problem. Hence, the third set of regressions
estimate:

            log wit = Xit φ + δ1 Ait + δ2 A2it + α0 Ii + α1 (Ii · Y SMit ) + α2 (Ii · Y SMit2 ) + β1 Ii Cit
                                                                                                               (4.4)
                      + γ1 1[1996]t + γ2 1[2001]t + γ3 1[2006]t + εit .

   Finally, to allow for potentially differing slopes in the assimilation profile, I interact Y SMit and
Y SMit2 with P OB indicator variables. Since the primary interest is on Hong Kong immigrants, as
before, I create the three immigrant categories of Hong Kong, China, and Other. Moreover, the cohort
year term is replaced with categorical variables for the groups of immigrants who arrived before or in
1984, from 1985 to 1989, from 1990 to 1997, and 1998 and beyond. These years were specifically chosen
to fit the historical context as reviewed in Section 2. These dummy variables are then interacted with




                                                           10
the immigrant group indicators. The following equation is thus estimated:
                                                        X            h                                   
           log wit = Xit φ + δ1 Ait + δ2 A2it +                          α0,j 1[j]i + α1,j 1[j]i · Y SMit
                                                  j=HK,China,Other
                                                                                             
                     + α2,j 1[j]i · YSMit2  + θ1,j 1[j]i · 1[1985-89]i + θ2,j 1[j]i · 1[1990-97]i             (4.5)
                                            i         X
                     +θ3,j 1[j]i · 1[≥ 1998]i +                   γk 1[k]t + εit .
                                                     k=1996,2001,2006


The base immigrant cohort group is those that arrived before 1985. Joint hypothesis tests can be carried
out to test whether or not the coefficients on the assimilation variables are indeed significantly different
from one another.


5.      Regression results

5.1.     Cross-section and pooled cross-section results
Table 6 presents the results for the estimation of equations (4.2a) and (4.2b), using the four Census data
separately as cross-sections. The signs of the coefficients on the Xi variables, age, and age-squared are
what one would expect. (Note that all of the t-statistics in both Tables 6 and 7 are very large, so I have
chosen not to indicate statistical significance in these two tables.) Being married, male, not of visible
minority, and living in a metropolitan area all contribute to having higher wages, while age is associated
with diminishing returns. Not surprisingly, we find positive returns for years of schooling.
     For any year of Census data, being an immigrant puts the individual at a severe disadvantage in
earnings upon entry. From the 1991 Census cross-section result in column 1, the average immigrant
faces an initial wage gap of roughly 36% in comparison with a native with identical socioeconomic
characteristics and age. This gap tends to increase over the Census years, and is greater than 50%
with the 2001 and 2006 Census data. In even columns, place of birth fixed effects are used instead
to differentiate the origin country of the migrant. Although there are 16 P OB categories in total
(excluding the base group of the Canadian-born), the United States and China have been selected as
groups of comparison for immigrants from Hong Kong (the other coefficients have been suppressed to
ease exposition).13 While the magnitudes of the coefficients on the Hong Kong indicator variables are
consistently smaller than those on China, they are only less than those of the United States in 1991.
Moreover, the magnitudes are slightly smaller than the immigrant dummies, indicating that migrants
from Hong Kong faced a lower earnings gap upon landing relative to the average immigrant. Again, this
gap is increasing from 1991 to 2006, suggesting other factors like cohort quality should be considered.
Lastly, examining the coefficients on the years since migration variables, the length of stay in Canada
for an immigrant definitely has a positive effect on earnings, but this declines over time. From this
cross-sectional evidence, it would appear that the annual rate of earnings convergence is quite low for
immigrants, beginning at around 2% to 3% in the first year and decreasing thereafter (because of the
squared term).
     Next, I turn to estimations that make use of all four Censuses. In Table 7, the regressions of columns
1 and 2 combine all of the cross-sections, but do not include the immigrant’s year of arrival. Conversely,
columns 3 to 4 do include the cohort year, and correspond respectively to the estimations of equation (4.3)
and its variant with P OB dummies. Columns 5 and 6 control for period effects as well as cohort quality,
and therefore present the regression results of equation (4.4).14 The Xit and age variables have been


                                                        11
Table 6: OLS Estimates, Cross-Section Regression Results

                                                                    Census Year
                                      1991                  1996                   2001                   2006
                                  (1)        (2)        (3)        (4)        (5)         (6)        (7)         (8)
   Immigrant                    -0.360                -0.491                -0.520                 -0.508
                                (-33.2)               (-39.7)               (-43.2)                (-39.7)
   Hong Kong                               -0.301                -0.448                 -0.483                 -0.496
                                           (-12.6)               (-17.9)               (-20.7)                (-15.7)
   China                                   -0.418                -0.512                 -0.644                 -0.757
                                           (-16.9)               (-20.7)               (-28.5)                (-29.0)
   United States                           -0.358                -0.435                 -0.455                 -0.400
                                           (-19.6)               (-20.3)               (-21.0)                (-14.6)
   Yrs since migration           0.024      0.021      0.035      0.032      0.035      0.033       0.033       0.034
                                (22.9)     (19.7)     (30.2)     (27.3)     (30.2)      (27.5)     (23.7)      (23.0)
   Yrs since migration2 /100    -0.032     -0.025     -0.052     -0.046     -0.047      -0.046     -0.041      -0.048
                                (-12.9)    (-9.87)    (-19.8)    (-17.0)    (-18.7)    (-17.3)     (-11.7)    (-12.7)
   Yrs of school                 0.073      0.074      0.077      0.078      0.077      0.078       0.102       0.103
                               (134.2)    (132.8)    (121.2)    (120.8)    (117.2)     (116.8)     (97.4)      (97.3)
   Married                       0.135      0.134      0.163      0.163      0.139      0.139       0.075       0.081
                                (36.4)     (36.2)     (39.2)     (39.2)     (34.3)      (34.3)     (13.5)      (14.6)
   Female                       -0.551     -0.552     -0.489     -0.490     -0.466      -0.467     -0.480      -0.482
                               (-172.0)   (-172.2)   (-135.9)   (-136.2)   (-130.7)   (-131.0)     (-95.9)    (-96.3)
   Visible minority             -0.154     -0.136     -0.176     -0.182     -0.186      -0.164     -0.204      -0.144
                                (-21.0)    (-11.7)    (-21.9)    (-15.2)    (-25.1)    (-15.3)     (-20.9)    (-10.5)
   Metropolitan area             0.218      0.218      0.189      0.190      0.188      0.188       0.181       0.182
                                (63.2)     (62.9)     (48.7)     (48.7)     (46.9)      (46.8)     (31.6)      (31.7)
   Age                           0.208      0.207      0.228      0.228      0.221      0.221       0.266       0.266
                               (229.3)    (229.1)    (216.5)    (215.9)    (216.4)     (215.7)    (163.5)     (163.2)
   Age2 /100                    -0.228     -0.227     -0.249     -0.248     -0.244      -0.243     -0.307      -0.307
                               (-199.7)   (-199.3)   (-186.7)   (-186.0)   (-186.9)   (-186.2)    (-146.3)   (-146.1)
   Constant                      4.786      4.775      4.187      4.179      4.446      4.440       3.364       3.359
                               (280.7)    (279.3)    (209.6)    (208.5)    (231.5)     (230.7)    (108.2)     (108.0)
   R2                            0.326      0.327      0.322      0.323      0.315      0.315       0.188       0.189
   N                           412,905    412,905    371,764    371,764    374,418     374,418    432,349     432,349
   Notes: The dependent variable is (log) real wage (in 2005 CDN). The sample is restricted to persons aged 15-65.
   All POB categories are included in regressions with country fixed-effects. Base group is male native. T-statistics
   in parentheses.


suppressed since they are not the variables of interest. They are similar in magnitude across the columns.
   First, the results of columns 1 and 2 are compared with columns 3 and 4 to examine the effects
of controlling for cohort quality. However, the intercept coefficient is not directly comparable between
say columns 1 and 3, since cohort year must be specified. As an example, to interpret the Immigrant
coefficient in column 3, for an immigrant arriving in Canada in 1980, the wage gap upon entry is -0.491
(−9.657 × 1.980 + 18.63), or approximately 49%. This disparity with the natives is increasing in the date
of arrival: the negative coefficient on cohort year clearly indicates that there has been deterioration in
the quality of the average Canadian immigrant. Thus, the initial wage gap tends to be overestimated in
column 1 for the majority of cohorts under analysis. Moreover, comparing the linear coefficient of years
since migration (which will dominate the quadratic term for shorter time periods), it is obvious that
omitting the cohort-specific effects will overestimate the rate of assimilation by a rather large amount.
   From column 4, a Hong Kong immigrant landing in 1980 would have about 43% less annual wage
income than a native, ceteris paribus. The same migrant arriving 10 years later would experience a much
larger gap at close to 52%. However, the initial advantage of Hong Kong immigrants over immigrants
from the United States or China still holds, and it is around 6 percentage points over the former and 14
over the latter. The growth rate of wages in comparison with the native-born for Hong Kong immigrants
starts at around 3.6% in the first year, and falls thereafter. Comparing column 2 and 4, the same
conclusions regarding the estimation bias of ignoring declining cohort quality are reached. This finding
of negative cohort effects is consistent with the research on immigrants in Canada (e.g., Baker and


                                                          12
Table 7: OLS Estimates, Cohort Quality, and Period Effects

                                                                                          Pooled Cross-Section
                                          Cross-Section           Pooled Cross-Section     With Period Effects
                                         (1)          (2)            (3)         (4)          (5)        (6)
         Immigrant                     -0.666                       18.63                   10.48
                                       (-81.8)                     (20.5)                   (10.4)
         Hong Kong                                  -0.600                      17.22                   8.946
                                                   (-42.0)                     (18.7)                  (8.86)
         China                                      -0.750                      17.08                   8.806
                                                   (-53.2)                     (18.5)                  (8.71)
         United States                              -0.654                      17.16                   8.882
                                                   (-48.1)                     (18.6)                  (8.80)
         Yrs since migration            0.048        0.045          0.038       0.036       0.042       0.040
                                       (64.1)       (59.1)         (43.8)      (40.8)       (46.8)     (43.9)
         Yrs since migration2 /100     -0.073       -0.068         -0.072      -0.067       -0.071     -0.066
                                       (-43.3)     (-39.7)         (-42.7)     (-39.0)     (-42.0)     (-38.3)
         Cohort year/1000                                          -9.657      -8.916       -5.576     -4.775
                                                                   (-21.1)     (-19.3)     (-11.1)     (-9.44)
         I(1996)                                                                            -0.096     -0.097
                                                                                           (-39.2)     (-39.6)
         I(2001)                                                                            -0.018     -0.019
                                                                                           (-7.11)     (-7.41)
         I(2006)                                                                            -0.097     -0.097
                                                                                           (-29.5)     (-29.7)
         R2                            0.263        0.263           0.263       0.263       0.264       0.264
         N                           1,564,485    1,564,485       1,564,485   1,564,485   1,564,485  1,564,485
         Notes: The dependent variable is (log) real wage (in 2005 CDN). The sample is restricted to persons
         aged 15-65. All regressions include controls Xit as listed in the text, Age, and Age2 . All POB cate-
         gories are included in regressions with country fixed-effects. Base group is male native. T-statistics
         in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level.


Benjamin (1994), Bloom et al. (1995), and Campolieti et al. (2013)), as well as studies in other contexts
such as Abramitzky et al. (2014). As new immigrants arrive, they expand the networks created by their
predecessors. In particular, many Hong Kong immigrants have settled in areas with a high concentration
of Hong Kong and Chinese residents and have greatly benefited from established ethnic businesses and
specialized services (Li, 1998, p.112). These communities and networks make the adjustment for new
immigrants easier, and so the lower quality of more recent cohorts is less detrimental to their assimilation
experience. This idea is consistent with the evidence for negative emigrant selection over time.
   Next, comparing columns 3 and 4 with columns 5 and 6, the bias from omitting period effects is
determined. The coefficients on the Census year indicator variables in columns 5 and 6 are all highly
statistically significant. Furthermore, the addition of period effects in the regression lowers the magnitude
of the coefficients on the immigrant or P OB dummies and cohort year, as compared to columns 3 and
4. In particular, the magnitude of the coefficient on cohort year in, for example, column 6 is roughly
half of that in column 4. The period-specific effects may have been absorbed through these variables,
resulting in an overestimation of the size of their effects. Repeating the same calculation as above, a
migrant from Hong Kong entering Canada in 1980 would face a wage gap of approximately 51%. The
annual rate of earnings convergence to natives is again around 4%, a figure comparable to those in Grant
(1999). Comparing the results of Tables 6 and 7, it is clear that controlling for cohort quality and period
effects will change the estimated rate of assimilation. Of course, a more complete picture of assimilation
must take into account both the entry gap of wages and its subsequent growth. The empirical results
of column 6 imply that for an immigrant from Hong Kong who arrived in 1980, it will take more than
18 years for their earnings to overtake a native, all else equal. By contrast, for the same entry year, an
American will take 24 years, and according to this specification, a migrant from China will actually find
it impossible for their earnings to reach the level of natives. Of course, those migrating later will take


                                                             13
Table 8: OLS Estimates, Province Fixed Effects

                                         (1)           (2)
Immigrant                             10.14***
                                       (10.2)
Hong Kong                                           8.510***
                                                     (8.44)
China                                               8.411***
                                                     (8.34)
United States                                       8.549***
                                                     (8.48)
Yrs since migration                    0.042***     0.040***
                                        (47.1)       (44.1)
Yrs since migration2 /100             -0.071***    -0.066***
                                        (-42.6)      (-38.5)
Cohort year/1000                      -5.434***    -4.605***
                                        (-10.9)      (-9.12)
British Columbia                       0.086***     0.086***
                                        (23.8)       (23.8)
Ontario                                0.105***    0.104***
                                        (43.1)       (43.0)
Immigrant×British Columbia              -0.004
                                        (-0.49)
Immigrant×Ontario                     0.073***
                                        (11.8)
Hong Kong×British Columbia                            0.003
                                                     (0.08)
Hong Kong×Ontario                                   0.159***
                                                     (4.69)
China×British Columbia                                0.019
                                                     (0.60)
China×Ontario                                       0.084***
                                                     (2.82)
United States×British Columbia                     -0.128***
                                                     (-4.17)
United States×Ontario                                 0.025
                                                     (0.96)
POB FE                                    N             Y
R2                                      0.265         0.266
N                                     1,564,485    1,564,485
Notes: The dependent variable is (log) real wage (in 2005
CDN). The sample is restricted to persons aged 15-65.
All regressions include controls Xit as listed in the text,
Age, Age2 , and dummies for Census years. All POB cate-
gories are included in regressions with country fixed-effects,
and all their interactions are included in regressions with
province fixed-effects. Base group is male native in 1991
Census; the omitted native/immigrant group is persons re-
siding in provinces other than British Columbia and On-
tario. T-statistics in parentheses. ***, **, and * indicate
statistical significance at the 1%, 5%, and 10% level.




                             14
even longer to achieve this parity. Since the quadratic functional form on years since migration induces
diminishing returns to the length of stay, I also present results using a linear specification for comparison
in Table 9.
   Motivated by the suggestive evidence in Figures 3 and 4 that geographic location within the host
nation Canada may have an important impact on wages, Table 8 introduces categorical variables for
provinces into the regression analysis.15 Specifically, three categories are created: British Columbia,
Ontario, and other provinces. The base group in both specifications is the male native not residing in
British Columbia nor Ontario in the 1991 Census year. Thus, a Canadian-born individual earns 8 to
10% more per year when living in either British Columbia or Ontario. Examining column 2, upon entry
in 1980, a Hong Kong migrant not residing in Ontario would face an earnings disadvantage of roughly
61% compared to a native also located in the same province category (the coefficient on the interaction
term between Hong Kong and British Columbia is not statistically significant). By contrast, Hong Kong
immigrants living in Ontario fare much better upon landing and experience an earnings gap almost
16 percentage points smaller even when comparing within the province of Ontario. For a Hong Kong
migrant arriving in 1980 and residing in Ontario, the length of time required for assimilation within the
province is reduced to 15 years.


5.2.    Cohort quality and assimilation profiles
The specifications in Tables 7 and 8 have restricted a common coefficient for the cohort year variable
across immigrant groups. To explore whether there is decline in cohort quality for Hong Kong immigrants
associated with the relevant historical events, this restriction is relaxed and equation (4.5) is estimated.
When interpreting these results for the non-Hong Kong immigrants, one should keep in mind that the
cutoff years for defining the cohort groups may have been relevant for people from Hong Kong, but not
necessarily for the other groups. Table 9 presents the results. The omitted immigrant cohort group is
those who arrived before or in 1984. For example, a Hong Kong immigrant in this category had an initial
earnings disadvantage of around 52% relative to natives according to the estimate in column 1. This is
a severe gap requiring a long period of time to overcome, consistent with earlier results; and as before,
the negative entry effect is even larger for a Chinese immigrant of this cohort. Hong Kong migrants that
arrived from 1985 to 1989 appear to have fared much better initially in the labor market, in comparison
to their predecessors; the coefficient is positive and large. However, between 1990 and 1997, immigrants
of significantly lower quality entered Canada from Hong Kong. The coefficient on the interaction term
of the 1990-97 Cohort dummy and the Hong Kong dummy indicates the earnings disadvantage of this
later cohort compared to the earlier ≤1984 Cohort is approximately 17.8 percentage points larger. This
gap is even greater when compared with the 1985-89 Cohort. Leavers following the handover in 1997
were also of lower quality. The combined pressures of the Tiananmen Square incident in 1989 and the
handover in 1997 may have led to the observed fall in Hong Kong emigrant quality. Although the main
analysis focuses on immigrants from Hong Kong, a similar pattern is observed for Chinese and Other
immigrants. (This correlation across the groups might suggest that Canadian immigration policies had
similar impact on immigrants from around the world.)
   The above discussion has focused on the cohort effects and in particular the initial wage gap, one
component of the immigrant assimilation profile. Next, the possibility of differential slopes of assimilation
across immigrant groups is explored. The interaction terms of the years since migration variables with
the immigrant group indicators provide evidence on whether or not the rate of assimilation varies. The
annual growth rate of wages above that of the natives is close to 4.5% in the first year for Hong Kong


                                                     15
Table 9: Differential Rates of Earnings Convergence and Cohort Quality

                                                      Excluding
                                        Full Sample       2006    Full Sample   Full Sample
                                             (1)           (2)         (3)            (4)
Hong Kong                                -0.521***    -0.456***    -0.330***     -0.150***
                                           (-10.2)      (-7.60)      (-7.98)       (-2.97)
China                                    -0.732***    -0.652***    -0.436***     -0.252***
                                           (-13.3)      (-9.73)      (-11.4)       (-6.19)
Other                                    -0.546***    -0.438***    -0.243***     -0.171***
                                           (-36.9)      (-27.1)      (-26.7)       (-18.1)
1985-89 Cohort×Hong Kong                  0.114***      0.094**     0.096***       -0.124*
                                           (3.48)        (2.37)      (2.92)        (-1.90)
1990-97 Cohort×Hong Kong                 -0.178***    -0.215***    -0.221***     -0.595***
                                           (-4.88)      (-5.09)      (-6.16)       (-9.34)
≥1998 Cohort×Hong Kong                    -0.217**       -0.178    -0.309***     -0.533***
                                           (-2.17)      (-0.98)      (-3.11)       (-2.68)
1985-89 Cohort×China                       0.068*        0.078*       0.021         -0.110
                                           (1.72)        (1.65)      (0.55)        (-1.44)
1990-97 Cohort×China                       -0.027        -0.031    -0.125***     -0.421***
                                           (-0.73)      (-0.66)      (-3.64)       (-7.57)
≥1998 Cohort×China                       -0.293***    -0.343***    -0.482***     -1.103***
                                           (-5.92)      (-4.89)      (-11.3)       (-16.9)
1985-89 Cohort×Other                      0.045***        0.017    -0.043***     -0.192***
                                           (4.48)        (1.53)      (-4.61)       (-11.1)
1990-97 Cohort×Other                     -0.039***    -0.105***    -0.169***     -0.461***
                                           (-3.78)      (-8.67)      (-19.1)       (-29.2)
≥1998 Cohort×Other                       -0.176***    -0.222***    -0.393***     -0.781***
                                           (-10.9)      (-9.71)      (-28.2)       (-30.9)
Yrs since migration×Hong Kong             0.045***     0.042***     0.017***      0.009***
                                           (8.66)        (7.41)      (9.17)         (4.04)
Yrs since migration2 /100×Hong Kong      -0.074***    -0.075***
                                           (-5.19)      (-5.30)
Yrs since migration×China                0.054***      0.047***    0.018***        0.011***
                                           (12.1)        (8.45)     (12.2)          (6.83)
Yrs since migration2 /100×China          -0.083***    -0.073***
                                           (-8.82)      (-6.65)
Yrs since migration×Other                 0.038***     0.029***    0.010***        0.007***
                                           (33.6)        (23.8)     (30.8)          (21.7)
Yrs since migration2 /100×Other          -0.054***    -0.040***
                                           (-25.2)      (-17.8)
Yrs since migration×1985-89 Cohort×HK                                           0.012***
                                                                                  (2.81)
Yrs since migration×1990-97 Cohort×HK                                           0.033***
                                                                                  (6.18)
Yrs since migration×≥1998 Cohort×HK                                               0.019
                                                                                  (0.54)
                                                                    Continued on next page




                                           16
Table 9: Concluded

                                                                         Excluding
                                                         Full Sample       2006       Full Sample     Full Sample
                                                              (1)           (2)            (3)             (4)
        Yrs since migration×1985-89 Cohort×China                                                         0.003
                                                                                                         (0.55)
        Yrs since migration×1990-97 Cohort×China                                                       0.023***
                                                                                                         (4.68)
        Yrs since migration×≥1998 Cohort×China                                                         0.124***
                                                                                                         (10.4)
        Yrs since migration×1985-89 Cohort×Other                                                       0.010***
                                                                                                         (7.31)
        Yrs since migration×1990-97 Cohort×Other                                                       0.030***
                                                                                                         (19.4)
        Yrs since migration×≥1998 Cohort×Other                                                         0.087***
                                                                                                         (16.6)
        R2                                                  0.264          0.322          0.264          0.264
        N                                                 1,564,485      1,146,535      1,564,485      1,564,485
        Notes: The dependent variable is (log) real wage (in 2005 CDN). The sample is restricted to persons
        aged 15-65. All regressions include controls Xit as listed in the text, Age, Age2 , and dummies for Cen-
        sus years. Base group is male native in 1991 Census. Base immigrant group is the cohort that arrived
        before or in 1984. T-statistics in parentheses. ***, **, and * indicate statistical significance at the 1%,
        5%, and 10% level.


immigrants, but drops to 3% in 10 years. Because the magnitude of the linear term dominates the
quadratic in determining the steepness of the assimilation profile, these figures are much larger for the
Chinese than either Hong Kong or Other immigrants. Some simple statistical tests can confirm that the
assimilation profiles are indeed different across the immigrant groups. First, the entry earnings gap can
be compared. For instance, the null hypothesis for the ≤1984 Cohort would be H0 : Hong Kong = China
= Other. Similarly, for the 1985-89 Cohort, we would test H0 : Hong Kong + 1985-89 Cohort × Hong
Kong = China + 1985-89 Cohort × China = Other + 1985-89 Cohort × Other. These types of Wald
tests are easily rejected at the 1% confidence level for all cohort groups. Second, the null hypothesis for
testing the equality of assimilation profile slopes is

                 H0 : Y SM × HK = Y SM × China = Y SM × Other                          AND
                        Y SM /100 × HK = Y SM /100 × China = Y SM 2 /100 × Other.
                                2                         2



The F -statistic for this test is 3.93 with an accompanying p-value of 0.0034, so this null hypothesis is
rejected at the 1% confidence level as well.
   Column 2 of Table 9 excludes the 2006 Census from the sample and is intended as a robustness check,
for concerns of data quality as explained in Section 3. Note that with this restriction, data is available up
to the 2001 Census; hence, the sample for immigrant groups arriving post-1997, i.e., the ≥1998 Cohort,
is smaller. The magnitudes of the coefficients are comparable across columns 1 and 2, and the similar
qualitative findings emerge.
   Figure 5 plots the assimilation profiles of different cohorts implied by the full sample estimates of
column 1. As expected, the 1985-89 Cohort achieves the quickest convergence with natives at 11.1 years
for immigrants from Hong Kong, while it takes somewhat longer for others (16.5 years for Chinese and
17.6 for Other). For the earlier cohort group arriving before or in 1984, years to equality are 15.6, 19.3,
and 20.1 for Hong Kong, Chinese, and Other immigrants respectively. However, the negative entry effect
is too great to overcome for the immigrants from Hong Kong that arrived after 1989. Full assimilation
for these later cohorts is never achieved according to these estimates (i.e., there is no intersection with
the x-axis). This is a pessimistic result given the perceived success of Hong Kong immigrants overseas


                                                              17
≤ 1984 Cohort                                          1985−89 Cohort
           0.5                                                  0.5

             0                                                     0

          −0.5                                  HK              −0.5                                 HK
                                                China                                                China
            −1                                  Other            −1                                  Other
             0         5        10      15            20            0         5        10      15        20
                      Years Since Migration                                  Years Since Migration
                          1990−97 Cohort                                         ≥ 1998 Cohort
           0.5                                                  0.5

             0                                                     0

          −0.5                                  HK              −0.5                                 HK
                                                China                                                China
            −1                                  Other            −1                                  Other
             0         5        10      15            20            0         5        10      15        20
                      Years Since Migration                                  Years Since Migration


                     Figure 5: Immigrant Assimilation Profiles (Base group is natives (x-axis).)



during or after the handover transition period.
   However, this result may depend on the functional form chosen for the estimation equation. As
alluded to earlier, including the quadratic term on years since migration is likely to imply a lower
rate of assimilation as more time is spent in the host nation. If instead we only have a linear term
or we add a cubic term along with the quadratic, there would most likely be an intersection with
the x-axis (there would be no guarantee even with just the linear term because negative assimilation
is theoretically possible.) Hence, column 3 of Table 9 presents estimates excluding the years since
migration-squared variables. The initial earnings gap shrinks for all immigrant groups, while the earlier
finding of increasing negative cohort effects still holds. Moreover, the rankings of the assimilation profile
slopes across the immigrant groups remain unchanged. According to these estimates, for Hong Kong
immigrants arriving before 1985, earnings convergence with natives requires 19.4 years; for the 1985-1989
Cohort, 13.8 years; for the 1990-97 Cohort, 32.4 years; and for the ≥1998 cohort, 37.6 years.16 Thus, even
with this specification, the claim that Hong Kong immigrants represent a success story of immigration
is challenged.
   In the empirical specifications analyzed thus far, the cohort effect was captured by the initial wage gap.
The quality of immigrants, however, may also be judged based on their rate of earnings convergence with
natives in the host nation. Thus, I relax the assumption that the slope of the assimilation profile is equal
across cohorts within an immigrant category. Column 4 in Table 9 therefore includes the triple interaction
terms between years since migration, a cohort group, and an immigrant group. Most of the coefficients on
these terms are statistically significant, demonstrating that the rate of earnings convergence may differ
even when immigrants are of the same origin. With this specification, the numbers of years to achieve
equality with natives for immigrants from Hong Kong, in increasing order of cohort year, are respectively:
16.0, 12.8, 17.6, and 24.0. Although these figures differ from those computed for the earlier specifications,
if immigrant quality is summarized by this statistic, the conclusion of declining cohort quality after 1989
still holds. With differing assimilation profile slopes however, immigrant quality is no longer defined
simply by the entry wage gap. While it is true that the initial earnings gap for post-1989 immigrants
is significant, the growth rate of their wages is also correspondingly larger. This might suggest that the
recent immigrants adapted more quickly to the new environment than their predecessors.


                                                           18
From various studies (e.g., Friedberg (1993), Schaafsma and Sweetman (2001)), we know that the
factor of age at immigration may have an impact on the results. Since this the not the focus of the analysis
here, I only mention how this might bias the estimates. Children who enter the host nation at an early
age will most likely have an advantage in terms of learning the language, customs, etc. of the country
before entering the labor force. Therefore, if this is controlled for, the estimate of the rate of assimilation
may decrease. This would seemingly only affect the later stages of the assimilation profile though since
it takes many years for children to join the labor force. (Of course, age at immigration cannot simply
enter the regression equations specified since it would be collinear with years since migration and age.)
In fact, given many children of ethnically Chinese families have tertiary education before joining the
labor force to work, this bias can be somewhat downplayed.
   Another potential source of bias, perhaps one with more sizable impact, is non-random or selective
return migration. The analysis by Abramitzky et al. (2014) comparing assimilation outcomes of pooled
cross-section data with panel data gives an indication of the possible severity of this type of bias. Although
there is no absolute definition for what the length of stay must be for the immigrant to be considered
“permanent,” the performance of this type of immigrant is what one is arguably concerned with. If
there is negative selection of return migrants, the average quality of immigrant population increases,
thus producing a pattern that resembles assimilation. The rate of earnings convergence for permanent
immigrants using pooled cross-section data would then be overestimated. The opposite would be true of
positively selected return migration.
   Estimates of the number of Canadian citizens in Hong Kong today is large, within the vicinity of
250,000 to 500,000 (Sussman, 2011). Remigration and the phenomenon of “astronaut” parents shuttling
back and forth between Hong Kong and Canada were not uncommon (Skeldon, 1997, p.140). Many mi-
grants still had ties to Hong Kong and were ambiguous about their national identity as well. Moreover,
economic opportunities were better in Hong Kong for many of them (Salaff et al., 2008). Admittedly, a
qualitative theoretical prediction on the quality of return migrants is difficult to make. With anecdotal
evidence, one may make an argument for positive selection in remigration. Entrepreneurs and profession-
als who were successful in Hong Kong actually had difficulty in Canada due to the lack of social networks
and the non-transferability of overseas experience and qualifications (Salaff et al., 2010). Furthermore,
returnees were generally well-educated, bilingual, relatively young, and possessed considerable earnings
capacity (Ley and Kobayashi, 2005). If the argument of positive selection in return migration is in fact
true, it would imply that assimilation was underestimated and the performance of the immigrants would
have been more optimistic. Quantifying the magnitude of the selection bias would be extremely helpful
in further understanding the different assimilation profiles of the different cohorts. This is a topic left
for future research.


5.3.    Other labor market outcomes
While annual earnings is an informative measure of performance in the labor market, other outcomes such
as labor force participation and the number of weeks worked are useful indicators as well that capture
different margins of labor market activity. Before turning to such measures, column 1 of Table 10 serves
as another robustness check to the regression of Table 9 column 1, restricting the sample to persons
working 40 or more hours per week (in the week prior to Census enumeration). Thus, the results provide
evidence on the labor market assimilation of the subset of immigrants who obtained full time employment.
The sample of natives is restricted in the same manner. The estimates show that the initial wage gap
faced by the Hong Kong cohorts that arrived after 1997 is significantly reduced. The implied number of


                                                      19
You can also read
Next part ... Cancel