Printer Friendly

The changing geography of the Canadian manufacturing sector in metropolitan and rural regions, 1976-1997.

Introduction

It is often argued that new technologies and improvements in transportation are making it easier to decentralize manufacturing away from the centres of large metropolitan centres to their fringes (Coffey 1994) or away from large cities to smaller urban centres and rural regions (Kilkenny 1998). Alternatively, it is argued that there may be strong centralizing tendencies associated with the advantages of large labour pools and advantageous interactions that take place in large cities, as well as the benefits that their large markets provide to producers (Krugman 1991; Fujita et al. 1999).

Despite this interest in the changing importance of manufacturing in cities and rural regions, we know surprisingly little about the changing location of manufacturing production in Canada. This paper fills this gap in out knowledge by tracking the actual course of economic development in the manufacturing sector of Canadian cities over the last quarter century. Using longitudinal data on output and wages along with a specially constructed fixed geographic code, this paper provides a picture of changes in the manufacturing sector that have occurred across various units in the rural/urban hierarchy--large central cities, their suburban fringes, medium and small cities and rural areas.

The objective of this paper is to measure not only shifts in aggregate manufacturing employment across the rural/urban hierarchy, but also the industrial composition of these changes. That is, we are interested in whether these shifts are broad-based, encompassing a wide selection of industries, or whether they are driven by a small selection of industries. Broad-based change may reflect structural shifts in the economy that favour one or more parts of the rural/urban hierarchy over others. For example, falling transportation and communication costs can make rural parts of the country a more attractive location for a broad selection of manufacturing industries (Kilkenny 1998). On the other hand, change may be more industry-specific. For example, large urban regions may provide the access to skilled workers and specialized suppliers necessary to attract and keep more knowledge-based manufacturing industries.

Aggregate employment levels are only one measure of economic performance. Production worker wages offer an alternate and important measure because of their relationship to the economic base (1) provided by industry. There is evidence that wages decline substantially moving down the rural/urban hierarchy. For example, in the United States, production worker wages in rural areas are markedly lower than those paid to workers in urbanized places (Gale 1997, 1998). In this study, we are interested in whether there are substantial differences in wage levels across the rural/urban hierarchy in Canada and whether we observe a similar urban-rural wage gradient seen in the United States.

We should add that the analysis concentrates on the manufacturing sector because of its central importance in the economic system. While manufacturing only accounts for about 20 percent of total employment, it is still one of the largest sectors in most areas. (2) As of 1986 in Canada, the percentage of the rural labour force in manufacturing was only slightly behind the percent in the primary sector (Ehrensaft and Beeman 1992). By 1992 in the United States, manufacturing had replaced agriculture as the primary economic base for much of the rural Midwest (Bernat 1997). Therefore, even in predominantly rural areas, manufacturing is a significant source of employment.

The remainder of the paper is organized as follows. Initially, we review and place into empirical context several formal and informal theories that attempt to explain shifts in the location of manufacturing employment and wage differences across metropolitan and rural regions. A description of the methodology for defining the rural/ urban hierarchy is then presented. This is followed by a discussion of the results, which focuses on trends in employment shares and wages across the Canadian rural/urban hierarchy over time, and by a brief conclusion.

Theoretical and Empirical Context

This paper attempts to measure the performance of urban and rural manufacturing economies using employment and production worker wages as indicators of their health. The purpose of this section is to place these performance measures within a broader theoretical and ing the theoretical discussion are two questions. First, why would we expect industries to shift from the centres of large cities to their suburbs and/or to smaller urban and rural regions? Second, why would wages systematically vary depending on a region's place within the rural/urban hierarchy?

Diffusion and dispersion of manufacturing employment

In the empirical analysis to follow, we measure shifts in manufacturing employment from the centres of large metropolitan areas to their fringes and down the rural/urban hierarchy. Slater (1961) refers to the former as diffusion, or suburbanization (shifts in employment from the centres of large cities to their suburban peripheries), and the latter as dispersion (shifts in employment from larger to smaller cities and, in the context of this study, rural areas). The distinction between diffusion and dispersion is useful because different economic processes likely underlie both patterns.

There are two views on whether the diffusion of manufacturing employment in large urban centres has led or lagged the movement of the population to the suburbs. One holds that manufacturing employment has tended to be concentrated in the cores of large urban centres (Slater 1961; Pred 1964; Scott 1982) and that only after the development of the suburbs did manufacturing activity more to the peripheries of these large cities (see Slater 1961). The second and more recent view, articulated by Walker and Lewis (2001), is that industry has undergone a continuous process of suburbanization since the middle of the 19th century: industry has not followed workers to the suburbs, but has moved simultaneously with them.

Regardless of whether industry followed the population to the urban fringe or moved concurrently with them, it is apparent that the centres of urban areas no longer attract large quantities of manufacturing investment. This suggests that the advantages that the present cores of Canada's large cities might have had in the late 19th and early 20th centuries are no longer relevant. There are several reasons why the old centres of cities are no longer attractive locations for manufacturing production. First, the development of automobiles and truck freight transport has made industries less tied to central urban locations with railway or water access. Second, the adoption of mass-production techniques by manufacturers favours production in single-story buildings on large sites (Slater 1961; Pred 1964; Scott 1982; Coffey 1994). It is suburban areas in which access to highway transport is best and in which large tracts of affordable land are available for new single-story factories, giving suburbs a clear advantage over central cities.

The third reason, as noted above, is that workers have also moved to the suburbs in North America (Walker and Lewis 2001). The old centres of cities no longer provide the only large pool of workers. The benefits of potentially less militant labour in suburban areas of cities and the active promotion by land developers and suburban governments of suburban industrial development may also be factors driving the movement of manufacturing employment to the suburbs (ibid.). Taken as a whole, all of the factors identified as driving location of industries within cities tend to push industries towards the suburbs and away from the cores of large urban areas. All forces seem to be pushing in the direction of diffusion.

Of course, the choice of where to locate a new plant is not limited to the intrametropolitan scale. As circumstances change, firms may choose to locate in smaller urban or in rural regions rather than in large metropolitan centres. Yet, unlike the forces behind diffusion, which all clearly push in the same direction, the same is not true of those factors thought to influence dispersion.

Most theoretical treatments of the forces driving the dispersion or concentration of industry focus on the influence of declining transportation costs. Krugman (1991) has argued that the declining cost of transporting industrial goods will tend to concentrate industries in urban areas in order to take advantage of increasing returns to scale while minimizing transportation costs. In other words, falling transportation costs unambiguously lead to the concentration of industry in cities. (3) On the other hand, Kilkenny (1998) argues that there is a nonlinear relationship between declining transportation costs and the concentration/dispersion of industry. She demonstrates in her model that as industrial goods transportation costs fall relative to agricultural goods, (4) industries tend to concentrate in urban regions. However, unlike Krugman's model, (5) Kilkenny's shows that below some threshold level, manufacturers begin shifting their production back to rural areas. Using a similar model, Fujita and colleagues (1999) obtain the same result. Therefore, based on theory, it is unclear whether we would expect industry to be shifting away from urban to rural regions or vice versa as transportation and other distance-related costs fall.

Taken as a whole, the theory on diffusion and dispersion of manufacturing activity suggests that there will be a continued trend towards more suburban employment in large cities, but there is no clear theoretical expectation regarding the movement of industrial production to small urban and rural regions. Only empirical evidence can shed further light on whether manufacturing is dispersing. It is to the empirical evidence of dispersion (and diffusion) of manufacturing employment in Canada and the United States that we now turn.

In the United States, the share of manufacturing employment in older metropolitan areas has declined consistently since the end of World War II. The fringes of older metropolitan areas, new metropolitan areas (those counties more recently classified as urban) and rural areas have all gained shares (Nucci and Long 1996, 1997). In short, the US experience provides evidence of both diffusion and dispersion.

At the urban and rural scales of analysis in Canada, Slater (1961), using the equivalent of the current 'Annual Survey of Manufacturers', has found evidence of the dispersion of manufacturing activity between 1932 and 1956, although much of this has been concentrated in Ontario. He also finds strong evidence that manufacturing employment in large industrial centres (Toronto and Montreal) has diffused from their cotes to their suburban peripheries. More recent analyses also find evidence of the diffusion and dispersion of manufacturing employment. Using employment data from Dunn and Bradstreet, Coffey (1994) finds strong evidence that employment grew more rapidly between 1981 and 1991 in the outskirts of Montreal than in its core areas (i.e., the Island of Montreal). Unfortunately, it is unclear from Coffey's analysis whether manufacturing has followed the same trend as overall employment. (6) The Census of Population also provides some evidence that the place of residence of workers in manufacturing shifted from larger metropolitan to rural regions between 1971 and 1981 (Coffey and Polese 1988; Coffey 1994). Therefore, there is evidence of patterns of diffusion and dispersion of manufacturing employment in Canada similar to those in the United States, but with the exception of Slater's analysis, this evidence uses different sources of data (i.e., Dunn and Bradstreet versus the census) and comes mainly from place-of-residence data, not from information on the location of activity of the employer.

This paper contributes to the empirical literature on the geography of manufacturing location in Canada by providing a comprehensive and consistent view of the changing patterns of manufacturing location in large urban centres and across the whole rural/urban hierarchy, using data on plants' locations of employment. In so doing, it provides us with a unique perspective on the extent of both diffusion and dispersion over the past quarter-century.

Wage differences across the rural/urban hierarchy

Wage levels serve as an additional performance indicator for regional manufacturing economies that may vary across rural and urban areas. Gale (1997) has found that there is a strong positive hierarchical relationship between nominal production-worker wage levels and urban size in the United States. The reason that wage levels are positively related to urban size remains unclear; it may be due to differences in labour market conditions across rural and urban areas and/or industrial structure.

There are several factors that may affect local labour market conditions. First, in smaller urban and rural areas, manufacturers may be operating as monopsonil in the local labour market. This would allow them to pay lower wages than firms in more competitive, larger urban labour markets. Second, the nominal wages paid to workers in urban areas may also be higher because of higher costs of living in urban areas (see Kilkenny 1998). Finally, employers may be able to afford to pay these higher wages because of stronger urbanization (Jacobs 1969) and localization economies (Marshall 1920; Krugman 1991) in larger urban centres (see Glaeser et al 1995). (7) Industrial structure may also play a role if industries that tend to pay higher wages also tend to be found in large urban centres. There is evidence that suggest this is the case. In the United States, industrial structure partially explains wage differences across rural and urban areas (Gale 1997).

With the possible exception of industrial structure, all of the other factors hypothesized to influence wage levels suggest nominal wages will tend to increase with the size of region. Industrial structure is the exception because it is unclear whether variations in this structure will contribute to the urban/rural differential. It may be that the types of industries that benefit most from agglomeration economies that are present in larger cities are those that require the highest-skilled workers and therefore pay the highest wages. However, there are two factors that may run counter to this argument. First, urban labour markets in Canada often have large pools of immigrant labour that, because of lower skills and/or discrimination, may be forced to accept low wages. Second, many capital-intensive, high-wage industries (e.g., pulp and paper producers) may locate in rural regions because they require access to raw-material resources. The implication is that if there are substantial differences between the structure of demand (industrial makeup) or supply (urban labour markets) in Canada and in the United States, we may not observe the same rural/urban wage gradient in Canada. As will become apparent in the analysis to follow, this is the case.

Methodology

Measuring the Canadian rural/urban hierarchy: Beale codes

In this paper, we ask, in part, whether smaller cities or rural areas have become more important over time because of the gradual shift of industry out of large metropolitan centres and down the rural/urban hierarchy. This requires a rural/ urban classification system that can be consistently applied over time. For this purpose, we use a modified version of the Beale rural-urban coding system that was originally developed by the US Department of Agriculture to identify the 'location' of counties within the rural-urban continuum, or what might be roughly considered the rural/urban hierarchy (see GAO 1989; Butler 1994). (8) Out Beale coding system consists of 11 categories of rural and urban places and uses census divisions, which are largely equivalent to US counties, as its base geographic unit. (9) For the purpose of this paper, we have collapsed these 11 categories into six. We have done so for ease of exposition and because many of the rural classifications include few plants, which would have resulted in the suppression of some of our results to preserve the confidentiality of respondents. The six categories are summarized in Table 1.

The Beale coding system classifies census divisions based on their relationship to the Canadian rural/urban hierarchy as defined by the size of census metropolitan areas (CMAs) and census agglomerations (CAs). The census divisions are classified first by whether they belong to a metropolitan area and then by the population of that metropolitan area. Outside the metropolitan area, they are classified on the basis of their location relative to metropolitan regions (i.e., Nonmetro-Adjacent versus Nonmetro-Nonadjacent; see Table 1). Therefore, the Beale coding system contains both hierarchical (size) and geographic (location) components. It provides us with a perspective on the influence of location and position within the rural/urban hierarchy on industrial change.

Figure 1 illustrates the geographic pattern of the Beale-coded census divisions for 1976. There were three Large-Metro-classified census divisions in that year: Montreal, Toronto and Vancouver. Associated with each of these are census divisions that overlap or are encompassed within their CMA boundaries, the Large Metro Fringe. More numerous are Medium Metro and Small Metro areas, which are found in all provinces except for Prince Edward Island. Medium Metro areas include cities such as Halifax, Ottawa and Calgary. Small Metro areas include cities such as Fredericton (New Brunswick), Kingston (Ontario) and Kelowna (British Columbia). Often bordering metropolitan areas are rural Nonmetro-Adjacent-classified census divisions. Note, however, that census divisions are classified Nonmetro-Adjacent only if they border on a CMA or CA boundaries-hence the large number of cases in which metropolitan-classified census divisions border on Nonmetro-Nonadjacent census divisions (see Figure 1). Nonmetro-Nonadjacent areas are the most common census divisions; they cover most of Canada's landmass.

[FIGURE 1 OMITTED]

Initially, census divisions were given Beale codes based on their population characteristics and relative locations for each of the census years 1976, 1981, 1986 and 1991. (10) As towns or smaller metro areas grow larger, their Beale codes change over time. In effect, census divisions more up the rural/urban hierarchy. However, for some types of longitudinal analysis, this reclassification can be a problem.

The primary objective of this paper is to measure change in the location and composition of industry through time. Therefore for out purposes it is important to be able to distinguish between two forces that have been behind the change in manufacturing activity across Canada's rural/ urban hierarchy: change that results from the growth or decline of industry; and change that results from the reclassification of census divisions. Allowing the classification of census divisions to change hampers the interpretation of shifts in the importance of economic activity across different geographic units because of resulting discontinuities at the census years. It means that changes in the importance of manufacturing in a particular area can be caused either by inherent growth in that region or by reclassification. For example, growth in Metro Fringe areas can be caused either by the fact that industry in these areas was inherently more dynamic or because smaller metro or rural areas were being reclassified or absorbed into metropolitan areas. Since we want to know, for example, the extent to which industry that was in rural regions at the beginning of the period grew more or less quickly than industry that was in larger centres, we need to remove the effect of reclassification on our measures of changing activity. To do so, it is important that we be able to hold the original classification constant since 1976--the start of our study.

To hold the classification of census divisions constant through time, two problems had to be resolved. First, through the study period, many census divisions across Canada grew in size, resulting in their reclassification and, at times, that of their neighbours. (11) Second, the boundaries of census divisions have not been constant over time. This was particularly true of census divisions in Quebec, which were completely redrawn in 1991.

We have taken two steps to overcome these difficulties. First, for those census divisions whose borders have not changed over time, we have maintained a constant 1976 Beale classification throughout the study period. This eliminates the reclassification problem. When a census division's borders changed, we followed a more complicated procedure. In cases where census divisions were split, we combined them to match their 1976 boundaries. In the few instances where census divisions were amalgamated, we adopted the new boundaries rather than those from 1976. In many instances, census division boundaries were completely redrawn, which made it impossible to recombine them to recreate earlier or later census geographies. To address this problem, plants were assigned point locations using postal codes. The point locations, in turn, were used to allocate plants to the 1976 census geography. A detailed discussion of the procedures we used to maintain a constant geography over time can be round in Baldwin and Brown (2001).

Results

At issue is the extent to which there have been substantial shifts in the location of production across different levels of the rural/urban hierarchy over the last 22 years. In particular, is there the same loss of importance in large urban cores in Canada as in the United States? If so, has the deterioration in the relative size of these central urban centres been matched by corresponding growth in the major urban suburbs, or has the decline been associated with growth of small metropolitan and rural areas? We address these questions by observing changes in manufacturing employment across the rural/urban hierarchy.

We divide this subsection of our paper into three parts. The first reviews the broad trends in employment and employment shares experienced by Canada's urban and rural areas. Since national trends can mask significant regional differences, the second section looks into employment changes across the rural/urban hierarchy in Canada's two most important manufacturing regions, Ontario and Quebec. In the third part of this section, employment is broken down by industrial sector. Here, we wish to determine whether shifting employment shares are driven by a few sectors or whether we are observing across-the-board shifts in manufacturing employment.

National employment trends

We analyze employment trends in two ways: first, by reporting both their levels and their shares of employment across the Beale categories over time (Table 2), and second, by testing whether employment-share trends are statistically significant (Table 3). We test for two types of trends, linear and nonlinear (quadratic). If the trend is linear, the regression coefficient on the TREND (12) variable and its level of statistical significance is reported. However, if the trend is nonlinear, we report the slopes and statistical significance for two variables, TREND and TRENDSQ (13) (see Table 3). We define nonlinearity narrowly as only those cases in which the coefficients for TREND and TRENDSQ are significant and take on opposite signs. A negative parameter for TREND and a positive one for TRENDSQ indicate that, although the Beale category's employment share may have fallen early in the period, this trend had become less negative over time and might have reversed itself by the end of the period. On the other hand, if TREND is positive and TRENDSQ is negative, the Beale category's share may have increased early on, but this trend had slowed or been reversed by the end of the study period. It should be emphasized that any trend reversal seen in the data may be confirmed by TREND and TRENDSQ taking on opposite signs, but such a reversal cannot be identified solely by observing the parameter estimates of TREND and TRENDSQ; such a result may simply reflect a slowing trend. Finally, autocorrelation has been identified as a problem with the data. We have corrected for autocorrelation by estimating the linear and nonlinear models specified in endnotes 12 and 13 using the Prais-Winsten estimator (14) (see Gujarati 1995). Throughout the rest of the paper, we employ this methodology to test the statistical significance of trends.

Between 1976 and 1997, the most dramatic change in employment share occurred in the cotes of large metropolitan areas and their suburban fringes. The employment share of Large Metro areas fell over the period, while the Large Metro Fringe increased its share (see Table 2). Both regions' trends were linear and significant (see Table 3). Increasing employment in their suburban fringes roughly compensated for the absolute and relative decline of large metropolitan core areas. Taken together, the Large Metro's and Large Metro Fringe's shares of employment changed little over the period. Therefore, although there have been major shifts within Canada's largest metropolitan regions, from their cores to their suburbs, taken together these regions have not experienced a serious decline in their share of manufacturing employment relative to smaller urban and rural areas. (15)

For most of the smaller urban and rural classifications, we also observe significant, albeit smaller, shifts in employment (see Table 2). The share of employment in Small Metro areas fell through most of the study period, in both relative and absolute terms. On the other hand, the two rural categories gained employment between 1976 and 1997 (see Tables 2 and 3), resulting in a 10-percent increase in the rural share of employment. This is a modest share increase and one that is roughly comparable to that experienced by the United States over a similar period of rime (Nucci and Long 1996, 1997)

Although there was a general shift towards rural employment, the rural categories did not follow the same trend: Nonmetro-Adjacent areas consistently increased their employment share, while significant shifts in employment towards Nonmetro-Nonadjacent did not occur until the mid- to late 1990s (see Tables 2 and 3). Therefore, although there has been an apparent shift towards rural manufacturing employment, this trend has been strongest in rural regions that are in the shadow of metropolitan areas. Employment may be moving out of urban regions, but it is not moving far away.

Regional employment trends

In the United States, Nucci and Long (1996, 1997) have found that shifts in manufacturing production across the rural/urban hierarchy vary considerably depending on the region under study. Slater (1961) has found it to be the same case for Canada over the period between 1933 and 1956. Thus, the national shifts in employment and employment shares across the rural/urban hierarchy that we have documented to this point may obscure significant regional differences. To explore this possibility, we compare employment trends in Ontario and Quebec. We focus on these provinces for two reasons. First, Ontario and Quebec together account for three-quarters of Canada's manufacturing employment (Statistics Canada 1999), and therefore have a strong influence on the estimated national average. Second, Ontario and Quebec have the most diverse urban structures of all the provinces (see Figure 1), which permits us to compare trends across all rural/urban classifications.

Employment trends by Beale classification are reported for both provinces in Table 4. In the table, we break the study period into three shorter periods that roughly correspond to the business cycles the Canadian economy has experienced since 1976: 1976-1980, 1981-1989, and 1990-1997. For each period, the average level of employment and its share of average total employment are reported by Beale category. We also report employment levels for the beginning and end years of the study period.

There are substantial differences between the rural/urban hierarchies in Ontario and Quebec in terms of the weights of their components and their underlying dynamics. The most apparent difference is the top-heavy nature of Quebec's rural/ urban hierarchy compared to that of Ontario. In both provinces, there is one Large Metro classified census division (see Figure 1). In 1976, the Large Metro core of Toronto accounted for 31 percent of manufacturing employment in Ontario, while the Large Metro core of Montreal accounted for about 49 percent of employment in Quebec (see Table 4). Although both trend downward over time, the decline is some 10 percentage points in Quebec and only about 5 percentage points in Ontario. The Island of Montreal has experienced a much more dramatic relative decline than the central core of Toronto. Moreover, the relative decline of Montreal's urban cote has been steady throughout the study period. Toronto's core experienced a falling share of employment only in the 1990s (see Table 4). In contrast to their urban cores, the fringes of Toronto and Montreal grew both in absolute and relative terms, but the Large Metro Fringe in Ontario increased its share of employment by 10 percent, while it increased by only 4 percent in Quebec between 1976 and 1997 (see Table 4).

The employment trends experienced by the Large Metro and Large Metro Fringe in Ontario and Quebec reflect the differing fortunes of the Toronto and Montreal metropolitan regions. That is, when added together, the Large Metro and Large Metro Fringe census divisions in both provinces form what we might call the Toronto and Montreal 'city-regions'. (16) The Toronto city-region share of employment in Ontario increased over the study period (see Figure 2). Toronto's gains through the 1970s and 1980s were at the expense of all other rural/urban categories except Nonmetro-Adjacent (see Table 4). Unlike Toronto, Montreal's share of employment fell consistently over the period (see Figure 2). Montreal's falling share was the result of declining employment in the Montreal region and increasing employment in rural regions of Quebec (see Table 4). Combined, the Nonmetro-Adjacent and Nonmetro-Nonadjacent categories increased their employment shares in Quebec by 9 percent between 1976 and 1997. In Ontario, these rural regions' shares remained essentially static.

[FIGURE 2 OMITTED]

Finally, we should note that our results for Quebec and Ontario contrast sharply with Slater's (1961) findings. (17) He found that for the period from 1932 to 1956, Toronto region's share of employment declined, while the share of employment in Ontario's smallest (rural) manufacturing counties increased. In Quebec, he found that the Montreal region maintained its share of provincial employment over the same period. The contrasting patterns of dispersion across regions and over time suggest that these patterns, although undoubtedly influenced by declining transportation costs, cannot be fully explained by them.

Why Montreal and Toronto have followed such different paths since 1976 is beyond the scope of this paper. Vinodrai (2001) analyzes in greater detail the changing industrial structures of Montreal and Toronto, as well as those of Vancouver. She finds that one reason why Montreal and Toronto followed such different trajectories lies in the difference in their industrial structures. Montreal has experienced large job losses in labour-intensive industries such as clothing and textiles, while growing employment in industries such as aerospace have not been enough to compensate. Toronto, on the other hand, was not as specialized in the labour-intensive industries. Differences in the industrial structures of Toronto and Montreal do not explain the whole story, however. Often, Montreal experienced declining employment and Toronto experienced the opposite for the same industries, which is an indication that Toronto was more successful at attracting investment capital.

Changes in industrial structure

As we have demonstrated above, measuring changes in the location of manufacturing employment on a national basis may hide considerable regional differences. Similarly, only looking at changes in aggregate manufacturing employment may mask variation in the types of industries round in rural and urban places over time--that is, changes in the industrial structure of rural and urban regions. In particular, we are interested in knowing whether employment shifts across the rural/urban hierarchy are driven by a broad spectrum of manufacturing industries or whether they are being driven by just a few industries. A broad-based shift suggests a structural change that favours one part of the rural/urban hierarchy. On the other hand, a narrower, industry-specific shift means it might be more appropriate to study the causes of the growth (or decline) of particular industries to understand the reasons for the growth of certain regions.

In order to study whether there are underlying differences in the broad trends that we have observed so far, we break the manufacturing sector down into five subsectors: (18) natural resource-based, labour-intensive, scale-based, product-differentiated and science-based. (19) The five groups are distinguished on the basis of the primary factors affecting the competitive process in each sector. For the natural resource-based sector, the primary factor affecting the competitive process is access to natural resources. These are industries in which the ratio of value added to materials inputs is small, because there is little value added beyond the raw materials stage. For the labour-intensive sector, the primary factor is labour costs. These industries pay relatively low wages. For the scale-based sector, the primary factor is the existence of scale economies. These are industries that are capital-intensive. The include both iron and steel, which are concentrated in urban areas, as well as forest industries which are based in rural areas. For product-differentiated industries, the primary factor is the ability to tailor production to highly varied demand conditions. These tend to be industries with higher advertising/sales ratios. For science based industries, the primary factor is the rapid application of scientific advances. These tend to be industries with higher research-and-development/sales ratios.

We summarize employment trends for these five industries in Table 5. As in Table 4, we break the study period down into shorter periods that correspond to the business cycle, and we test the trends for statistical significance. For each period, the average level of employment and its share of average total employment are reported by Beale category and industry. To provide a basis of comparison, the period averages are also reported for all industries together.

Are we observing a broad-based shift in manufacturing employment, or is it more industry--or sector-specific? Clearly, there has been an across-the-board shift in employment away from Large Metro areas and towards the Large Metro Fringe. Only in science-based industries were Large Metro areas able to increase employment in absolute terms (see Table 5). Employment change in the other rural/urban categories tended to be more idiosyncratic. For example, Medium Metro areas have increased their share of employment in labour-intensive and science-based industries, but have reduced their share of scale-based employment.

Overall, it is apparent that within large metropolitan regions, production in every industrial sector has shifted towards their suburban fringes. Therefore, the forces driving diffusion--for example, access to suburban labour pools and lower land prices--applies across all industries. Outside of these large centres, the pattern is more complicated. No general industry shifts can be identified as driving the relative rise or decline of smaller urban and rural Beale categories. Furthermore, for most industries there is no apparent shift up or down the rural/urban hierarchy. The one exception is science-based industries, which have concentrated over time in the top half of the rural/urban hierarchy (see Table 5). Therefore, dispersion does not appear to be a broad-based process. Declining transportation and other distance-related costs over the study period have not resulted in any clear patterns of dispersion or concentration across the rural/urban hierarchy. The factors at work here are industry-specific.

To summarize the results of this section, employment has shifted away from the Large Metro cores and towards the fringes of these metropolitan regions and rural areas of the country. This general shift parallels trends in the United States (Nucci and Long 1996, 1997) and is consistent with broader intrametropolitan employment in Canada (c.f. Coffey 1994). Like Nucci and Long, we find that trends in the location of manufacturing employment vary depending on the region in question. Specifically, employment in Ontario has tended to concentrate in the Toronto urban region, especially its urban fringe. In Quebec, the Montreal urban region has experienced a falling share of provincial employment. It is the rural regions of Quebec that have been the most dynamic. Finally, breaking employment down by manufacturing sector makes it apparent that there has been a broad shift in manufacturing production across many industries, away from the centres of large metropolitan areas and to their fringes. The pattern for other rural/urban regions is not as clear. The pattern of change depends on the industry and the rural/urban classification in question.

Variations in wage across the rural/urban hierarchy

Size, as measured by employment or employment shares, provides us with one measure of performance. Wages provide an additional measure of performance, and one that has been round to vary significantly across the rural/urban hierarchy in the United States (Gale 1997, 1998). In order to examine the difference in relative wages paid across rural and urban areas, we compare wage rates for each Beale category to the national wage (see Table 6). (20) Differences in relative wage rates across regions will reflect, inter alia, differences in industry composition, differences in relative skill levels and differences in the relative demand for workers.

The relative wages of production workers vary significantly depending on their location within the rural/urban hierarchy. The spread between the highest and lowest relative wages by Beale category averaged 17 percent between 1976 and 1997, ranging from a maximum of 22 percent to a low of 12 percent. Large Metro and Nonmetro-Adjacent areas tended to have below-average wage levels throughout the study period. Small Metro had the highest wage levels, and Medium Metro and Nonmetro-Nonadjacent tended to have wages that were above the national average. These results are quite different from those reported for the United States. Gale (1997) reports that, in 1992, US core metro areas paid the highest hourly wage and rural, nonmetropolitan areas paid from 12 to 20 percent less. In Canada, wages are lowest in the cores of large metropolitan areas and are near the national average in rural parts of the country.

Also reported in Table 6 are relative wage trends over time, which also vary across rural and urban areas. Of note are the wage trends for Large Metro and Nonmetro-Adjacent areas. Relative wages in the Large Metro cores fell consistently over the period, drifting further below the national average.

The opposite was true of Nonmetro-Adjacent rural areas, which increased their relative wages levels from 9 percent below the national average at the beginning of the period to about the national average by the end. At least in these two categories, wage trends have tended to match employment trends.

The differences in wage levels and trends across urban/rural regions may reflect a variety of different factors. It may be because a region generally attracts industries that pay higher wages, or it may be because of a higher wage structure for all industries in that region. Figure 3 illustrates the significant impact that industrial structure can have on wage levels. It shows wage levels across Beale categories and industries relative to the aggregate national wage, averaged over the study period. What is apparent is that most of the variation in wages is across the industry categories, rather than the Beale categories.

[FIGURE 3 OMITTED]

To investigate the extent to which industry structure affects wage differences, we graph the average relative wage rate for the period across the Beale categories and the average that is corrected for industry structure in Figure 4. The latter is obtained by first calculating the relative wage rate for each industry across Beale codes, using the five-sector definition, and then averaging the result for each Beale code across all industry sectors. When this correction is made, the rural/urban wage profile changes dramatically. The four urban areas now have relative average wages that are slightly above 1 and are very similar one to another. The two rural areas have relative average wages that are below 1. Thus, once industry characteristics are taken into account, a difference emerges between the group of urban areas and the two rural areas, one that more closely resemble the differences that exist in the United States (see Gale 1997, 1998).

[FIGURE 4 OMITTED]

Relative wage changes are also muted when the effect of industry is taken into account. In Figure 5, we graph the 'corrected' mean relative wage at the beginning and at the end of the study period. The central cote of the urban area still declines, but the fringe increases over time, rather than decreasing. Both the Medium and the Small Metro areas show a slight decline in their relative wage rate, but the changes are not large. The rural adjacent area still experiences improvements in its relative wage rate. In summary, correcting for industry mix shows that the urban area that was gaining manufacturing employment over this period, the Large Metro Fringe, also saw an increase in its relative wage compared to the central core area.

FIGURE 4 OMITTED]

The analysis of production-worker wages shows us that wages in Canada, unlike those in the United States, do not decline as we more down the urban/rural hierarchy. Only when differences in industrial structure are taken into account are wages lower in rural areas than in urban areas. More importantly, there has been little change in the wage differences between urban and rural centres over the period once industrial structure is taken into account. This result explains, at least in part, why we see very little dispersion of employment from large metropolitan regions down the rural/urban hierarchy.

Conclusion

Between 1976 and 1997, there was a shift in the share of manufacturing employment sector away from the central areas of large metropolitan regions towards their suburban fringes. This trend was broad-based, encompassing most sectors, and was consistent with several factors hypothesized to drive the continued outward movement of industry--for example, access to labour and transportation infrastructure and lower land costs.

The diffusion of employment in these large metropolitan regions was not accompanied by a large dispersion of employment towards those places further down the rural/urban hierarchy, nor was the dispersion of employment driven by a broad selection of industries or one industry. Of the two kinds of rural regions, the Nonmetro-Adjacent areas increased their share the most. Therefore, as was the case in the United States, there is some evidence of dispersion, but this phenomenon has been largely limited to those places in the shadow of metropolitan regions.

Wages in the analysis were used as an additional measure of economic performance that speaks to the quality of jobs and the economic base provided by manufacturing industries in rural and urban areas of Canada. In contrast to the United States, rural regions of Canada tended to have production worker wages that were at or near the national average. At the top of the rural/urban hierarchy, wages were lower. Most of these variations were due to differences in industrial structures between rural and urban regions. Once the differences in structure were taken into account, wages tended to be higher in urban areas and lower in rural areas. This is consistent with the pattern observed in the United States.

These results suggest that the industrial structure of large urban areas and rural regions in Canada and the United States may be different. The large presence of the pulp and paper industry, which tends to pay above-average wages, in rural parts of Canada may partially explain the fact that Canadian rural wages tend to be higher. It might also be that labour-intensive industries in Canada have tended to remain in large cities, possibly because they may have access to large pools of lower-cost labour in these centres. In the United States, it could be that labour-intensive industries have moved to more rural regions, in particular the rural South. In Canada, there is likely no equivalent to such a large pool of rural labour.

In closing, there are at least two implications that might be drawn from the analysis. First, this research points to the importance of industrial structure as a significant determinant of wage differences across regions. This finding re-emphasizes the significance of understanding the underlying cause of firms' location decisions because of the welfare consequences that follow. From a policy perspective, this finding also suggests that efforts to reduced earned-income disparities across regions cannot ignore the importance of industrial structure. Second, the analysis shows that beyond the intrametropolitan scale, we lire in a relatively static environment. That is, manufacturing industry has tended to remain in large urban regions. This suggests that the forces that might lead to dispersion of industry, such as falling transportation costs, are not able to overwhelm the benefits of access to large urban markets. The tendency of industry to locate in rural census divisions adjacent to metropolitan regions is evidence of this effect.
Table 1

Description Beale coding system

Code    Name                    Description

0       Large Metro             Central and most populous census divi-
                                  sion of a CMA with a population
                                  greater than one million.
1       Large Metro fringe      Remaining census division(s) within or
                                  partially within a CMA with a popula-
                                  tion grater than one million
2       Medium Metro            Census division(s) containing, par-
                                  tially within a CMA with a population
                                  between 250,000 and 999,999
3       Small Metro             Census division(s) containing, within
                                  or partially within a CMA/CA with a
                                  population between 50,000 and 249,999
4       Nonmetro-Adjacent       Census division that share  a boundary
                                  with a CMA/CA that has a population
                                  greater than 50,000
5       Nonmetro-Nonadjacent    Census division that do not share a
                                  boundary with a CMA/CA that has a
                                  population greater than 50,000

NOTE: Because CMA and CA boundaries are different than census-division
boundaries, census divisions may: (1) contain entire CMA/Cas; (2) be
found completely within CMA/CA boundaries; or (3) be only partially
within the territory of a CMA or CA. In all cases, the census division
is classified using the Beale code that is associated with the size of
that CMA/CA.

Table 2

Total manufacturing employment by Beale code, 1976-1997

                         Employment (Percent Share)

        Large Metro     Large Metro        Medium          Small
            (0)          Fringe (1)       Metro (2)      Metro (3)

1976    595,746 (34)    175,625 (10)    324,465 (19)    261,090 (15)
1977    571,773 (34)    173,098 (10)    317,075 (19)    263,130 (15)
1978    596,521 (33)    186,463 (10)    330.631 (19)    267,737 (15)
1979    614,004 (33)    200,680 (11)    342,597 (19)    269,152 (15)
1980    611,881 (33)    201,929 (11)    347,527 (19)    264,351 (14)
1981    611,645 (33)    202,849 (11)    349,033 (19)    265,173 (14)
1982    572,013 (34)    184,912 (11)    323,157 (19)    243,254 (14)
1983    558,223 (33)    192,452 (12)    306,960 (18)    238,325 (14)
1984    563,854 (33)    210,723 (12)    315,333 (18)    241,142 (14)
1985    573,851 (32)    226,995 (13)    321,311 (18)    246,124 (14)
1986    584,077 (32)    236,935 (13)    327,650 (18)    250,049 (14)
1987    598,313 (32)    247,426 (13)    335,702 (18)    257,060 (14)
1988    619,923 (32)    267,301 (14)    351,621 (18)    266,453 (14)
1989    617,774 (31)    281,155 (14)    361,263 (18)    265,688 (13)
1990    588,573 (32)    270,848 (14)    341,118 (18)    248,524 (13)
1991    542,169 (31)    246,945 (14)    323,249 (19)    233,815 (13)
1992    514,793 (31)    240,953 (14)    311,122 (19)    221,338 (13)
1993    494,154 (30)    239,361 (15)    304,308 (19)    217,784 (13)
1994    489,702 (29)    244,271 (15)    309,644 (19)    225,011 (13)
1995    494,194 (29)    253,300 (15)    319,692 (19)    234,064 (14)
1996    502,040 (28)    273,042 (15)    329,521 (19)    241,426 (14)
1997    507,127 (28)    290,923 (16)    345,986 (19)    250,089 (14)

                   Employment (Percent Share)

          Nonmetro-        Nonmetro-           Total
        Adjacent (4)    Nonadjacent (5)

1976    179,639 (10)     203,035 (12)     1,739,600 (100)
1977    176,695 (10)     199,860 (12)     1,701,631 (100)
1978    190,522 (11)     215,146 (12)     1,787,020 (100)
1979    199,788 (11)     224,594 (12)     1,850,815 (100)
1980    195,894 (11)     223,925 (12)     1,845,507 (100)
1981    196,844 (11)     223,129 (12)     1,848,673 (100)
1982    176,571 (10)     202,265 (12)     1,702,172 (100)
1983    175,906 (11)     199,140 (12)     1,671,006 (100)
1984    185,452 (11)     205,417 (12)     1,721,921 (100)
1985    188,507 (11)     210,057 (12)     1,766,845 (100)
1986    194,569 (11)     215,496 (12)     1,808,776 (100)
1987    201,387 (11)     223,878 (12)     1,863,776 (100)
1988    211,298 (11)     229,743 (12)     1,946,339 (100)
1989    212,584 (11)     230,405 (12)     1,968,869 (100)
1990    201,829 (11)     217,574 (12)     1,868,466 (100)
1991    187,827 (11)     203,150 (12)     1,737,155 (100)
1992    183,674 (11)     201,434 (12)     1,673,314 (100)
1993    186,592 (11)     201,641 (12)     1,643,840 (100)
1994    194,107 (12)     207,192 (12)     1,669,927 (100)
1995    200,841 (12)     212,783 (12)     1,714,874 (100)
1996    207,156 (12)     222,563 (13)     1,775,748 (100)
1997    215,890 (12)     230,908 (13)     1,840,923 (100)

NOTE: Shares may not add to 100 due to rounding.

SOURCE: Special tabulation of microdata from Statistics Canada's Annual
Survey of Manufacturers.

Table 3

Manufacturing employment share trend analysis, preferred models

Beale Code                     CONSTANT            TREND

Large Metro (0)             34.89 (0.0000)    -0.3137 (0.0000)
Large Metro Fringe (1)       9.72 (0.0000)     0.2764 (0.0000)
Medium Metro (2)            18.84 (0.0000)    -0.0859 (0.0000)
Small Metro (3)             15.49 (0.0000)    -0.2258 (0.0000)
Nonmetro-Adjacent (4)       10.20 (0.0000)     0.0629 (0.0000)
Nonmetro-Nonadjacent (5)    11.64 (0.0000)     0.0359 (0.0000)

Beale Code                      TRENDSQ        Adj [R.sup.2]    n

Large Metro (0)                                     0.88        22
Large Metro Fringe (1)                              0.96        22
Medium Metro (2)                                    0.26        22
Small Metro (3)             0.0037 (0.0634)         0.92        22
Nonmetro-Adjacent (4)       0.0062 (0.0000)         0.75        22
Nonmetro-Nonadjacent (5)                            0.25        22

NOTE: p-values are in parentheses.

Table 4

Total manufacturing employment by Beale code and periods

                         Employment (Percent Share)

             Large Metro     Large Metro Fringe    Medium Metro
Period           (0)                (1)                (2)

Quebec
1976         257,567 (49)       51,491 (10)        27,548 (5)
1976-1980    247,291 (47)       53,885 (10)        27,567 (5)
1981-1989    224,877 (45)       58,535 (12)        26,290 (5)
1990-1997    195,297 (41)       60,243 (13)        24,821 (5)
1997         195,341 (39)       66,507 (13)        23,522 (5)
TREND        Negative ***       Positive ***         No Trend
TRENDSQ

Ontario
1976         268,281 (31)       120,102 (14)       192,964 (23)
1976-1980    276,633 (31)       128,684 (15)       196,350 (22)
1981-1989    292,748 (32)       162,539 (18)       192,095 (21)
1990-1997    245,195 (29)       189,068 (22)       171,155 (20)
1997         234,160 (26)       215,183 (24)       171,588 (19)
TREND        Positive ***       Positive ***       Negative ***
TRENDSQ      Negative **

                         Employment (Percent Share)

                                                   Nonmetro-
             Small Metro     Nonmetro-Adjacent    Nonadjacent
Period           (3)                (4)               (5)

Quebec
1976         55,098 (11)        60,327 (12)       71,740 (14)
1976-1980    55,967 (11)        62,464 (12)       74,162 (14)
1981-1989    52,415 (10)        65,440 (13)       77,137 (15)
1990-1997    46,760 (10)        67,065 (14)       80,203 (17)
1997         49,956 (10)        75,564 (15)       90,016 (18)
TREND        Negative ***      Positive ***       Positive ***
TRENDSQ

Ontario
1976         149,033 (17)       82,050 (5)        40,080 (5)
1976-1980    149,271 (17)       85,799 (5)        42,171 (5)
1981-1989    143,139 (16)       88,090 (10)       42,373 (5)
1990-1997    129,161 (15)       87,027 (10)       36,739 (4)
1997         139,174 (16)       91,899 (10)       38,799 (4)
TREND        Negative ***        No Trend          No Trend
TRENDSQ      Positive ***

             Employment (Percent Share)

                        Total
Period

Quebec
1976                523,771 (100)
1976-1980           521,335 (100)
1981-1989           504,695 (100)
1990-1997           474,388 (100)
1997                500,906 (100)
TREND
TRENDSQ

Ontario
1976                852,510 (100)
1976-1980           878,906 (100)
1981-1989           920,984 (100)
1990-1997           858,344 (100)
1997                890,803 (100)
TREND
TRENDSQ

NOTE: Shares may not add to 100 due to rounding.

* Significant at the 5% level.

** Significant at the 5% level.

*** Significant at the 0.1% level.

Table 5

Total manufacturing employment by Beale code and industry, selected
periods

                       Period Average (Percent Share)

Period       Large Metro     Large Metro Fringe    Medium Metro
                 (0)                (1)                (2)

All Industries
1976-1980    597,985 (34)       187,559 (10)       332,459 (19)
1981-1989    588,853 (33)       227,861 (13)       332,448 (18)
1990-1997    516,594 (30)       257,455 (15)       323,080 (19)

Natural Resource-Based
1976-1980    156,116 (34)        39,921 (9)         76,049 (17)
1981-1989    151,779 (32)        49,686 (11)        76,514 (16)
1990-1997    140,288 (30)        62,180 (13)        79,551 (17)
TREND        negative ***       positive ***         no trend
TRENDSQ

Labour-Intensive
1976-1980    170,079 (45)        30,102 (8)         67,275 (18)
1981-1989    159,740 (43)        36,798 (10)        66,130 (18)
1990-1997    126,056 (39)        38,181 (12)        61,329 (19)
TREND        negative ***       positive ***       positive ***
TRENDSQ

Scale-Based
1976-1980    124,213 (21)        63,291 (11)       121,390 (21)
1981-1989    122,505 (21)        73,185 (13)       116,160 (20)
1990-1997    108,196 (20)        80,025 (15)       103,768 (19)
TREND        negative ***       positive ***       negative ***
TRENDSQ

Product-Differentiated
1976-1980     68,670 (35)        23,098 (12)        40,796 (21)
1981-1989     64,737 (32)        30,851 (15)        41,509 (21)
1990-1997     55,080 (27)        36,822 (18)        43,027 (21)
TREND        negative ***       positive ***         no trend
TRENDSQ                         negative **

Science-Based
1976-1980     78,906 (47)        31,148 (18)        26,950 (16)
1981-1989     90,091 (47)        37,340 (19)        32,135 (17)
1990-1997     86,974 (45)        40,249 (21)        35,406 (18)
TREND        positive *         positive ***       positive ***
TRENDSQ      negative **

                    Period Average (Percent Share)

                              Nonmetro-       Nonmetro-
Period       Small Metro       Adjacent      Nonadjacent
                 (3)             (4)             (5)

All Industries
1976-1980    265,092 (15)    188,508 (11)    213,312 (12)
1981-1989    252,585 (14)    193,680 (11)    215,503 (12)
1990-1997    234,006 (13)    197,240 (11)    212,156 (12)

Natural Resource-Based
1976-1980     67,619 (15)     49,365 (11)     69,890 (15)
1981-1989     66,032 (14)     52,206 (11)     74,008 (16)
1990-1997     61,022 (13)     53,460 (11)     74,552 (16)
TREND        negative ***    positive ***    positive ***
TRENDSQ

Labour-Intensive
1976-1980     41,581 (11)     42,137 (11)     30,290 (8)
1981-1989     40,025 (11)     42,137 (11)     30,474 (8)
1990-1997     35,966 (11)     33,827 (10)     28,296 (9)
TREND          no trend        no trend      positive ***
TRENDSQ

Scale-Based
1976-1980    109,803 (19)     67,647 (12)     92,699 (16)
1981-1989    103,512 (18)     70,478 (12)     89,574 (16)
1990-1997     92,721 (17)     76,919 (14)     86,978 (16)
TREND        negative **>    positive ***      no trend
TRENDSQ      positive *

Product-Differentiated
1976-1980     30,085 (15)     19,595 (10)     15,280 (8)
1981-1989     29,204 (14)     20,418 (10)     14,924 (7)
1990-1997     32,078 (16)     22,526 (11)     15,839 (8)
TREND        negative ***    positive ***    negative *
TRENDSQ      positive ***                    positive **

Science-Based
1976-1980     16,005 (10)      9,763 (6)       5,153 (3)
1981-1989     13,812 (7)      11,462 (6)       6,524 (3)
1990-1997     12,220 (6)      10,508 (5)       6,490 (2)
TREND        negative ***      no trend      negative ***
TRENDSQ      positive ***                    positive ***

             Period Average (Percent Share)

Period                   Total

All Industries
1976-1980           1,784,915 (100)
1981-1989           1,810,930 (100)
1990-1997           1,740,531 (100)

Natural Resource-Based
1976-1980             458,960 (100)
1981-1989             470,225 (100)
1990-1997             471,052 (100)
TREND
TRENDSQ

Labour-Intensive
1976-1980             381,464 (100)
1981-1989             372,284 (100)
1990-1997             323,654 (100)
TREND
TRENDSQ

Scale-Based
1976-1980             579,043 (100)
1981-1989             575,414 (100)
1990-1997             548,607 (100)
TREND
TRENDSQ

Product-Differentiated
1976-1980             197,524 (100)
1981-1989             201,643 (100)
1990-1997             205,372 (100)
TREND
TRENDSQ

Science-Based
1976-1980             167,924 (100)
1981-1989             191,364 (100)
1990-1997             191,847 (100)
TREND
TRENDSQ

NOTE: Shares may not add to 100 due to rounding.

* Significant at the 5% level.

** Significant at the 1% level.

*** Significant at the 0.1% level.

SOURCE: Special tabulation of micro data from Statistics Canada's
Annual Survey of Manufacturers.

Table 6

Relative production-worker wages in manufacturing, 1976-1997

                           Large Metro
Year       Large Metro       Fringe       Medium Metro    Small Metro
               (0)             (1)            (2)             (3)

1976           0.94           1.04            1.05           1.09
1977           0.94           1.03            1.05           1.09
1978           0.94           1.03            1.05           1.09
1979           0.94           1.03            1.05           1.07
1980           0.95           1.01            1.05           1.07
1981           0.95           1.01            1.05           1.07
1982           0.94           1.01            1.06           1.09
1983           0.93           1.02            1.05           1.10
1984           0.92           1.02            1.06           1.14
1985           0.92           1.02            1.04           1.14
1986           0.93           1.03            1.04           1.12
1987           0.93           1.03            1.04           1.11
1988           0.92           1.03            1.04           1.11
1989           0.93           1.03            1.03           1.10
1990           0.93           1.01            1.04           1.10
1991           0.92           1.02            1.04           1.09
1992           0.93           1.01            1.04           1.10
1993           0.92           1.04            1.04           1.08
1994           0.91           1.05            1.04           1.08
1995           0.92           1.04            1.03           1.09
1996           0.92           1.03            1.02           1.10
1997           0.91           1.05            1.02           1.09
TREND      negative ***    negative *     negative ***    positive *
TRENDSQ         --         positive **         --         negative *

Year       Nonmetro-Adjacent    Nonmetro-Nonadjacent
                 (4)                     (5)

1976             0.91                   1.02
1977             0.92                   1.02
1978             0.93                   1.03
1979             0.94                   1.03
1980             0.95                   1.04
1981             0.94                   1.02
1982             0.95                   1.02
1983             0.95                   1.03
1984             0.95                   1.00
1985             0.95                   1.00
1986             0.96                   1.01
1987             0.96                   1.02
1988             0.97                   1.03
1989             0.97                   1.03
1990             0.97                   1.03
1991             0.99                   1.04
1992             0.98                   1.03
1993             0.99                   1.03
1994             0.99                   1.01
1995             0.99                   1.03
1996             0.99                   1.03
1997             1.00                   1.01
TREND        positive ***             no trend
TRENDSQ           --

* Significant at the 5% level.

** Significant at the 1% level.

*** Significant at the 0.1% level.

SOURCE: Special tabulation of micro data from Statistics Canada's
Annual Survey of Manufacturers.


Acknowledgements

We would like to thank Professor Ehrensaft of the University of Quebec at Montreal, Tara Vinodrai of the University of Toronto and Robert Campbell of Statistics Canada for their work on the Beale geographic codes used herein. We would also like to thank William Anderson of Boston University for his helpful comments on a previous version of this paper, as well as three anonymous referees for their comments.

This paper represents the views of the authors and does not necessarily reflect the opinions of Statistics Canada.

Notes

(1) The impact of the manufacturing sector on communities is more than its direct employment. A portion of worker wages is spent on local goods and services that create additional jobs. Therefore, the wages paid to production workers is an important base for many local economies. Consequently, industries that pay higher wages will also provide a stronger economic base (see Hewings 1977 for an explanation of the economic base model).

(2) While services are more important, it is probably inappropriate to make comparisons to the service sector as a whole. Services contain such diverse production activities as communications, transportation, retailing and wholesaling. Measured in terms of value added, manufacturing is larger than each of these sectors.

(3) Technically, Krugman's (1991) model is a two-region model rather than a rural-urban model. However, we can think of the region where the industry tends to concentrate as a 'city' and the other region as 'rural'.

(4) Kilkenny (1998) argues that the premise that the cost of transporting industrial goods has fallen relative to agricultural goods is reasonable. She notes that in the past, the transportation of cut meat was far more expensive than moving cattle, leading to a market oriented industry. As technologies have changed and developed (e.g., refrigeration), the cost of transporting meat has fallen relative to the cost of transporting lire cattle. This, at least partially, explains the movement of meatpacking to rural areas.

(5) The Krugman and Kilkenny models differ in a several respects, the most important being that Krugman treats the transportation of agricultural goods as costless, while Kilkenny includes transportation costs for both agricultural and industrial goods.

(6) Coffey's (1994) analysis concentrates on broader employment trends and that of services. His analysis does not lotus on the diffusion of manufacturing employment with in the Montreal census metropolitan area (CMA).

(7) In Canada, there is relatively little evidence that higher productivity levels result in higher production worker wages. Rather, increases in productivity are passed along as lower prices to consumers (Baldwin et al 2001). In Baldwin and Brown (2001), we do find a positive correlation between wage and productivity growth over time. However, the statistical significance of this relationship across rural/urban categories is often weak.

(8) See McGranahan (1989) for a US comparison of social and economic characteristics by Beale code classes.

(9) For their application to Canada, see Ehrensaft and Beeman (1992).

(10) For intercensus years, codes remained fixed based on their classification from the previous census.

(11) For example, the reclassification of a Nonmetro-Nonadjacent census division to Small Metro might also result in the reclassification of those census divisions that border Its new CA boundaries.

(12) The Linear model is specified as follows: Y=a+bTREND+[epsilon] where Y is the variable the trend of which we are analyzing, TREND is a time-based variable that starts at 1 for 1976 and increases by 1 for every year of the study, and e is an error term.

(13) The nonlinear model is specified as Y=a+bTREND+ cTRENDSQ+[epsilon], where TRENDSQ is simply TREND squared. By including TRENDSQ in the regression equation, we are able to determine whether there is a nonlinear (quadratic) relationship between rime and the dependent variable.

(14) The Prais-Winsten estimator takes into account for first-order (AR1) autocorrelation. In almost all cases, the Durbin-Watson statistic indicated that the Prais-Winsten method successfully accounted for autocorrelation, reducing the possibility that the parameters' standard errors are biased.

(15) It should be kept in mind that the diffusion of manufacturing employment experienced by large urban centres may also be occurring in smaller centres as well. However, we cannot observe these trends, because out geographic units--census divisions---are too large to discern such small geographic shifts in employment.

(16) The city-region metropolitan area boundaries used here do not correspond exactly to the CMA boundaries (see Vinodrai 2001). Consequently, the results presented here are not directly comparable to other published information that uses the Toronto and Montreal CMA boundaries.

(17) Some caution should be used when comparing Slater's (1961) results and those of this study. Although our urban definitions are similar (see Slater 1961, 75), they are not exactly the same. Therefore, making direct comparisons is difficult. Nevertheless, the classifications are similar enough to infer differences in trends between the two periods of study with some confidence.

(18) See Baldwin and Rafiquzzaman (1994) for a discussion of the definition of these sectors.

(19) In this paper, industry classifications are held constant to the first year a plant is included in the Annual Survey of Manufactures, rather than allowing each plant's classifications to change over the period. Consequently, statistics reported by industrial sector may differ slightly from those published elsewhere.

(20) We use production workers for this exercise.

References

BALDWIN, J.R., and BROWN, W.M. 2001 'Dynamics of the Canadian manufacturing sector in metropolitan and rural regions' Research Paper no. 169 (Ottawa: Analytical Studies Branch, Statistics Canada)

BALDWIN, J.R., DURAND, R. and HOSEIN, J. 2001 'Restructuring and productivity growth in the Canadian business sector' in Productivity Growth in Canada Catalogue 15-204-XPE (Ottawa: Statistics Canada) 25-37

BALDWIN, J.R., and RAFIQUZZAMAN, M. 1994 'Structural change in the Canadian manufacturing sector (1970-1990)' Research Paper no. 61 (Ottawa: Analytical Studies Branch, Statistics Canada)

BEALE, OZ. 1984 'Poughkeepsie's complaint, or, defining metropolitan areas' American Demographics 6, 28-31

BERNAT, G.A. 1997 "Manufacturing and the Midwest rural economy: recent trends and implications for the future" Rural Development Perspectives 2, 2-12

BUTLER, M.A. 1994 'Rural-urban continuum codes for metro and non-metro counties, 1993' Staff Report no, AGES 9425 (Washington, DC: Economic Research Service, United States Department of Agriculture)

COFFEY, W.J. 1994 The Evolution of Canadian Metropolitan Economies (Montreal: The institute for Research on Public Policy [IRPP]) COFFEY, W., AND POLESE J.M. 1988 'Locational shifts in Canadian employment, 1971-1988: decentralization v. decongestion' The Canadian Geographer 32,248-256.

EHRENSAFT, P., and BEEMAN, J, 1992 'Distance and diversity in non-metropolitan economies' in Rural and Small Town Canada ed. Ray Bollman (Ottawa: Statistics Canada, Supply and Services Canada and Thompson Educational Publishing Press) 193-224

FUJITA, M., KRUGMAN, P.R. and VENEBLES, A. 1999 The Spatial Economy: Cities, Regions and International Trade (Cambridge, MA: MIT Press)

GALE, H.F. 1997. 'The rural-urban gap in manufacturing productivity and wages: effects of industry mix and region' Staff Paper no. 9710 (Washington, DC: Rural Economy Division, Economic Research Service, United States Department of Agriculture)

--. 1998 'Labour productivity and wages in rural and urban manufacturing plants' Review of Regional Studies 28, 13-26 GENERAL ACCOUNTING OFFICE (GAO) 1989 Rural Development: Programs that

Focus on Rural America and Its Economic Development (Washington, DC: General Accounting Office)

GLAESER, E.L,, SCHEINKMAN, J.A. and SHLEIFER, A. 1995 'Economic growth in a cross-section of cities' Journal of Monetary Economics 36, 117-143.

GUJARATI, D.N. 1995 Basic Econometrics (New York; McGraw-Hill)

HEWINGS, J. 1977 Regional Industrial Analysis and Development (New York: 8t. Martin's Press)

KILKENNY, M. 1998 'Transportation costs and rural development' Journal of Regional Science 38,293-312.

KRUGMAN, P.R. 1991 'Increasing returns and economic geography' Journal of Political Economy 99, 483499.

JACOBS, J 1969 The Economy of Cities (New York: Random House) MARSHALL, A. 1920 Principles of Economics 8th ad. /London: The Macmillan Press Ltd.)

MCGRANAHAN, D., HESSION J,C., HINES F.K, and JORDAN M.F. 1986 'Social and economic characteristics of the population of metro and non-metro counties, 1970-80' Report no. 58 (Washington, DC: Rural Development Research, Economics Research Service, United States Department of Agriculture)

NUCCI, A., and LONG, L. 1996 'Are U.S. manufacturing firms moving to the countryside?' Paper presented to the annual meeting of the Southern Regional Science Association, Baltimore, 11 April

--. 1997 'Regional restructuring in manufacturing: the role of establishment formation and dissolution' Paper presented to the Southern Regional Science Association, Memphis, 18 April

PRED, A.R. 1964 'The intrametropolitan location of American manufacturing' Annals of the Association of American Geographer. 54,165-180.

SCOTT, A.J. 1982 'Production system dynamics and metropolitan development' Armais of the Association of American Geographers 72,185-200.

SLATER, D.W 1961 'Decentralization of urban peoples and manufacturing activity in Canada' Canadian Journal of Economics and Political Science 27, 72-84.

STATISTICS CANADA 1999 Manufacturing Industries of Canada: National and Provincial Areas (Ottawa: Statistics Canada)

VINODRAI, T. 2001 'A tale of three cities: the dynamics of manufacturing in Toronto, Montreal and Vancouver, 1976-1997' Research Paper no. 177 (Ottawa: Analytical Studies Branch, Statistics Canada)

WALTER, R., and LEWIS, R.D. 2001 'Beyond the crabgrass frontier: industry and the spread of North American cries, 1850-1950' Journal of Historical Geography 27, 3-19

W. MARK BROWN

Micro-Economic Analysis Division, Statistics Canada, Ottawa, Ontario, Canada K1A 0T6 (e-mail: mark.brown@statcarl.ca)

JOHN R. BALDWIN

Micro-Economic Analysis Division, Statistics Canada, Ottawa, Ontario, Canada K1A 0T6 (e-mail: john.baldwin@statcan.ca)
COPYRIGHT 2003 Canadian Association of Geographers
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2003 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Brown, W. Mark; Baldwin, John R.
Publication:The Canadian Geographer
Geographic Code:1CANA
Date:Jun 22, 2003
Words:11058
Previous Article:Disability activism and the politics of scale.
Next Article:Innovations in resource and environmental management.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters