SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Local Economies, Resource Curse, and the Intrusive Rentier Hypothesis
  4. Data and Methodology
  5. Results
  6. In Search of the Intrusive Rentier: A Brief Look at Individual Cases
  7. Conclusion
  8. References
  9. Appendix A Resource Industry Definitions with NAICS Codes

Looking at 135 Canadian urban areas over a 35-year period (1971–2006), the paper examines the relationship between initial specialisation (using employment) in resource industries and various growth indicators via a mix of descriptive statistics and econometric modelling. The paper differentiates between two resources sectors: resource extraction (mining, logging, etc.); primary resource transformation (paper mills, foundries, smelters, etc.). The evidence for a “resource curse” is mixed. Resource transformation industries are found to be associated with slower population growth, also depressing growth in college-educated cohorts. However, no such relationship is found for resource extraction. We find no evidence for a durable Dutch Disease wage effect. Wages fluctuate in response to resource demand as do working-age populations. Many relationships hold only for the short run. In the end, we argue, the impact of resource specialisation depends on the particular resource and type of industry it spawns, as well as location. There is no generalisable resource curse, valid for all resources and all places.

An abundant literature has emerged on the so-called “Resource Curse” (Boyce and Emery 2011; Bulte et al. 2005; Freeman 2009; Goodman and Worth 2008; Lederman and Maloney 2007; Neumayer 2004; Ross 1999; Sachs and Warner 2001) pointing to the contradictory effects of natural resource abundance on the development of national economies. One of the channels by which this occurs is Dutch Disease, a term first coined by The Economist (1997) following the discovery of natural gas deposits in Dutch territorial waters, driving up the value of the Dutch currency, in turn undermining the international competitivity of Dutch manufacturing. The term has since been applied generically to the relationship between resource specialisation, high exchange rates, and slower growth, notably in manufacturing (Beine, Bos, and Coulombe 2009; Krugman 1987).

The question addressed here is whether similar reasoning can be applied to local economies where high wages generated by resource rents replace exchange rates as the principal channel by which negative impacts are transmitted to local economies. We examine the evidence for 135 urban areas in Canada over a 35-year period (1971–2006), using a mix of descriptive statistics and econometric modelling.

Local Economies, Resource Curse, and the Intrusive Rentier Hypothesis

  1. Top of page
  2. Abstract
  3. Local Economies, Resource Curse, and the Intrusive Rentier Hypothesis
  4. Data and Methodology
  5. Results
  6. In Search of the Intrusive Rentier: A Brief Look at Individual Cases
  7. Conclusion
  8. References
  9. Appendix A Resource Industry Definitions with NAICS Codes

For national economies, rich natural resource endowments become a “curse” when they distort the allocation of resources (i.e., away from knowledge-rich industries) and undermine the efficient functioning of political institutions. The latter has all too often been the outcome in developing nations. Revenues generated by resource rents retard the introduction of effective tax systems and cause governments to neglect investments in human capital. Gylfason (2001) observes a negative relationship between the weight of natural resources in national economies and various indicators of educational attainment, also pointing to the risks of rent seeking and protectionism. Maloney (2007) coins the term of “Cultural Dutch Disease,” essentially a rentier mentality where rents act as a disincentive to innovation and institutional reform. In short, for nations as for individuals, the corrupting effects of easy riches are all too real.

However, the evidence for a universal resource curse is weak. Abstracting from straightforward exchange rate impacts, institutions, and human capital are almost always cited as the chief determinants of whether natural resources will spawn a “curse” or not. Blomström and Kokko (2007), Bravo-Ortega and de Gregorio (2007), Gylfason (2001), and Maloney (2007) all make the same point: Natural resources have not been a “curse” for nations like Canada, Australia, Finland, and Sweden. There is little evidence of a “curse” where natural resources are combined with human capital and properly functioning institutions. Furthermore, Dutch Disease alone is not a sufficient condition for an enduring “curse” as the evolution of the Dutch economy since the 1970s demonstrates. Beine, Bos, and Coulombe (2009) show that Canada suffers from Dutch Disease, attributing 54 percent of manufacturing job losses between 2002 and 2007 to a rising Canadian dollar fuelled by high commodity prices. However, they add that it only becomes a “disease” if the manufacturing sector fails to rebound.

The evidence is mixed for states and provinces in federations. For Canadian provinces, Coulombe (2011) focuses on the relationship with productivity, observing that the technologies of resource exploitation can induce innovation and investments in human capital, citing impressive productivity growth of the Newfoundland economy since the discovery of offshore oil in the 1990s. Royalties and other revenues generated by natural resources can contribute to productivity growth via improvement of public infrastructures, including public education. Using panel data for U.S. states for 1970–2001, Boyce and Emery (2011) conclude that resource abundance is negatively correlated with growth rates but positively correlated with income levels. Freeman (2009) finds evidence of a “curse” for U.S. states where the principal channel is the crowding out of manufacturing, producing slower growth.

Systematic examinations of the impact of natural resource specialisation are rare below the provincial or state level. Again, the evidence is mixed. The fate of single industry resource towns is a recurring theme in Canadian local development literature (Barnes and Hayter 1994; Hayter 2000; Lucas 1971; Slack, Bourne, and Gertler 2003). Looking at 99 communities in Canada, Gaudreault (2011) finds a negative correlation between resource specialisation (measured by employment shares) and business start-ups. For U.S. counties, James and Aaland (2011) find a negative relationship between initial resource specialisation (measured by earnings shares) and per capita income growth, but note that effects dissipate over time. For Australian government areas, Hajkowics, Heyenga, and Moffat (2011) find a positive relationship between the gross value of local mineral production, incomes, and other welfare indicators. For U.S. counties, the negative results may in part be attributable to the predominance coal mining, with high concentrations in Kentucky and West Virginia (James and Aaland 2011). Looking at the local impacts of the coal boom and bust, Black, McKinnish, and Sanders (2005) find only weak evidence for the crowding out of manufacturing during boom years, observing also that wages and workers react to demand swings. Wages rose, predictably, during the boom and declined during the bust with accompanying outmigration. Deller and Schreiber (2012) examine the local impacts of mining, including rapidly growing sand frack mining, for non-metropolitan U.S. counties for the 2000–2007 boom years, finding a positive impact on wages but not on employment and population growth. They point out that fracking is both less labour intensive than more traditional mining activity with potentially higher paying jobs and generally more reliant on a transient labour force. The uncertainty generated by boom and bust swings introduces a disincentive, they suggest, to long-term investments in retailing and other industries, all of which combine to explain low overall growth effects.

This brief literature review suggests two preliminary conclusions, also pointing to a potential difficulty. First, impacts are sensitive to the nature of the resource industry studied. We should not, e.g., expect aluminum production to produce the same outcomes as oil drilling or coal mining. Second, the prevalence of booms and busts means that outcomes are sensitive to observed time periods, not only their length but also the location of endpoints. Both point to the difficulty of making board generalisations about the impact of local resource specialisation on growth.

Polèse and Shearmur (2002, 2006a) coined the term “Intrusive Rentier Syndrome” to describe why resource communities in Eastern Canada failed to diversify into other industries. Although not a formalised model, it provides a basis for conceptualising the possible local impacts of resource specialisation. Figure 1 depicts the core elements. The principal modification is the addition of education to the human capital box. Polèse and Shearmur (2002, 2006a) argue that large well-paying firms act as a disincentive for smaller firms to invest in worker training. Workers, once trained, leave for better paying jobs in larger firms. We add that the presence of well-paying middle-skill jobs may also act as a disincentive for local youth to pursue a post-secondary education.

figure

Figure 1. Intrusive Rentier Syndrome: Schematic Presentation.

Adapted from Polèse and Shearmur (2002, 2006a) and Polèse (2009).

Download figure to PowerPoint

In Canada, the dominant resources industries have traditionally been logging and mining with attendant pulp and paper mills, smelters, foundries, and, more recently, oil and natural gas. A common feature is the weight of scale economies, spawning large plants (or mines), highly capital intensive, in which wages comprise a fraction of total costs. The production of aluminium ingots, e.g., involves a costly process called electrolysis requiring vast quantities of electricity, far outweighing wage costs. The result is large operations that pay relatively high wages for relatively middle-skill jobs, not only because of resource rents but also because the technology allows it.

Summarising Figure 1, the effects of highly capitalised resource industries on local economies play out at two levels. First, they push up wages, undermining community competitiveness for non-resource industries, a Dutch Disease effect. Second, they produce environments that discourage local business start-ups and act as a disincentive to post-secondary education. Large plants facilitate unionisation possibly further fuelling wages. The outcome is a labour market with a restricted number of high-paying jobs, but fewer jobs overall, a generally less educated population, and below-average growth.

However, the Intrusive Rentier hypothesis is founded on the same implicit premise as much of the literature on Dutch Disease and the Resource Curse: that there is something unique about resource industries in general. In the analysis below, the resource sector is divided into two broad classes, corresponding to the first and second stages of resource exploitation.

Data and Methodology

  1. Top of page
  2. Abstract
  3. Local Economies, Resource Curse, and the Intrusive Rentier Hypothesis
  4. Data and Methodology
  5. Results
  6. In Search of the Intrusive Rentier: A Brief Look at Individual Cases
  7. Conclusion
  8. References
  9. Appendix A Resource Industry Definitions with NAICS Codes

The methodology is based on a mix of descriptive statistics and an econometric growth model for multiple dependent variables using a seemingly unrelated regressions (SUR) approach. The data are drawn from five censuses of Canada (1971, 1981, 1991, 2001, 2006) based on special Statistics Canada tabulations.1 To ensure comparability over time, urban areas only are examined, defined as urban places with populations over 10,000 in 2006,2 defined as Census Agglomerations or Census Metropolitan Areas by Statistics Canada. Non-urban spatial units are excluded because boundary changes make time comparisons impossible. The urban data set contains 135 observations accounting for 81 percent of Canadian employment in 2006.

Industry classes are standardised over time. Canada introduced the North American Industrial Classification System (NAICS) in 1997, changing from the previous Standard Industrial Classification (SIC). The initial pre-1996 data set contained 141 SIC-coded industry classes. The updated data set has 126 industry classes, recoded to match NAICS. The basic datum is total employment by industry. Two resource industry classes were created, defined in Appendix A. The first identifies Resource Extractive industries: mining, logging, drilling, etc. Agriculture and fishing, although in the primary sector, are excluded because they are not typically considered in the Resource Curse literature. The second class identifies Primary Transformation Manufacturing: sawmills, pulp and paper mills, smelters, foundries, and other basic wood and metal product industries. Oil refining and petrochemicals are excluded because they are not necessarily produced from local or nearby resources. Much of the oil refined in Eastern Canada is imported. The two resource variables are expressed as employment location quotients (abbreviated as LQs).3 Thus, it is not resource abundance that is being measured, but relative resource specialisation. To test whether initial specialisation affects growth, the quotients enter as independent variables in the model presented below.4

Growth can be measured using several variables. The model is based on a set of five dependent variables expressed as changes in: population; wages; non-resource manufacturing; other non-resource employment; college-educated populations. To ensure in each case that the effects of resource specialisation are adequately isolated, control variables consider other initial local attributes, including location. Provincial dummies indirectly capture resource demand cycles, specifically for Alberta, the heart of Canada's oil industry. Finally, in the model, all values are expressed as logs.

Econometric model

The model is an adaptation of a general growth model, which captures the effect of initial independent variables grouped in a matrix of variables Xt-p of dimension (N × K), where N is the total number of observations (135 urban areas in this case) while K is the total number of independent variables affecting the growth of given dependent variables, yt − yt-p, grouped in a vector of dimension (N × 1). The general expression for such a model is given by the following equation (1):

  • display math(1)

where β is a vector of unknown parameter of dimension (K × 1) measuring the effect of the independent variable on the growth of the dependent variable, αi is a vector of provincial fixed effects controlling for substantial differences in the structure of the local economies of dimension (N × I*) where I* is the total number of provinces and territories (11), and εt is a vector of error terms, of dimension (N × 1). It should be noted that the provincial fixed effects account for demand swings, related to historical differences and difference in industrial composition.5

To adapt this framework to multiple dependent variables, a SUR (Zellner 1962) approach is used. The general growth SUR specification is based on G (G = 1, …, g) dependent variables (equation (2)).

  • display math(2)

where ε1,t, ε2,t, …, εg,t are vectors of error terms possibly correlated among equations. Expressed differently, a shock influencing the growth of wages in region n may also influence the growth in employment in the same region. This influence is captured through a general variance-covariance matrix capturing the relations among the equations.

Since the model seeks to capture the possible effects of resource specialisation on the growth of urban areas, matrix of independent variables included the following variables:

  • Eit-p: Extractive industries. LQ for employment in extractive industries, urban area i time t-p.
  • Tit-p: Primary transformation. LQ for employment in primary resource transformation manufacturing, urban area i time t-p.

The dependent variables are defined as follows:

  • POPit − POPit-p: Population (log): Growth of the working-age population (ages 15–64), urban area i between time t and t-p. The 15-64 population was preferred as it indirectly captures migratory movements of the labour force.
  • Lit − Lit-p: Non-resource manufacturing (log): Growth in manufacturing employment, excluding primary transformation, urban area i between time t and t-p.
  • eit − eit-p: Other employment (log): Growth in remaining employment, urban area i between time t and t-p (excluding manufacturing and extractive industries). The exclusion is necessary to avoid overlap with other variables.
  • wit − wit-p: Weekly mean wage (log): Growth of average weekly earnings per employed worker, urban area i between time t and t-p;
  • Bit − Bit-p: B.A. degree holders (log): Growth in the number of persons aged 15 and over with at least a B.A. (or B.Sci.) degree, urban area i between time t and t-p.

Other independent variables, controlling for local attributes, also allow us to further control for possible demand shifts:

  • Non-resource manufacturing LQ: LQ for employment in non-resource manufacturing, urban area i time t-p.
  • Centre-periphery: “0” if located more than 90 min (by road) from an urban area with a population over 500,000; “1” for the remainder.
  • Accessibility (two variables): Continental market potential, based on a gravity-model approach, first, for road travel time distances and, second, for rail-travel time and populations for U.S. counties and Canadian urban areas; taken from Apparicio et al. (2007).
  • Regional and provincial dummies: As explained above, these capture mean provincial effects.6

Two models are estimated. The first specification leaves out local initial attributes relative to dependant variables, while the second controls for initial conditions related to population size (ages 15–64), wages,7 employment rates, and percentage of B.A. degree holders.

By combining the models (one for each dependant variable) into a unique equation system related through correlations between error terms (SUR), we obtain a complex model allowing for the simultaneous inclusion of the five dependent variables in an integrated model specification. The final specification is estimated using a standard SUR procedure available on any econometric software (such as Stata), constructing the necessary variables prior to estimation.

Recalling the discussion of the Dutch Disease and Intrusive Rentier hypothesis, we expect to find that resource specialisation (measured by the βG parameters) pushes up wage rates, depresses employment growth, depresses education levels, and crowds out non-resource manufacturing. Finally, the econometric model allows us to distinguish between short-term cyclical and long-term effects by changing the time frame (t and p). The model is applied to four time periods: 1971–2006 (long term) and 1971–1981, 1981–1991, and 1991–2006 (short term).

Notwithstanding the robust (we believe) specification of the model, data availability limits our ability to adequately test the Intrusive Rentier hypothesis. This would have required variables on establishment size and on business start-ups, neither of which exists for our spatial units and time frames. Moreover, census data do not tell us what occurs between census years.8 Finally, the analysis is limited to urban areas, excluding rural resource communities possibly prone to Dutch Disease. Note, however, that the fairly low population threshold (10,000) allows us to capture a large number of small resource-based communities. The majority of observations have populations below 50,000 (2006 census). With these caveats in mind, we now turn to the results.

Results

  1. Top of page
  2. Abstract
  3. Local Economies, Resource Curse, and the Intrusive Rentier Hypothesis
  4. Data and Methodology
  5. Results
  6. In Search of the Intrusive Rentier: A Brief Look at Individual Cases
  7. Conclusion
  8. References
  9. Appendix A Resource Industry Definitions with NAICS Codes

Findings are presented in two stages, beginning with descriptive statistics followed by econometric modelling results.

Descriptive statistics

Table 1 shows selected statistics for the two resource sectors over the 35-year study period. The evolution of employment in resource extraction is indicative of boom and bust cycles, with employment rising sharply during the 1971–1981 decade, largely fuelled by oil and gas extraction, followed by stagnation and decline, and then an upswing again beginning in the late 1990s. The principal driver of the latest turnaround is the voracious appetite of emerging economies, notably China and India, for oil and mineral resources. The decline in the forestry and logging shares after 1991 also reflects the negative impact of the Internet on newsprint consumption and the weakening U.S. demand.9

Table 1. Selected Statistics for two Resource Sectors (Canadian Urban System. n = 135)
EmploymentYears
197119811991199620012006
Resource extraction (total)93,329134,410136,687115,142120,100167,320
% oil and gas extraction16.729.931.026.527.631.1
% mining59.348.044.750.050.553.1
% forestry and logging24.122.124.323.522.015.9
Primary transformation (total)234,192283,796230,485210,617215,045198,020
% paper manufacturing37.636.537.634.933.231.3
% primary metal manufacturing41.138.737.832.934.233.1
% wood products21.424.824.632.232.635.5
Correlation coefficients: matched location quotients
Extraction-extraction (2006)0.590.640.830.920.981.00
Transformation- transformation (2006)0.850.880.940.960.971.00
Extraction-transformation (compared)0.080.040.040.020.030.03

Employment change over time is more regular for resource transformation; steady employment decline since 1981 with a hiatus in 2001, again largely technology induced: information technology impact on newsprint demand plus improvements in labour productivity in paper, aluminium, and steel production. The sector also underwent a structural transformation with wood products industries steadily increasing their share, which includes sawmills and producers of wood panels, doors and windows, etc … , industries typically less capital intensive and paying lower wages.

The last three rows in Table 1 give correlation coefficients for LQs compared to the year 2006, respectively, for employment in resource extraction, transformation, and between the two industry groups. Comparing the two resource sectors (last row), correlation coefficients are consistently near zero, suggesting two different geographies. If nothing else, this vindicates their use as separate variables. Although resource extraction and transformation may be linked, they do not necessarily occur in the same locations. Many extractive industries (i.e., oil drilling, diamond and gold mining) are not automatically tied to downstream activities. Unprocessed ores and lumber are often directly exported, notably where water transport is available.

The evolution (1971–2006) of industry LQs also echoes two different spatial dynamics, noticeably more volatile for extractive employment, a sign of more frequent booms and busts. The distribution of extractive employment in 2006 is only mildly indicative of that some 35 years earlier, a correlation coefficient of 0.59. In contrast, the geography of primary transformation manufacturing has remained largely stable over time (correlation coefficient of 0.85 with 1971), a reflection, we may assume, of the weight of large plants and sunk costs.

Correlations with selected variables

Table 2 shows correlation coefficients between LQs for the two resource sectors and selected variables.

Table 2. Correlation Coefficients: Between Location Quotients (LQs) for Two Resource Sectors and Selected Variables. 1971–2006
Location quotient197119811991199620012006
  1. Note: Coefficients above 0.22 are significant at 0.05 or better.

Average weekly wage (log)
Extraction0.3860.4400.4270.4050.3640.513
Transformation0.3700.3700.3640.3260.2300.080
% population B.A. degree or higher (log)
Extraction−0.269−0.223−0.263−0.271−0.301−0.336
Transformation−0.324−0.379−0.437−0.451−0.469−0.432
Non-resource manufacturing employment (log)
Extraction−0.446−0.429−0.403−0.432−0.436−0.419
Transformation−0.325−0.370−0.382−0.367−0.390−0.381
Employment rate (excluding manufacturing and resource sectors) (log)
Extraction−0.332−0.253−0.279−0.073−0.056−0.060
Transformation−0.564−0.561−0.617−0.589−0.657−0.673

For wages, correlation coefficients with extractive industries are consistently positive from 1971 to 2006, with some volatility. For primary transformation manufacturing, the correlations are also positive, but weaker and with a systematic decline after 1991, possibly reflecting the declining weight of the industry in local economies and also of wage adjustments in reaction to declining demand.

For the percentage of B.A. degree holders, the relationship is consistently negative, but weak, for extractive industries. The negative relationship is stronger for primary transformation and increasing over time.

For employment in non-resource manufacturing, the coefficients for extractive industries are consistently negative, consistent with a crowding out effect. However, caution is in order. Urban areas specialised in extractive industries are more frequent in Western Canadian locations while non-resource manufacturing is concentrated in southern Ontario and Quebec (Bourne et al. 2011). Crowding out can only be said to occur where industry groups are in competition for the same locations. Spatial competition cannot be ruled out, but is it is difficult to argue that it is the principal source of the negative coefficients. The correlations for resource transformation are also systematically negative, although weaker.

The coefficients for employment rates are mildly negative for extractive industries before 1996 and then cease to be significant. However, the coefficients for resource transformation are strongly negative and consistently so. The difference between the two suggests that the employment-inhibiting effects of resource specialisation, if confirmed, are not an attribute of resource extraction, but rather of industries involved in the subsequent (manufacturing) stages of resource exploitation.

Summarising the evidence so far, we find positive correlations between local resource specialisation and wages, notably for extractive industries, consistent with Dutch Disease. However, we also observe volatility in coefficients over time together with a declining relationship for resource transformation, suggesting that wages react to changing demand conditions. The negative coefficients with non-resource manufacturing for both resource sectors are consistent with crowding out scenarios, but need to be interpreted with caution given the differing geographies of resource and non-resource industries. The evidence for possible employment-inhibiting effects is weak for extractive industries, but stronger for primary resource transformation industries, pointing to two different scenarios. On the other hand, the negative coefficients with the education variable hold for both resource industries, but again stronger for resource transformation. The correlations do not yet allow us to infer causation. For that, we turn to the econometric model.

Econometric model results

Table 3 shows results for the five simultaneous regressions. Each regression aims to capture a posited effect (dependent variable) of local resource specialisation. The five dependent variables are expressed in similar terms, measuring growth over the given time period. Results are shown in turn for four time periods: 1971–2006 and 1971–1981; 1981–1991; 1991–2006. Each regression is presented for two models (labelled 1 and 2) where model 2 includes additional variables controlling for initial attributes. We focus on model 2.10

Table 3. Econometric Model ResultsThumbnail image of

The evidence for a local Resource Curse, strictly defined, is weak, i.e., slower growth or decline due to an initial specialisation in resource extraction. No significant negative coefficients appear for extractive industries (Extraction LQ) for the population equation for any period. Indeed, the coefficient is positive for 1971–1981 (the first oil boom). On the other hand, the coefficients for resource transformation (Transformation LQ) are negative for the long-run model and earlier two periods. In other words, initial specialisation in primary transformation (chiefly of wood and minerals in the Canadian case) generally produced slower growth. If a curse can be said to exist, it is associated with resource transformation manufacturing, not the presence of resources. Indeed, there is no necessary link; resources to be transformed may be imported from other locations.

The evidence for a crowding out effect of non-resource manufacturing is also weak. In no period does the extraction variable show a significant negative coefficient in the non-resource manufacturing equation. Indeed, the relationship is positive for 1981–1991. For resource transformation, the evidence is also weak with no significant coefficients for any period. A look at other independent variables in the non-resource manufacturing equation suggests that the principal factors driving out employment (or retarding its growth) are initial specialisation in non-resource manufacturing, urban size (population), and higher wages, entirely consistent with the literature on the location dynamics of manufacturing in Canada and elsewhere (Desmet and Fafchamps 2005, Henderson 1997, Polèse and Shearmur 2006b). Neither is there evidence for a systematic downward effect on other employment. Indeed, the only significant coefficient is positive (extraction: 1971–1981), a sign that resource booms also drive up employment in other industries, similar to the findings of Black, McKinnish, and Sanders (2005) for the Appalachian coal boom.

The evidence for Dutch Disease driving up wages is somewhat stronger, but still weak. Neither extractive nor resource transformation industries exhibit a significant positive coefficient for the long-term model (1971–2006); initial resource specialisation did not, in other words, exert a durable upward pressure on wages. For the wage equation, the three shorter term models are systematically better predictors (the R2 for 1981–1992 is 0.70 compared to 0.57 for 1991–2006), a sign of volatility, but also of wage flexibility. Furthermore, looking at wages from the other side (as an independent variable), initial higher wages inhibit wage growth (negative coefficients in all but the 1991–2006 model). All this reinforces the perception of local labour markets in which wage rates adjust to changing demand conditions; perhaps with some lag and stickiness, but adjust nonetheless. The absence of a durable upward pressure on wages suggests that the growth-inhibiting effects of resource specialisation, where observed, cannot be attributed (at least not primarily) to Dutch Disease type scenarios. On the other hand, the strong positive coefficient for the Alberta dummy (1971–2006) for wage growth, driven by that province's oil-based economy, suggests that long-term upward pressures on wages cannot be totally discounted.

Table 3 confirms the negative relationship with the education variable, but again only for resource transformation. The relationship holds both for the long-run model and two earlier periods. In sum, initial local specialisation in resource transformation retarded the growth of college and university educated persons. For resource transformation, the negative relationship with the education variable is the most consistent across the five equations, the population equation aside, largely attributable, it is reasonable to assume, to outmigration. A look at other independent variables in the education equation reveals that the negative relationship is a general attribute of manufacturing, not solely of resource transformation manufacturing. The negative coefficients for non-resource manufacturing (1971–2006, 1991–2006) imply that the Intrusive Rentier, if present, is not exclusive to resource transformation manufacturing, a perception reinforced by the negative coefficient for non-resource manufacturing (independent variable) in the population equation.

An examination of the results by time frame brings out the periodicity of many relationships. None of the five equations (model 1 or 2) reveal a significant relationship (negative or positive) with extraction for the long-run model (1971–2006). Significant coefficients appear only for shorter time periods, pointing to “boom and bust” scenarios, with equally short-term impacts. The provincial dummies are particularly useful in bringing out the influence of cyclical swings. The first oil boom primarily benefited oil-rich Alberta, mirrored in high-positive coefficient in the population equation for the Alberta dummy in the 1971–1981 model. Coefficients are also positive for the employment and wage equations. In the following decade (1981–1991), in which oil prices fell, the relationship with population growth reverses, becoming negative, with now also negative coefficients for the employment, wage, and education equations (the most negative of any province), a characteristic “bust” scenario: Employment and wages fall, and the educated leave. During the last period (1991–2006), all three coefficients reverse again, associated now with growing working-age populations, rising employment rates, and growing educated populations. More significantly, the high-positive coefficient for Alberta in the population equation (1971–2006 model) suggests that resource specialisation is not necessarily incompatible with long-term growth (compared to Ontario), at least for the period studied, although marked by booms and busts. Only the future will tell whether Alberta's performance is truly durable.

The weight of the provincial dummies in all the models, notably Alberta and to a lesser extent British Columbia (its economy largely dependent on the logging industry), is a reminder that the performance of the 135 spatial units (urban areas) cannot be explained solely by local economic structures and resource endowments, but also by activities in surrounding rural areas, not captured in the model. In Alberta, much oil and gas exploitation takes place in rural areas, but is nonetheless a driver (or the inverse) of nearby urban economies. By the same token, interaction between neighbouring provinces influences the results. The negative coefficients for the education variable for Saskatchewan during the 1971–1981 oil boom, the mirror opposite of the values for Alberta, are most probably a reflection, at least in part, of educated populations moving to neighbouring Alberta. The coefficients for British Columbia on the population equation with the highest for the most recent model (1991–2006) point to yet another growth sequence, driven not only by resources but also by location. Vancouver is the principal port of entry and point contact with the rising Asian economies. For both Western provinces, growth in educated populations was more rapid than in Ontario (reference point).

The importance of location is further brought home by the consistently positive coefficients for the rail access variable, measuring accessibility at the continental level, in the population, non-resource manufacturing, and employment equations. In other words, Canadian urban areas with superior rail access to North American markets saw higher rates of non-resource employment growth, a result not inconsistent with the growing integration of the Canadian and U.S. economies. For long-distance (overland) trade in merchandise, rail remains the dominant transport mode. Once we abstract from the continued draw of Alberta and British Columbia, North American rail accessibility appears as the only positive (significant) predictor of long-term growth in the model (population equation: 1971–2006), and initial specialisation in resource transformation is the only negative predictor.

Summing up, we find little evidence for a systematic Resource Curse linked to the extraction of local resources. Boom and bust scenarios typically associated with resource extraction exhibit negative effects during downturns, as would be expected; but are not necessarily incompatible, the results suggest, with long-term growth, at least over the period examined (1971–2006). On the other hand, initial specialisation in primary resource transformation does display some of the symptoms posited by the Intrusive Rentier hypothesis: slower growth (or possible decline) in working-age populations and, specifically, in college and university educated persons, but which cannot necessarily be traced to a Dutch Disease wage effect.

A comparison of the econometric model results with the correlation results in Table 2 points to an apparent paradox for wages, notably for resource extraction. Table 2 revealed that wages are generally higher in urban areas specialised in resource extraction. Yet, the model results point to an absence of Dutch Disease driving up wages. Part of the paradox, we argue, lies in the reactivity of wages, which are high only so long as demand is high and rise only so long as demand is rising. Part of the answer can also be found in location. High wages are needed to attract workers to what are often fairly remote locations, which considerably diminish the probability of Intrusive Rentier crowding out effects. The rentier is only intrusive if there are alternative economic activities to be intruded upon, which is not necessarily the case for all locations. With that in mind, we now return to a more descriptive mode with a brief examination of particular cases.

In Search of the Intrusive Rentier: A Brief Look at Individual Cases

  1. Top of page
  2. Abstract
  3. Local Economies, Resource Curse, and the Intrusive Rentier Hypothesis
  4. Data and Methodology
  5. Results
  6. In Search of the Intrusive Rentier: A Brief Look at Individual Cases
  7. Conclusion
  8. References
  9. Appendix A Resource Industry Definitions with NAICS Codes

The ultimate outcome of Dutch Disease and its Intrusive Rentier variant (Figure 1) are communities that exhibit both high wages and net outmigration, a priori an incompatible combination. In support of the Intrusive Rentier hypothesis, Polèse and Shearmur (2006a) cite the example of the Quebec North Shore community of Baie-Comeau, almost totally dependent on a large paper mill and aluminum plant, with average wages well above those in Montreal, but with continuing net outmigration. Is Baie-Comeau an anomaly? We focus on the most recent period: 1991–2006.

Table 4 lists urban areas selected as follows: All 135 urban areas are first sorted by 1991 wage rates, keeping those with above-average wages (index of 1.0 or above); within this group, urban areas are then sorted by working-age population change (%) for 1991–2006, keeping those that declined or grew considerably below the system average, the cut-off placed at 4 percent.11 Indicators for each urban area were expressed as indexes (system = 1.0) for wages, % B.A. degree holders, and employment ratio. The right-hand columns show LQs for the two resource sectors for 1991 and 2006, shaded for values above 3.0, indicating high resource specialisation.

Table 4. Urban Areas Exhibiting Both High Wages and Slow Growth (1991–2006)—Selected Indicators
Urban areaPopulationPop Δ (%) age 15–64Wage indexbEducation indexEmployment rate indexLQ extractionaLQ transformationa
20061991–2006 (%)1991–1996 (%)1991Δ 1991–20061991Δ 1991–20061991Δ 1991–20061991200619912006
  1. a

    Shaded = LQ < 3.0.

  2. b

    Shaded = wage index 1.00 or above in both years.

Elliot lake11,440−33.3−5.51.16−0.320.48−0.080.74−0.0325.165.950.600.59
Prince Rupert13,275−25.91.91.12−0.210.53−0.010.91−0.031.181.504.640.90
Kitimat8,950−19.5−1.51.30−0.010.54−0.130.870.021.260.8220.6226.39
Kapuskasing8,350−18.9−5.01.14−0.090.41−0.010.850.022.004.8111.549.48
Powell River16,270−13.16.21.07−0.180.470.070.860.056.503.879.296.68
Baie-Comeau29,460−9.90.61.08−0.060.46−0.060.910.060.971.6710.4013.78
Thompson13,540−9.5−4.51.130.000.59−0.080.92−0.0319.7716.051.601.77
Timmins42,455−7.40.91.01−0.010.49−0.060.890.0415.149.941.452.08
Port Alberni25,075−5.50.31.07−0.240.370.000.860.038.674.559.627.16
Saguenay149,600−4.61.51.00−0.140.64−0.020.800.100.841.026.375.98
Kenora14,950−2.41.81.01−0.040.63−0.070.970.020.961.344.113.23
Terrace18,450−2.210.31.11−0.200.50−0.010.89−0.036.683.602.993.58
Sudbury156,400−1.11.11.03−0.030.69−0.040.940.008.125.792.121.63
Thunder Bay121,055−0.50.21.00−0.060.730.001.00−0.031.601.752.863.05
Sarnia87,6951.6−3.61.08−0.020.73−0.160.960.021.070.560.170.39
Sorel-Tracy47,1402.5−2.71.02−0.140.380.010.810.110.670.566.378.43
Rouyn-Noranda39,4403.81.81.03−0.140.61−0.050.870.029.837.241.231.82

Baie-Comeau appears in Table 4, but is not alone. Seventeen communities are on the list. More significantly, only one, Sarnia in southern Ontario, is not resource based, the heart of Canada's petrochemical industry. In short, the high wage-slow growth combination is almost exclusively associated with high resource specialisation.12 However, the corollary does not follow. Out of the 135 urban areas, 55 registered LQs above 3.0 in resource industries in 1991. While almost all high-wage/slow growth communities are resource communities, most resource communities are not high-wage/slow growth communities. The high-wage/slow growth couple is not a generalisable attribute of resource-dependent communities.

Taking a closer look at the 16 cases,13 all but one saw their wage index decline between 1991 and 2006. Communities are not frozen in a syndrome of sticky wages, consistent with our preceding findings. Of the 16 communities, 12 are primarily or partly specialised in mining and in forestry-related industries, the latter in decline over the study period (1991–2006). Prince Albert, second in Table 4, saw its paper mill close in early 2006, causing its primary transformation LQ to fall below 3.0 and would no longer be classified as a resource community. The severest case of decline, Elliot Lake, saw the collapse of its chief resource base, uranium mining, during the mid-1990s reflected in a sharply falling LQ for extraction from 26.2 to 6.0. In sum, the evidence does not support the hypothesis of a Dutch Disease whose effects persist after the plant (mine) has closed, although it does not necessarily point to happy endings.

Five cases (also shaded) come closest to the Intrusive Rentier model: persistent above-average wages (although they may be falling); below-average education; below-average employment rates; continued resource dependency; slow growth or decline. Kitimat in Northern British Columbia shows a wage index of 1.30 in 1991 (1.29 in 2006), an education index of 0.54 in 1991 (0.41), and an employment index of 0.87 in 1991 (0.89); the working-age population declined by some 20 percent. Kitimat also exhibits an extremely high level of resource specialisation: an LQ of 26.4 for primary transformation in 2006 (20.6 in 1991), almost wholly due to a large aluminum plant and large paper mill, the same mix as Baie-Comeau. For the Intrusive Rentier model to apply, in short, resource specialisation must remain high, keeping wages up and/or drive them up, not a common case.

Common to all urban areas in Table 4 is the combination of above-average wages and below-average educational attainment. More troubling still, the education index has either fallen or remained stable since 1991 with only two exceptions. The econometric results pointed in the same direction for resource transformation, although for two earlier periods, suggesting that this is a fairly deep-seated attribute of resource-based manufacturing. However, the results revealed that this is also true for non-resource manufacturing, to which the presence of Sarnia in Table 4 bears witness.

With the exception of Sarnia and Sorel-Tracy, the latter near Montreal, all communities in Table 4 are located in the northern reaches of their respective provinces far from major markets (recall the weight of the continental rail access variable in the econometric model). Their resource endowment may be their chief and perhaps only comparative advantage. “Curse,” “syndrome,” and “disease” all imply that they would be better off without the resource, which is difficult to defend looking at the communities in Table 4. It is the combination of small size, peripheral location, and industry-specific attributes of resource extraction or transformation, Table 4 suggests, that creates the particular circumstances that produce outmigration under conditions of high wages. However, the presence of Saguenay, Sudbury, and Thunder Bay (population over 100,000) indicates that even mid-sized urban areas are not immune.

To conclude, this foray into individual cases reveals that resource specialisation is sometimes associated with a combination of high wages (although they may be falling) and low human capital, not an attractive mix for attracting other industries, especially in more remote locations. Such communities do in part conform to the Intrusive Rentier hypothesis, but they are unique cases and cannot be extrapolated to all resource-based communities.

Conclusion

  1. Top of page
  2. Abstract
  3. Local Economies, Resource Curse, and the Intrusive Rentier Hypothesis
  4. Data and Methodology
  5. Results
  6. In Search of the Intrusive Rentier: A Brief Look at Individual Cases
  7. Conclusion
  8. References
  9. Appendix A Resource Industry Definitions with NAICS Codes

Looking at 135 Canadian communities over a 35-year period (1971–2006), we ask whether initial specialisation in resource industries stunts growth. Employment LQs were calculated for two resource industry classes: 1) resource extraction (drilling, mining, forestry, etc.); and 2) primary resource transformation manufacturing (paper mills, aluminum plants, metal ores smelting, etc.). Analysis involved the use of descriptive statistics and application of an econometric model for 1971–2006 and sub-periods where initial local resource specialisations act as independent variables and posited effects as dependant variables.

The evidence is mixed. The findings do not point to a generalisable relationship (valid for both resource industries) between resource specialisation, higher wages, and stunted growth. Slower growth (of the working-age population, other sector employment, and college-educated cohorts) was found to be associated with local initial specialisation in resource transformation, but not with resource extraction. However, the absence of a generalisable relationship does not preclude the existence—in given communities under given conditions—of negative side effects resulting from resource specialisation. The findings for Canada suggest that the mechanics are more complex than straightforward Dutch Disease crowding out wage effects.

For both resource sectors, the findings fail to reveal a positive, durable relationship with rising wages, suggesting that upward pressures are generally short lived. Statistically significant relationships with wages appear only for short-run models. The explanation, we suggest, lies on the reactivity of wages. Wages fall and labour emigrates during periods of declining demand. Local economies are different from national economies in their ability to adjust via migration. Boom and bust cycles, specifically for resource extraction, also help to explain the absence of a long-term relationship not only with wages, but also with other growth variables.

The different outcomes for resource extraction and resource transformation bring out the importance of industry-specific and local attributes. In the case at hand, the results reflect the particular resource mix and geography of Canadian communities. The industry specificity of outcomes goes some way in explaining why different authors looking at different places observe different results. Coal mining produces a different work environment from fishing or forestry, a point brought home by James and Aaland (2011) in their comparison of the Wyoming and Maine economies. The Australian literature on local resource economics implicitly places the onus on mining (Goodman and Worth 2008). The results for Canada, where resource transformation (mainly paper mills and aluminum and other primary metal production) is the apparent culprit, suggest that size, sunk costs, and the period over which investments need to be amortised are factors in explaining negative growth impacts over extended time periods. Aluminium plants are typically not only large and highly capitalised, paying high wages for middle-skill jobs, but also imply a certain permanency, different from smaller scale foraging or exploration activities founded on a more transient labour force.

In short, there is no reason to assume that all resource industries produce similar results. Deller and Schreiber (2012) make very much the same point, noting that even within the bounds of the mining industry outcomes vary with the nature of mining. Sand frack mining, currently on the rise in U.S., e.g., relies on a generally more transient and higher paid labour force than more traditional mining.

Location also matters. High wages may be a necessary condition for attracting labour to remote locations in which case crowding out cannot be invoked as an explanation for slower growth, also implying a more mobile labour force. An examination of slow-growing/declining communities with high levels of resource specialisation (in one or both industries) revealed that all were located in fairly remote, generally northern, locations. All also exhibited above-average wages, although generally falling, and below-average college-educated populations. Would other industries have emerged in the absence of local resources? The answer is by no means clear. “Curse” can only be said to apply if resource specialisation prevented other activities with comparable growth prospects from developing.

Boom and bust swings associated with resource extraction are also in part industry specific with oil and gas the clearest example in Canada. Communities in oil-rich Alberta showed the clearest signs of booms and busts with growth coefficients reversing between periods. It can be argued the uncertainty generated by industries prone to violent demand swings creates a disincentive for long-term investment in non-resource sectors. But our results do not allow us to make such an inference. Over the (long-run) 35-year period examined, growth variable coefficients are systematically positive for Alberta-based communities.

To conclude, the findings point to the absence of a generalisable Resource Curse valid for all resource industries and for all communities. A corollary of the above is that the possible solutions for mitigating the negative side effects of resource-led growth will vary between communities. The challenges facing a remote community whose fate has been tied for the last 30 years to a single large paper mill providing well-paying jobs for an established blue-collar population are not the same as those of a centrally located community with a number of oil rigs at various stages of development, staffed largely by transient engineers, geologists, and drillers.

Notes
  1. 1

    At the time of writing, data for sub-provincial spatial units were not yet available for the 2011 census.

  2. 2

    Kitimat, BC, and Kapuskasing in Northern Ontario are exceptions with populations, respectively, of 8,950 and 8,350 in 2006. Both are in the historical data set as their populations were above 10,000 in 1971.

  3. 3

    We assume that the reader is familiar with the location quotient. In a nutshell, the quotient is the ratio of the % of employment in industry × locally over the % of employment in industry × at the system level.

  4. 4

    A sensitivity analysis was undertaken in which LQs were replaced by employment shares (the numerator). The results affected neither the direction nor significance of coefficient results in the regressions presented below. LQs are simply linear transformations of employment shares.

  5. 5

    An alternative method, suggested by an anonymous reviewer, for controlling for industry mix and demand swings is to introduce the industry mix component from shift-share analysis. This was done both with and without provincial fixed effects, but failed to produce significant coefficients.

  6. 6

    Note that the province of Ontario is used as reference. Thus, values related to the fixed effect must be interpreted as the difference relative to the reference province.

  7. 7

    Note that endogeneity is not a major concern since data are macro observations (mean wages) based on initial values and the models are estimated over medium- to long-run time periods.

  8. 8

    We implicitly assume that the transformation occurring between each 5-year time period is linear.

  9. 9

    A trade dispute with U.S. around softwood lumber exports further dampened demand.

  10. 10

    Model 1 is kept for comparison purposes. Model 2 controls for possible biases related to the omission of variable expressing initial conditions.

  11. 11

    The system average was 22 percent. The next two lowest growth urban areas in the high-wage class grew, respectively, by 8.7 and 14.5 percent.

  12. 12

    The opposite “normal” scenario—high wages and high growth—reveals a very different picture for 1991–2006. Twelve urban areas exhibited above-average growth (<22 percent), of which only four with resource LQs above 3.0, including Calgary, the heart of Canada's oil and gas industry.

  13. 13

    Community details referred to in this section are based on numerous sources, too varied to list, including our personal knowledge of certain communities.

References

  1. Top of page
  2. Abstract
  3. Local Economies, Resource Curse, and the Intrusive Rentier Hypothesis
  4. Data and Methodology
  5. Results
  6. In Search of the Intrusive Rentier: A Brief Look at Individual Cases
  7. Conclusion
  8. References
  9. Appendix A Resource Industry Definitions with NAICS Codes
  • Apparicio, P., G. Dussault, M. Polèse, and R. Shearmur. 2007. Transport Infrastructures and Local Economic Development. A Study of the Relationship between Continental Accessibility and Employment Growth in Canadian Communities, 1971–2001. Montreal: INRS.
  • Barnes, T.J., and R. Hayter. 1994. Economic restructuring, local development and resource towns: Forest towns in coastal British Columbia. The Canadian Journal of Regional Science 17: 289310.
  • Beine, M., C. Bos, and S. Coulombe. 2009. Does the Canadian Economy Suffer from Dutch Disease? Discussion Paper 2009-06. Center of Research in Economic Analysis, University of Luxembourg.
  • Black, D., T. McKinnish, and S. Sanders. 2005. The economic impact of the coal boom and bust. The Economic Journal 115: 449476.
  • Blomström, M., and A. Kokko. 2007. From Natural Resources to High-Tech Production: The Evolution of Industrial Competiveness in Sweden and Finland. In Lederman and Maloney op cit.
  • Bourne, L.S., C. Brunelle, M. Polèse, and J. Simmons. 2011. Growth and change in the Canadian urban system. In Canadian Urban Regions. Trajectories of Growth and Change, ed. L.S. Bourne et al., Toronto: Oxford University Press.
  • Boyce, J.R., and J.C.H. Emery. 2011. Is a negative correlation between resource abundance and growth sufficient evidence that there is a “resource curse”. Resources Policy 36(1): 113.
  • Bravo-Ortega, C., and J. de Gregorio. 2007. The Relative Richness of the Poor? Natural Resources, Human Capital, and Economic Growth. In Lederman and Maloney op cit.
  • Bulte, E.H., R. Damania et al. 2005. Resource intensity, institutions, and development. World Development 33(7): 10291044.
  • Coulombe, S. 2011. Lagging Behind: Productivity and the Good Fortune of Canadian Provinces. C.D. Howe Institute Commentary 331. C.D. Howe Institute, Toronto.
  • Deller, S.C., and A. Schreiber. 2012. Mining and community economic growth. The Review of Regional Studies 42: 121141.
  • Desmet, K., and M. Fafchamps. 2005. Changes in the spatial concentration of employment across US counties: A sectoral analysis: 1972–2000. Journal of Economic Geography 5: 261284.
  • Freeman, D.G. 2009. The “Resource Curse” and regional US development. Applied Economics Letters 16(5): 527530.
  • Gaudreault, S. 2011. Le syndrome du renier encombrant: Une évaluation de la situation au Canada (The intrusive rentier syndrome: An evaluation for Canada). Master's Thesis, Department of Economics, University of Quebec at Montreal.
  • Goodman, J., and D. Worth. 2008. The minerals boom and Australia's “Resource Curs”. Journal of Australian Political Economy 61: 201219.
  • Gylfason, T. 2001. Natural resources, education, and economic development. European Economic Review 45(4–6): 847859.
  • Hajkowics, S., S. Heyenga, and K. Moffat. 2011. The relationship between mining and socio-economic well-being in Australia's regions. Resources Policy 36: 3038.
  • Hayter, R. 2000. Single industry resource towns. In A Companion to Economic Geography, ed. E. Sheppard, and T.J. Barnes, 290308. Oxford: Blackwell.
  • Henderson, V. 1997. Medium sized cities. Regional Science and Urban Economics 27: 583612.
  • James, A., and D. Aaland. 2011. The curse of natural resources: An empirical investigation of U.S. counties. Resources and Energy Economics 33: 440453.
  • Krugman, P. 1987. The narrow moving band, the Dutch Disease, and the competitive consequences of Mrs. Thatcher. Notes on trade in the presence of dynamic scale economies. Journal of Development Economics 27(1–2): 4155.
  • Lederman, D., and W.F. Maloney, eds. 2007. Natural Resources-Neither Curse nor Destiny. Washington, DC, and Palo Alto: Stanford University Press/World Bank.
  • Lucas, R.A. 1971. Minetown, Milltown, Railtown. Life in Canadian Communities of Single Industry. Toronto: Toronto University Press.
  • Maloney, W.F. 2007. Missed Opportunities: Innovation and Resource-Based Growth in Latin America. In Lederman and Maloney op cit.
  • Neumayer, E. 2004. Does the “resource curse” hold for growth in genuine income as well? World Development 32(10): 16271640.
  • Polèse, M. 2009. The intrusive rentier syndrome. In The Wealth and Poverty of Regions. Why Cities Matter, ed. M. Polèse, 1823. Chicago: The University of Chicago Press.
  • Polèse, M., and R. Shearmur. 2002. The Periphery in the Knowledge Economy: The Spatial Dynamics of the Canadian Economy and the Future of Non-Metropolitan Regions in Quebec and the Atlantic Provinces. Montreal & Moncton: INRS/Canadian Institute for Research on Regional Development. http://www.ucs.inrs.ca/ucs/publications/collections/collection-regions-et-economie-du-savoir (accessed July 2014).
  • Polèse, M., and R. Shearmur. 2006a. Why some regions will decline: A Canadian case study with thoughts on local economic development. Papers in Regional Science 85(1): 2346.
  • Polèse, M., and R. Shearmur. 2006b. Growth and location of economic activity: The spatial dynamics of industries in Canada 1971–2001. Growth and Change 37(3): 362395.
  • Ross, M.L. 1999. The political economy of the resource curse. World Politics 51: 297322.
  • Sachs, J.D., and A.M. Warner. 2001. The curse of natural resources. European Economic Review 45(4–6): 827838.
  • Slack, E., L.S. Bourne, and M. Gertler. 2003. Small, rural, and remote communities: The anatomy of risk. Research Paper Series, RP (18), Panel on the Role of Government, Government of Ontario, Toronto.
  • The Economist. 1997. The Dutch Disease. The Economist, November 26: 8283.
  • Zellner, A. 1962. An efficient method of estimating seemingly unrelated regressions and tests of aggregation bias. Journal of the American Statistical Association 57(298): 348368.

Appendix A Resource Industry Definitions with NAICS Codes

  1. Top of page
  2. Abstract
  3. Local Economies, Resource Curse, and the Intrusive Rentier Hypothesis
  4. Data and Methodology
  5. Results
  6. In Search of the Intrusive Rentier: A Brief Look at Individual Cases
  7. Conclusion
  8. References
  9. Appendix A Resource Industry Definitions with NAICS Codes
Extractive industries
113 Forestry and logging
1153 Support activities for forestry
2122 Metal ore mining
2121 Coal mining
211 Oil and gas extraction
2123 Non-metallic mineral mining and quarrying
213 Support activities for mining and oil and gas extraction
219 Mining—unspecified
Primary resource transformation manufacturing
321 Wood product manufacturing
322 Paper manufacturing
331 Primary metal manufacturing