Per capita earnings in 2000 were highest in Ontario
($18,100) and lowest in Newfoundland and Labrador ($10,100). Across urban
and rural areas, per capita earnings were highest in large urban areas ($18,500)
and lowest in remote rural ($8,600). The $10,000 spread between
urban–rural is greater than the $8,000 spread across provinces
(Beckstead and Brown 2005).
Higher earnings in cities have also been linked to the proximity to human
capital, whereby proximity to college-educated workers increases earnings
while proximity to less than college-educated workers had the opposite effect
(Rosenthal and Strange 2008). Rauch (1993) found that each additional
year of metropolitan area education raises total factor productivity by 2.8%.
Firm size is not the only characteristic of firms
that influences wage levels. One of the more important characteristics of
firms in the Canadian context is foreign control. Foreign-owned firms may
operate at higher productivity levels and hence pay higher wages (Davies and
Lyons 1991). It remains unclear whether foreign-owned firms are more
likely to locate their operations in larger cities. Nonetheless, we will test
for their effect.
There is a consistent positive association
between firm size and wage levels (Idson and Oi 1999). As long as this
association is not driven entirely by the skilled composition of large firms,
part of the wage premium for larger cities may be due to this firm-size effect.
In this study we choose to account for nominal, rather than real (purchasing
power parity), earnings differences across the urban-rural spectrum. Real
earnings would be inappropriate because variation in the cost of living across
cities is largely determined by differences in shelter costs (see CANSIM Table 326-0015).
Higher shelter costs reflect, in part, the capitalization in land prices of
higher earnings, which in turn are driven by higher levels of productivity.
As such, a real income measure—that, by definition, takes into account
cost of living differences—would mask the effects of human capital and
agglomeration economies on earnings. Nominal earnings are used because they
capture the effect of higher productivity on earnings. Firms in larger cities
have to be more productive to pay higher wages. Their higher level of productivity
can come either from the ability to produce a higher volume of output per
worker or producing products that are more valuable to consumers.
It is unlikely that
daily commuting would take place beyond 200 kilometres.
The boundaries of some rural census divisions
overlap with boundaries of CMAs and CAs. For these, we only use their rural
Experience is defined as the age of each worker, minus six (an
approximate accounting for the pre-school years), minus their years of schooling.
The relatively lower mean weekly income
observed for new immigrants, especially those that immigrated to Canada in
the early 1990s, can be partially explained by declining returns to foreign
labour market experience, which occurred exclusively among immigrants from
non-traditional source countries (Aydemir and Skuterud 2004).
Large plants are defined as those with more
than 200 workers.
To illustrate, consider
two managers in the banking industry with the same number of years of education
and experience: one works at a senior level at the head office in Toronto,
while the other works in a regional office. As the model is specified, both
managers will be treated the same. However, the manager in Toronto likely
earns more, because of his/her higher position within the organization—a
position that they may have obtained by means of unobservable personal traits—e.g.,
superior analytical skills.
Moving costs will likely increase
with earnings, but this increase should be relatively small compared with
the greater (absolute) gains in earnings for those with higher earnings.
Fu and Ross (2007) provide an innovative solution
to this problem by arguing workers with similar skills and abilities—both
observed and unobserved—sort into the same neighbourhoods and therein
provide a means to control the effect of skills and abilities on wage levels.
Although this assumption may be appropriate for large urban areas, it would
likely not hold for the smaller cities and rural areas that form a significant
proportion of our dataset.
Kamloops had a population of 86,000 in 2001 (2001 Census
Note that the market potential of the location j does not include itself.
We would have preferred to have used employment for consistency
with our other measure of market size, but data on U.S. counties were more
readily available for population only. As the correlation coefficient between
employment and population for the Canadian census metropolitan areas and rural
census divisions was 0.99, population and employment can be treated as
Combes et al. (2007) obtain a similar but statistically significant point
estimate. However, some caution should be used comparing these estimates because
they use employment density instead of size to measure market potential.
and Sereda (2007) for a definition of these occupations.
Glaeser and Gottlieb's (2006) discussion
is with regards to real earnings. Nevertheless, nominal earnings may still
be influenced by amenities.